diff --git a/.agents/skills/e2e-cucumber-playwright/SKILL.md b/.agents/skills/e2e-cucumber-playwright/SKILL.md
index 4596f9062ea..75e79ea2e7f 100644
--- a/.agents/skills/e2e-cucumber-playwright/SKILL.md
+++ b/.agents/skills/e2e-cucumber-playwright/SKILL.md
@@ -45,6 +45,7 @@ Keep this skill focused on Cucumber, Playwright, and package-level E2E guidance.
    - Inspect the target feature area.
    - Reuse an existing step when wording and behavior already match.
    - Add a new step only for a genuinely new user action or assertion.
+   - Before adding several similar steps, scan the target capability for an existing domain noun that can be parameterized without hiding behavior.
    - Keep edits close to the current capability folder unless the step is broadly reusable.
 2. Write behavior-first scenarios.
    - Describe user-observable behavior, not DOM mechanics.
@@ -53,12 +54,16 @@ Keep this skill focused on Cucumber, Playwright, and package-level E2E guidance.
 3. Write step definitions in the local style.
    - Keep one step to one user-visible action or one assertion.
    - Prefer Cucumber Expressions such as `{string}` and `{int}`.
+   - Use a bounded regex only when the accepted values are a small explicit domain set and Cucumber Expressions would make the Gherkin less natural.
+   - Do not create one-off steps for each case variant when the same domain action or outcome applies to named surfaces, modes, or resources.
    - Scope locators to stable containers when the page has repeated elements.
    - Avoid page-object layers or extra helper abstractions unless repeated complexity clearly justifies them.
 4. Use Playwright in the local style.
    - Prefer user-facing locators: `getByRole`, `getByLabel`, `getByPlaceholder`, `getByText`, then `getByTestId` for explicit contracts.
    - Use web-first `expect(...)` assertions.
    - Do not use `waitForTimeout`, manual polling, or raw visibility checks when a locator action or retrying assertion already expresses the behavior.
+   - Use `expect.poll` for API persistence, backend eventual consistency, captured browser events, or other non-DOM state; prefer locator assertions for DOM readiness and visible UI state.
+   - If a product element has real user-facing semantics but no accessible name, prefer fixing that accessible contract over adding a test id.
 5. Validate narrowly.
    - Run the narrowest tagged scenario or flow that exercises the change.
    - Run `vpr lint --fix --quiet` from the repository root and `pnpm -C e2e type-check`.
diff --git a/.agents/skills/e2e-cucumber-playwright/references/cucumber-best-practices.md b/.agents/skills/e2e-cucumber-playwright/references/cucumber-best-practices.md
index d7a1a52852b..06177faa5c7 100644
--- a/.agents/skills/e2e-cucumber-playwright/references/cucumber-best-practices.md
+++ b/.agents/skills/e2e-cucumber-playwright/references/cucumber-best-practices.md
@@ -39,6 +39,7 @@ Prefer reuse when:
 - the user action is genuinely the same
 - the expected outcome is genuinely the same
 - the wording stays natural across features
+- the parameter is a real product domain value such as a named surface, mode, resource, or status
 
 Write a new step when:
 
@@ -46,6 +47,8 @@ Write a new step when:
 - reusing the old wording would make the scenario misleading
 - a supposedly generic step would become an implementation-detail wrapper
 
+Do not optimize for a low step count by making vague steps. Optimize for a small set of truthful, domain-owned steps.
+
 ### 4. Prefer Cucumber Expressions
 
 Use Cucumber Expressions for parameters unless regex is clearly necessary.
@@ -59,6 +62,8 @@ Common examples:
 
 Keep expressions readable. If a step needs complicated parsing logic, first ask whether the scenario wording should be simpler.
 
+Use regex for a bounded natural-language alternative only when it keeps Gherkin readable, for example `/(Web app|Backend service API)/`. Avoid broad regexes that accept unowned language.
+
 ### 5. Keep step definitions thin and meaningful
 
 Step definitions are glue between Gherkin and automation, not a second abstraction language.
diff --git a/.agents/skills/e2e-cucumber-playwright/references/playwright-best-practices.md b/.agents/skills/e2e-cucumber-playwright/references/playwright-best-practices.md
index 02e763d46b6..deb29e72430 100644
--- a/.agents/skills/e2e-cucumber-playwright/references/playwright-best-practices.md
+++ b/.agents/skills/e2e-cucumber-playwright/references/playwright-best-practices.md
@@ -41,6 +41,7 @@ Also remember:
 - repeated content usually needs scoping to a stable container
 - exact text matching is often too brittle when role/name or label already exists
 - `getByTestId` is acceptable when semantics are weak but the contract is intentional
+- when a real UI region, card, status, or icon lacks an accessible name, prefer adding that semantic contract in product code before falling back to `getByTestId`
 
 ### 3. Use web-first assertions
 
@@ -62,6 +63,8 @@ Avoid:
 
 If a condition genuinely needs custom retry logic, use Playwright's polling/assertion tools deliberately and keep that choice local and explicit.
 
+Use `expect.poll` for non-DOM truth such as API state, backend eventual consistency, generated resources, or captured browser events. For DOM state, use locator assertions so Playwright can apply actionability and web-first retry semantics.
+
 ### 4. Let actions wait for actionability
 
 Locator actions already wait for the element to be actionable. Do not preface every click/fill with extra timing logic unless the action needs a specific visible/ready assertion for clarity.
diff --git a/.github/workflows/main-ci.yml b/.github/workflows/main-ci.yml
index 69c540c54b4..b04b135ca7d 100644
--- a/.github/workflows/main-ci.yml
+++ b/.github/workflows/main-ci.yml
@@ -44,6 +44,7 @@ jobs:
       api-changed: ${{ steps.changes.outputs.api }}
       cli-changed: ${{ steps.changes.outputs.cli }}
       e2e-changed: ${{ steps.changes.outputs.e2e }}
+      external-e2e-changed: ${{ steps.changes.outputs.external_e2e }}
       web-changed: ${{ steps.changes.outputs.web }}
       vdb-changed: ${{ steps.changes.outputs.vdb }}
       migration-changed: ${{ steps.changes.outputs.migration }}
@@ -99,6 +100,25 @@ jobs:
               - 'docker/envs/middleware.env.example'
               - '.github/workflows/web-e2e.yml'
               - '.github/actions/setup-web/**'
+            external_e2e:
+              - 'e2e/features/agent-v2/**'
+              - 'e2e/scripts/**'
+              - 'e2e/support/**'
+              - '.github/workflows/web-e2e.yml'
+              - '.github/actions/setup-web/**'
+              - 'dify-agent/**'
+              - 'api/clients/agent_backend/**'
+              - 'api/core/app/apps/agent_app/**'
+              - 'api/core/workflow/nodes/agent_v2/**'
+              - 'api/controllers/console/agent/**'
+              - 'api/services/agent/**'
+              - 'api/core/plugin/**'
+              - 'api/services/plugin/**'
+              - 'api/core/tools/**'
+              - 'api/services/tools/**'
+              - 'web/features/agent-v2/**'
+              - 'web/app/(commonLayout)/roster/**'
+              - 'web/app/components/workflow/nodes/agent-v2/**'
             vdb:
               - 'api/core/rag/datasource/**'
               - 'api/tests/integration_tests/vdb/**'
@@ -322,15 +342,18 @@ jobs:
     needs:
       - pre_job
       - check-changes
-    if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.e2e-changed == 'true'
+    if: needs.pre_job.outputs.should_skip != 'true' && (needs.check-changes.outputs.e2e-changed == 'true' || needs.check-changes.outputs.external-e2e-changed == 'true')
     uses: ./.github/workflows/web-e2e.yml
+    with:
+      run-external-runtime: ${{ github.event_name == 'push' && github.ref == 'refs/heads/main' && needs.check-changes.outputs.external-e2e-changed == 'true' }}
+    secrets: inherit
 
   web-e2e-skip:
     name: Skip Web Full-Stack E2E
     needs:
       - pre_job
       - check-changes
-    if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.e2e-changed != 'true'
+    if: needs.pre_job.outputs.should_skip != 'true' && needs.check-changes.outputs.e2e-changed != 'true' && needs.check-changes.outputs.external-e2e-changed != 'true'
     runs-on: depot-ubuntu-24.04
     steps:
       - name: Report skipped web full-stack e2e
@@ -349,7 +372,7 @@ jobs:
       - name: Finalize Web Full-Stack E2E status
         env:
           SHOULD_SKIP_WORKFLOW: ${{ needs.pre_job.outputs.should_skip }}
-          TESTS_CHANGED: ${{ needs.check-changes.outputs.e2e-changed }}
+          TESTS_CHANGED: ${{ needs.check-changes.outputs.e2e-changed == 'true' || needs.check-changes.outputs.external-e2e-changed == 'true' }}
           RUN_RESULT: ${{ needs.web-e2e-run.result }}
           SKIP_RESULT: ${{ needs.web-e2e-skip.result }}
         run: |
diff --git a/.github/workflows/web-e2e.yml b/.github/workflows/web-e2e.yml
index 2447e9033a3..d50cce65d46 100644
--- a/.github/workflows/web-e2e.yml
+++ b/.github/workflows/web-e2e.yml
@@ -2,6 +2,11 @@ name: Web Full-Stack E2E
 
 on:
   workflow_call:
+    inputs:
+      run-external-runtime:
+        required: false
+        type: boolean
+        default: false
 
 permissions:
   contents: read
@@ -32,7 +37,9 @@ jobs:
         with:
           enable-cache: true
           python-version: "3.12"
-          cache-dependency-glob: api/uv.lock
+          cache-dependency-glob: |
+            api/uv.lock
+            dify-agent/uv.lock
 
       - name: Install API dependencies
         run: uv sync --project api --dev
@@ -51,12 +58,52 @@ jobs:
           E2E_INIT_PASSWORD: E2eInit12345
         run: vp run e2e:full
 
+      - name: Run external runtime E2E tests
+        if: ${{ inputs.run-external-runtime }}
+        working-directory: ./e2e
+        env:
+          E2E_ADMIN_EMAIL: e2e-admin@example.com
+          E2E_ADMIN_NAME: E2E Admin
+          E2E_ADMIN_PASSWORD: E2eAdmin12345
+          E2E_AGENT_DECISION_MODEL_NAME: ${{ vars.E2E_AGENT_DECISION_MODEL_NAME || 'gpt-5.5' }}
+          E2E_AGENT_DECISION_MODEL_PROVIDER: ${{ vars.E2E_AGENT_DECISION_MODEL_PROVIDER || 'openai' }}
+          E2E_AGENT_DECISION_MODEL_TYPE: ${{ vars.E2E_AGENT_DECISION_MODEL_TYPE || 'llm' }}
+          E2E_EXTERNAL_RUNTIME_SEED_SPECS: ${{ vars.E2E_EXTERNAL_RUNTIME_SEED_SPECS }}
+          E2E_EXTERNAL_RUNTIME_TAGS: ${{ vars.E2E_EXTERNAL_RUNTIME_TAGS }}
+          E2E_FORCE_WEB_BUILD: "1"
+          E2E_INIT_PASSWORD: E2eInit12345
+          E2E_MARKETPLACE_API_URL: ${{ vars.E2E_MARKETPLACE_API_URL }}
+          E2E_MARKETPLACE_PLUGIN_IDS: ${{ vars.E2E_MARKETPLACE_PLUGIN_IDS }}
+          E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS: ${{ vars.E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS }}
+          E2E_MODEL_PROVIDER_CREDENTIALS_JSON: ${{ secrets.E2E_MODEL_PROVIDER_CREDENTIALS_JSON }}
+          E2E_START_AGENT_BACKEND: ${{ vars.E2E_START_AGENT_BACKEND || '1' }}
+          E2E_STABLE_MODEL_NAME: ${{ vars.E2E_STABLE_MODEL_NAME || 'gpt-5-nano' }}
+          E2E_STABLE_MODEL_PROVIDER: ${{ vars.E2E_STABLE_MODEL_PROVIDER || 'openai' }}
+          E2E_STABLE_MODEL_TYPE: ${{ vars.E2E_STABLE_MODEL_TYPE || 'llm' }}
+        run: |
+          if [[ -z "${E2E_MODEL_PROVIDER_CREDENTIALS_JSON}" ]]; then
+            echo "E2E_MODEL_PROVIDER_CREDENTIALS_JSON is required for external runtime E2E." >&2
+            exit 1
+          fi
+
+          if [[ -d cucumber-report ]]; then
+            rm -rf cucumber-report-non-external
+            mv cucumber-report cucumber-report-non-external
+          fi
+
+          trap 'vp run e2e:middleware:down' EXIT
+          vp run e2e:middleware:up
+          vp run e2e:external:prepare
+          vp run e2e:external
+
       - name: Upload Cucumber report
         if: ${{ !cancelled() }}
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         with:
           name: cucumber-report
-          path: e2e/cucumber-report
+          path: |
+            e2e/cucumber-report
+            e2e/cucumber-report-non-external
           retention-days: 7
 
       - name: Upload E2E logs
diff --git a/api/clients/agent_backend/event_adapter.py b/api/clients/agent_backend/event_adapter.py
index e983aa336d7..54c63f15e46 100644
--- a/api/clients/agent_backend/event_adapter.py
+++ b/api/clients/agent_backend/event_adapter.py
@@ -67,6 +67,7 @@ class AgentBackendRunSucceededInternalEvent(AgentBackendInternalEventBase):
     type: Literal[AgentBackendInternalEventType.RUN_SUCCEEDED] = AgentBackendInternalEventType.RUN_SUCCEEDED
     output: JsonValue
     session_snapshot: CompositorSessionSnapshot
+    usage: dict[str, JsonValue] | None = None
 
 
 class AgentBackendDeferredToolCallInternalEvent(AgentBackendInternalEventBase):
@@ -76,6 +77,7 @@ class AgentBackendDeferredToolCallInternalEvent(AgentBackendInternalEventBase):
     deferred_tool_call: DeferredToolCallPayload
     message: str | None = None
     session_snapshot: CompositorSessionSnapshot
+    usage: dict[str, JsonValue] | None = None
 
 
 class AgentBackendRunFailedInternalEvent(AgentBackendInternalEventBase):
@@ -140,6 +142,7 @@ class AgentBackendRunEventAdapter:
                             deferred_tool_call=event.data.deferred_tool_call,
                             message=_deferred_tool_call_message(event.data.deferred_tool_call),
                             session_snapshot=event.data.session_snapshot,
+                            usage=_agent_run_usage(event.data.usage),
                         )
                     ]
                 return [
@@ -148,6 +151,7 @@ class AgentBackendRunEventAdapter:
                         source_event_id=event.id,
                         output=event.data.output,
                         session_snapshot=event.data.session_snapshot,
+                        usage=_agent_run_usage(event.data.usage),
                     )
                 ]
             case RunFailedEvent():
@@ -184,3 +188,13 @@ def _deferred_tool_call_message(payload: DeferredToolCallPayload) -> str:
             return title
 
     return f"Agent backend requested external input via deferred tool '{payload.tool_name}'."
+
+
+def _agent_run_usage(usage: object | None) -> dict[str, JsonValue] | None:
+    """Return JSON-safe usage metadata from optional Agent backend usage."""
+    if usage is None:
+        return None
+    dumped = _EVENT_DATA_ADAPTER.dump_python(usage, mode="json")
+    if not isinstance(dumped, dict):
+        return None
+    return cast(dict[str, JsonValue], dumped)
diff --git a/api/clients/agent_backend/request_builder.py b/api/clients/agent_backend/request_builder.py
index 0ed8af042dd..3b7b78f299b 100644
--- a/api/clients/agent_backend/request_builder.py
+++ b/api/clients/agent_backend/request_builder.py
@@ -171,7 +171,7 @@ class AgentBackendWorkflowNodeRunInput(BaseModel):
     knowledge: DifyKnowledgeBaseLayerConfig | None = None
     config_layer_config: DifyConfigLayerConfig | None = None
     # Drive Skills & Files declaration (dify.drive) — an index the agent pulls
-    # through the back proxy, never inline content; see AGENT_DRIVE_MANIFEST_ENABLED.
+    # through the back proxy, never inline content.
     drive_config: DifyDriveLayerConfig | None = None
     # Human-in-the-loop ask_human deferred tool (dify.ask_human). Present only when
     # the Agent Soul configures human involvement; a deferred call ends the run and
@@ -220,7 +220,7 @@ class AgentBackendAgentAppRunInput(BaseModel):
     knowledge: DifyKnowledgeBaseLayerConfig | None = None
     config_layer_config: DifyConfigLayerConfig | None = None
     # Drive Skills & Files declaration (dify.drive) — an index the agent pulls
-    # through the back proxy, never inline content; see AGENT_DRIVE_MANIFEST_ENABLED.
+    # through the back proxy, never inline content.
     drive_config: DifyDriveLayerConfig | None = None
     # Human-in-the-loop ask_human deferred tool (dify.ask_human). Present only when
     # the Agent Soul configures human involvement (ENG-635).
@@ -254,10 +254,11 @@ class AgentBackendRunRequestBuilder:
         """Build an Agent App conversation-turn run request.
 
         Layer graph: optional Agent Soul system prompt → user prompt →
-        execution context → optional history (multi-turn) → LLM → optional
-        plugin-direct tools / core-routed tools / knowledge search →
-        optional structured output. Mirrors the workflow-node layer ordering
-        minus the workflow-job / previous-node prompt.
+        execution context → optional shell / config / drive / history
+        (multi-turn) → LLM → optional plugin-direct tools / core-routed tools /
+        knowledge search / ask_human / structured output. Mirrors the
+        workflow-node layer ordering minus the workflow-job / previous-node
+        prompt.
         """
         layers: list[RunLayerSpec] = []
         if run_input.agent_soul_prompt:
@@ -354,11 +355,14 @@ class AgentBackendRunRequestBuilder:
         )
 
         if run_input.tools is not None and run_input.tools.tools:
+            plugin_tool_deps = {"execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID}
+            if include_shell:
+                plugin_tool_deps["shell"] = DIFY_SHELL_LAYER_ID
             layers.append(
                 RunLayerSpec(
                     name=DIFY_PLUGIN_TOOLS_LAYER_ID,
                     type=DIFY_PLUGIN_TOOLS_LAYER_TYPE_ID,
-                    deps={"execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID},
+                    deps=plugin_tool_deps,
                     metadata=run_input.metadata,
                     config=run_input.tools,
                 )
@@ -474,9 +478,9 @@ class AgentBackendRunRequestBuilder:
         """Build a workflow Agent Node run request without defining another wire schema.
 
         Layer graph mirrors the workflow surface: prompts → execution context →
-        optional drive/history → LLM → optional plugin-direct tools /
-        core-routed tools / knowledge search → optional auxiliary layers such
-        as ask_human, shell, and structured output.
+        optional shell / config / drive / history → LLM → optional
+        plugin-direct tools / core-routed tools / knowledge search /
+        ask_human / structured output.
         """
         layers: list[RunLayerSpec] = []
         if run_input.agent_soul_prompt:
@@ -581,11 +585,14 @@ class AgentBackendRunRequestBuilder:
         )
 
         if run_input.tools is not None and run_input.tools.tools:
+            plugin_tool_deps = {"execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID}
+            if include_shell:
+                plugin_tool_deps["shell"] = DIFY_SHELL_LAYER_ID
             layers.append(
                 RunLayerSpec(
                     name=DIFY_PLUGIN_TOOLS_LAYER_ID,
                     type=DIFY_PLUGIN_TOOLS_LAYER_TYPE_ID,
-                    deps={"execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID},
+                    deps=plugin_tool_deps,
                     metadata=run_input.metadata,
                     config=run_input.tools,
                 )
diff --git a/api/configs/extra/agent_backend_config.py b/api/configs/extra/agent_backend_config.py
index 58228cbfb02..b59350d5f0d 100644
--- a/api/configs/extra/agent_backend_config.py
+++ b/api/configs/extra/agent_backend_config.py
@@ -1,4 +1,4 @@
-from pydantic import Field
+from pydantic import Field, NonNegativeFloat
 from pydantic_settings import BaseSettings
 
 
@@ -32,12 +32,10 @@ class AgentBackendConfig(BaseSettings):
         default=False,
     )
 
-    AGENT_DRIVE_MANIFEST_ENABLED: bool = Field(
+    AGENT_APP_TEXT_DELTA_DEBOUNCE_SECONDS: NonNegativeFloat = Field(
         description=(
-            "Inject the dify.drive layer (Skills & Files drive manifest declaration) "
-            "into Agent runs. The declaration is an index only — the agent backend "
-            "pulls the actual SKILL.md / files through the back proxy. Set this to "
-            "false only when temporarily rolling back the drive integration."
+            "Buffer Agent App assistant text deltas for up to this many seconds before "
+            "publishing SSE chunks. Set to 0 to publish each delta immediately."
         ),
-        default=True,
+        default=0.5,
     )
diff --git a/api/controllers/console/agent/roster.py b/api/controllers/console/agent/roster.py
index 377c40877e5..349826e54d7 100644
--- a/api/controllers/console/agent/roster.py
+++ b/api/controllers/console/agent/roster.py
@@ -91,33 +91,31 @@ class AgentIdPath(BaseModel):
 class AgentAppCreatePayload(BaseModel):
     name: str = Field(..., min_length=1, description="Agent name")
     description: str | None = Field(default=None, description="Agent description (max 400 chars)", max_length=400)
-    role: str = Field(..., min_length=1, description="Agent role", max_length=255)
+    role: str | None = Field(default=None, description="Agent role", max_length=255)
     icon_type: IconType | None = Field(default=None, description="Icon type")
     icon: str | None = Field(default=None, description="Icon")
     icon_background: str | None = Field(default=None, description="Icon background color")
 
     @field_validator("role")
     @classmethod
-    def validate_role(cls, value: str) -> str:
-        role = value.strip()
-        if not role:
-            raise ValueError("Agent role is required.")
-        return role
+    def validate_role(cls, value: str | None) -> str | None:
+        if value is None:
+            return None
+        return value.strip()
 
 
 # Keep agent-app roster DTOs agent-specific instead of reusing the shared
 # /apps response/request models. The roster surface needs Agent-only fields such
 # as `role`, while the generic console/apps contracts must stay unchanged.
 class AgentAppUpdatePayload(GenericUpdateAppPayload):
-    role: str = Field(..., min_length=1, description="Agent role", max_length=255)
+    role: str | None = Field(default=None, description="Agent role", max_length=255)
 
     @field_validator("role")
     @classmethod
-    def validate_role(cls, value: str) -> str:
-        role = value.strip()
-        if not role:
-            raise ValueError("Agent role is required.")
-        return role
+    def validate_role(cls, value: str | None) -> str | None:
+        if value is None:
+            return None
+        return value.strip()
 
 
 class AgentAppCopyPayload(BaseModel):
@@ -133,10 +131,7 @@ class AgentAppCopyPayload(BaseModel):
     def validate_role(cls, value: str | None) -> str | None:
         if value is None:
             return None
-        role = value.strip()
-        if not role:
-            raise ValueError("Agent role is required when provided.")
-        return role
+        return value.strip()
 
 
 class AgentApiStatusPayload(BaseModel):
@@ -531,6 +526,7 @@ class AgentAppListApi(Resource):
             page=args.page,
             limit=args.limit,
             mode="agent",
+            sort_by=args.sort_by,
             name=args.name,
             tag_ids=args.tag_ids,
             creator_ids=args.creator_ids,
@@ -565,7 +561,7 @@ class AgentAppListApi(Resource):
             name=args.name,
             description=args.description,
             mode="agent",
-            agent_role=args.role,
+            agent_role=args.role or "",
             icon_type=args.icon_type,
             icon=args.icon,
             icon_background=args.icon_background,
diff --git a/api/controllers/console/app/completion.py b/api/controllers/console/app/completion.py
index c408569703a..1b54f9f9c65 100644
--- a/api/controllers/console/app/completion.py
+++ b/api/controllers/console/app/completion.py
@@ -1,7 +1,7 @@
 import json
 import logging
-from collections.abc import Generator
-from typing import Any, Literal
+from collections.abc import Generator, Iterator, Mapping
+from typing import Any, Literal, Protocol, runtime_checkable
 from uuid import UUID
 
 from flask import request
@@ -60,6 +60,11 @@ from services.errors.llm import InvokeRateLimitError
 logger = logging.getLogger(__name__)
 
 
+@runtime_checkable
+class _ClosableStream(Protocol):
+    def close(self) -> None: ...
+
+
 def _resolve_debugger_chat_streaming(
     *, app_mode: AppMode, response_mode: str, response_mode_provided: bool = True
 ) -> bool:
@@ -118,15 +123,12 @@ edit workspace files, run validation or debugging commands, make exploratory che
 Use only the current Build chat message history to identify changes that need to be persisted. Do not inspect, test, or
 validate old config unless the message history already shows that the old config is invalid.
 
-Persist only the build-draft config resources that need to change, using the Agent config CLI usage provided in the
-runtime prompt:
+Only update the build-draft config note when the current Build chat contains durable context that later runs need.
+Do not create, update, delete, inspect, or fill gaps in other Agent config resources, including config files, config
+skills, config env, tools, models, knowledge, or prompt settings.
 
-- config files for reusable artifacts that should be available later,
-- config skills for reusable procedures or tools that should be available later,
-- config env when environment keys or values need to be recorded,
-- config note for concise durable context when useful.
-
-When updating the config note, record only durable context needed by later runs, such as:
+When updating the config note with the Agent config CLI usage provided in the runtime prompt, record only durable
+context needed by later runs, such as:
 
 - what you installed or configured outside the workspace for this agent,
 - where those external updates live, including CLI tools, packages, and persistent $HOME paths,
@@ -437,7 +439,6 @@ def _drain_streaming_generate_response(response: RateLimitGenerator | Generator[
     changes the HTTP boundary: it drains the SSE stream server-side and returns
     success after the generated build-chat message reaches ``message_end``.
     """
-    close = getattr(response, "close", None)
     try:
         for chunk in response:
             for raw_event in chunk.split("\n\n"):
@@ -467,8 +468,8 @@ def _drain_streaming_generate_response(response: RateLimitGenerator | Generator[
                 if payload_event == "error":
                     raise CompletionRequestError(str(payload.get("message") or "Build chat finalization failed."))
     finally:
-        if callable(close):
-            close()
+        if isinstance(response, _ClosableStream):
+            response.close()
 
     raise CompletionRequestError("Build chat finalization did not complete.")
 
@@ -531,6 +532,8 @@ def _generate_chat_message_response(
         args=args,
         streaming=streaming,
     )
+    if AppMode.value_of(app_model.mode) == AppMode.AGENT and streaming:
+        response = _raise_agent_stream_error_before_response(response)
     return helper.compact_generate_response(response)
 
 
@@ -543,3 +546,67 @@ def _stop_chat_message(*, current_user_id: str, app_model: App, task_id: str):
     )
 
     return SimpleResultResponse(result="success").model_dump(mode="json"), 200
+
+
+def _raise_agent_stream_error_before_response(response):
+    """Surface immediate Agent App stream errors as HTTP errors before SSE starts.
+
+    The shared streaming helper always returns HTTP 200 once the SSE response is
+    created. Agent v2 configuration errors, such as an invalid model API key,
+    can be the first real stream event after the initial ping; pre-reading that
+    first non-ping event lets the console API return the existing 400 error
+    contract instead of a successful HTTP response carrying only an SSE error.
+    """
+    if isinstance(response, Mapping):
+        return response
+
+    buffered: list[str] = []
+    iterator = iter(response)
+    while True:
+        try:
+            chunk = next(iterator)
+        except StopIteration:
+            return iter(buffered)
+
+        if not isinstance(chunk, str):
+            return _prepend_stream_chunks(buffered, chunk, iterator)
+
+        if _is_sse_ping(chunk):
+            buffered.append(chunk)
+            continue
+
+        error_payload = _extract_sse_error_payload(chunk)
+        if error_payload is not None:
+            if isinstance(response, _ClosableStream):
+                response.close()
+            message = error_payload.get("message")
+            raise CompletionRequestError(str(message or "Agent App chat failed."))
+
+        return _prepend_stream_chunks(buffered, chunk, iterator)
+
+
+def _prepend_stream_chunks(buffered: list[Any], first: Any, iterator: Iterator[Any]) -> Generator[Any, None, None]:
+    yield from buffered
+    yield first
+    yield from iterator
+
+
+def _is_sse_ping(chunk: str) -> bool:
+    return chunk.strip() == "event: ping"
+
+
+def _extract_sse_error_payload(chunk: str) -> dict[str, Any] | None:
+    for raw_event in chunk.split("\n\n"):
+        data_lines: list[str] = []
+        for line in raw_event.splitlines():
+            if line.startswith("data: "):
+                data_lines.append(line.removeprefix("data: "))
+        if not data_lines:
+            continue
+        try:
+            payload = json.loads("\n".join(data_lines))
+        except json.JSONDecodeError:
+            continue
+        if isinstance(payload, dict) and payload.get("event") == "error":
+            return payload
+    return None
diff --git a/api/controllers/files/upload.py b/api/controllers/files/upload.py
index 661135f2954..249784dee92 100644
--- a/api/controllers/files/upload.py
+++ b/api/controllers/files/upload.py
@@ -28,6 +28,7 @@ class PluginUploadQuery(BaseModel):
     sign: str = Field(..., description="HMAC signature")
     tenant_id: str = Field(..., description="Tenant identifier")
     user_id: str | None = Field(default=None, description="User identifier")
+    conversation_id: str | None = Field(default=None, description="Conversation identifier")
 
 
 register_schema_models(files_ns, PluginUploadQuery)
@@ -92,6 +93,7 @@ class PluginUploadFileApi(Resource):
             mimetype=mimetype,
             tenant_id=tenant_id,
             user_id=user.id,
+            conversation_id=args.conversation_id,
             timestamp=timestamp,
             nonce=nonce,
             sign=sign,
@@ -105,7 +107,7 @@ class PluginUploadFileApi(Resource):
                 file_binary=file.stream.read(),
                 mimetype=mimetype,
                 filename=filename,
-                conversation_id=None,
+                conversation_id=args.conversation_id,
             )
 
             extension = guess_extension(tool_file.mimetype) or ".bin"
diff --git a/api/controllers/inner_api/plugin/plugin.py b/api/controllers/inner_api/plugin/plugin.py
index e1de82c8b68..4e6dbf5dcbe 100644
--- a/api/controllers/inner_api/plugin/plugin.py
+++ b/api/controllers/inner_api/plugin/plugin.py
@@ -434,6 +434,7 @@ class PluginUploadFileRequestApi(Resource):
             mimetype=payload.mimetype,
             tenant_id=tenant_model.id,
             user_id=user_model.id,
+            conversation_id=payload.conversation_id,
         )
         return BaseBackwardsInvocationResponse(data={"url": url}).model_dump()
 
diff --git a/api/core/app/apps/agent_app/app_generator.py b/api/core/app/apps/agent_app/app_generator.py
index 6e9a7a79af5..9d45ac71389 100644
--- a/api/core/app/apps/agent_app/app_generator.py
+++ b/api/core/app/apps/agent_app/app_generator.py
@@ -507,6 +507,7 @@ class AgentAppGenerator(MessageBasedAppGenerator):
             ),
             event_adapter=AgentBackendRunEventAdapter(),
             session_store=AgentAppRuntimeSessionStore(),
+            text_delta_debounce_seconds=dify_config.AGENT_APP_TEXT_DELTA_DEBOUNCE_SECONDS,
         )
 
     def _run_input_guards(
diff --git a/api/core/app/apps/agent_app/app_runner.py b/api/core/app/apps/agent_app/app_runner.py
index d90dad1f891..57dac07f761 100644
--- a/api/core/app/apps/agent_app/app_runner.py
+++ b/api/core/app/apps/agent_app/app_runner.py
@@ -16,6 +16,8 @@ from __future__ import annotations
 
 import json
 import logging
+import time
+from collections.abc import Mapping
 from decimal import Decimal
 from typing import Any, Literal
 
@@ -74,12 +76,23 @@ def _prompt_messages_from_query(user_query: str | None) -> list[PromptMessage]:
     return [UserPromptMessage(content=user_query)]
 
 
+def _llm_usage_from_agent_backend(usage: Mapping[str, Any] | None) -> LLMUsage | None:
+    if usage is None:
+        return None
+    try:
+        return LLMUsage.from_metadata(usage)
+    except (TypeError, ValueError):
+        logger.warning("Failed to parse Agent backend usage metadata: %s", usage, exc_info=True)
+        return None
+
+
 def publish_text_answer(
     *,
     queue_manager: AppQueueManager,
     model_name: str,
     answer: str,
     user_query: str | None = None,
+    usage: LLMUsage | None = None,
 ) -> None:
     """Publish a complete assistant answer as one chunk + message-end.
 
@@ -99,6 +112,7 @@ def publish_text_answer(
         model_name=model_name,
         answer=answer,
         user_query=user_query,
+        usage=usage,
     )
 
 
@@ -127,6 +141,7 @@ def publish_message_end(
     model_name: str,
     answer: str,
     user_query: str | None = None,
+    usage: LLMUsage | None = None,
 ) -> None:
     """Publish the terminal assistant result without emitting another delta."""
     prompt_messages = _prompt_messages_from_query(user_query)
@@ -136,13 +151,46 @@ def publish_message_end(
                 model=model_name,
                 prompt_messages=prompt_messages,
                 message=AssistantPromptMessage(content=answer),
-                usage=LLMUsage.empty_usage(),
+                usage=usage or LLMUsage.empty_usage(),
             ),
         ),
         PublishFrom.APPLICATION_MANAGER,
     )
 
 
+class _TextDeltaDebouncer:
+    """Batch assistant text deltas on stream-event boundaries for final SSE output."""
+
+    def __init__(self, *, debounce_seconds: float) -> None:
+        self._debounce_seconds = debounce_seconds
+        self._parts: list[str] = []
+        self._first_pending_at: float | None = None
+
+    def push(self, delta: str) -> str | None:
+        if not delta:
+            return None
+        if self._debounce_seconds <= 0:
+            return delta
+
+        now = time.monotonic()
+        if not self._parts:
+            self._first_pending_at = now
+        self._parts.append(delta)
+
+        if self._first_pending_at is not None and now - self._first_pending_at >= self._debounce_seconds:
+            return self.flush()
+        return None
+
+    def flush(self) -> str | None:
+        if not self._parts:
+            return None
+
+        text = "".join(self._parts)
+        self._parts = []
+        self._first_pending_at = None
+        return text
+
+
 class _AgentProcessRecorder:
     """Persist Agent v2 thinking/tool process events through the legacy thought model."""
 
@@ -430,11 +478,13 @@ class AgentAppRunner:
         agent_backend_client: AgentBackendRunClient,
         event_adapter: AgentBackendRunEventAdapter,
         session_store: AgentAppRuntimeSessionStore,
+        text_delta_debounce_seconds: float,
     ) -> None:
         self._request_builder = request_builder
         self._agent_backend_client = agent_backend_client
         self._event_adapter = event_adapter
         self._session_store = session_store
+        self._text_delta_debounce_seconds = text_delta_debounce_seconds
 
     def run(
         self,
@@ -512,6 +562,7 @@ class AgentAppRunner:
             answer=answer,
             query=query,
             streamed_answer=streamed_answer,
+            usage=_llm_usage_from_agent_backend(terminal.usage),
         )
         self._save_session(
             scope=scope,
@@ -724,19 +775,39 @@ class AgentAppRunner:
         model_name: str,
         query: str | None,
     ):
+        """Consume backend events while preserving raw recorder granularity.
+
+        Process events are recorded immediately for observability. Only the
+        final assistant text deltas sent through the EasyUI queue are debounced,
+        with flushes happening on later stream events or terminal boundaries.
+        """
         terminal = None
         streamed_answer_parts: list[str] = []
+        text_delta_debouncer = _TextDeltaDebouncer(debounce_seconds=self._text_delta_debounce_seconds)
         process_recorder = _AgentProcessRecorder(
             dify_context=dify_context,
             message_id=message_id,
             queue_manager=queue_manager,
         )
+
+        def flush_pending_text() -> None:
+            pending_text = text_delta_debouncer.flush()
+            if pending_text:
+                publish_text_delta(
+                    queue_manager=queue_manager,
+                    model_name=model_name,
+                    delta=pending_text,
+                    user_query=query,
+                )
+
         for public_event in self._agent_backend_client.stream_events(run_id):
             if queue_manager.is_stopped():
+                flush_pending_text()
                 self._cancel_run(run_id)
                 raise GenerateTaskStoppedError()
             for internal_event in self._event_adapter.adapt(public_event):
                 if queue_manager.is_stopped():
+                    flush_pending_text()
                     self._cancel_run(run_id)
                     raise GenerateTaskStoppedError()
                 if internal_event.type in (
@@ -758,18 +829,22 @@ class AgentAppRunner:
                         text_delta = self._extract_stream_text_delta(internal_event)
                         if text_delta:
                             streamed_answer_parts.append(text_delta)
-                            publish_text_delta(
-                                queue_manager=queue_manager,
-                                model_name=model_name,
-                                delta=text_delta,
-                                user_query=query,
-                            )
+                            debounced_delta = text_delta_debouncer.push(text_delta)
+                            if debounced_delta:
+                                publish_text_delta(
+                                    queue_manager=queue_manager,
+                                    model_name=model_name,
+                                    delta=debounced_delta,
+                                    user_query=query,
+                                )
                         continue
                     continue
+                flush_pending_text()
                 terminal = internal_event
                 break
             if terminal is not None:
                 break
+        flush_pending_text()
         return terminal, "".join(streamed_answer_parts)
 
     def _cancel_run(self, run_id: str) -> None:
@@ -793,10 +868,20 @@ class AgentAppRunner:
         answer: str,
         query: str | None,
         streamed_answer: str,
+        usage: LLMUsage | None,
     ) -> None:
         """Finish a successful streamed turn without duplicating the final text."""
+        if not answer and streamed_answer:
+            answer = streamed_answer
+
         if not streamed_answer:
-            self._publish_answer(queue_manager=queue_manager, model_name=model_name, answer=answer, query=query)
+            publish_text_answer(
+                queue_manager=queue_manager,
+                model_name=model_name,
+                answer=answer,
+                user_query=query,
+                usage=usage,
+            )
             return
 
         if answer.startswith(streamed_answer):
@@ -812,7 +897,13 @@ class AgentAppRunner:
                 "using terminal output for message persistence."
             )
 
-        publish_message_end(queue_manager=queue_manager, model_name=model_name, answer=answer, user_query=query)
+        publish_message_end(
+            queue_manager=queue_manager,
+            model_name=model_name,
+            answer=answer,
+            user_query=query,
+            usage=usage,
+        )
 
     def _save_session(
         self,
@@ -852,6 +943,8 @@ class AgentAppRunner:
         configured the value is a JSON object, which we serialize so the chat
         message always has a string body.
         """
+        if output is None:
+            return ""
         if isinstance(output, str):
             return output
         if isinstance(output, dict):
diff --git a/api/core/app/apps/agent_app/runtime_request_builder.py b/api/core/app/apps/agent_app/runtime_request_builder.py
index 4e850a4887a..fb1d054e750 100644
--- a/api/core/app/apps/agent_app/runtime_request_builder.py
+++ b/api/core/app/apps/agent_app/runtime_request_builder.py
@@ -46,7 +46,7 @@ from core.workflow.nodes.agent_v2.runtime_request_builder import (
 )
 from models.agent_config_entities import AgentSoulConfig, AgentSoulToolsConfig
 from models.provider_ids import ModelProviderID
-from services.agent.prompt_mentions import build_soul_mention_resolver, expand_prompt_mentions
+from services.agent.prompt_mentions import expand_prompt_mentions
 
 
 class AgentAppRuntimeRequestBuildError(ValueError):
@@ -124,17 +124,14 @@ class AgentAppRuntimeRequestBuilder:
                 "cli_tool_count": len(agent_soul.tools.cli_tools),
             }
 
-        config_layer_config = None
-        soul_prompt_resolver = build_soul_mention_resolver(agent_soul)
-        if dify_config.AGENT_DRIVE_MANIFEST_ENABLED:
-            config_layer_config, config_warnings = build_config_layer_config(
-                agent_soul,
-                agent_id=context.agent_id,
-                config_version_id=context.agent_config_snapshot_id,
-                config_version_kind=context.agent_config_version_kind,
-            )
-            append_runtime_warnings(metadata, config_warnings)
-            soul_prompt_resolver = build_config_aware_soul_mention_resolver(agent_soul)
+        config_layer_config, config_warnings = build_config_layer_config(
+            agent_soul,
+            agent_id=context.agent_id,
+            config_version_id=context.agent_config_snapshot_id,
+            config_version_kind=context.agent_config_version_kind,
+        )
+        append_runtime_warnings(metadata, config_warnings)
+        soul_prompt_resolver = build_config_aware_soul_mention_resolver(agent_soul)
         knowledge_config = build_knowledge_layer_config(agent_soul)
 
         request = self._request_builder.build_for_agent_app(
diff --git a/api/core/helper/marketplace.py b/api/core/helper/marketplace.py
index 2c7d520250c..0b77891ce16 100644
--- a/api/core/helper/marketplace.py
+++ b/api/core/helper/marketplace.py
@@ -1,5 +1,6 @@
 import logging
 from collections.abc import Sequence
+from urllib.parse import urlencode
 
 import httpx
 from yarl import URL
@@ -16,7 +17,8 @@ MARKETPLACE_TIMEOUT = 30
 
 
 def get_plugin_pkg_url(plugin_unique_identifier: str) -> str:
-    return str((marketplace_api_url / "api/v1/plugins/download").with_query(unique_identifier=plugin_unique_identifier))
+    query = urlencode({"unique_identifier": plugin_unique_identifier})
+    return f"{marketplace_api_url / 'api/v1/plugins/download'}?{query}"
 
 
 def download_plugin_pkg(plugin_unique_identifier: str):
diff --git a/api/core/plugin/entities/request.py b/api/core/plugin/entities/request.py
index fecad81c032..65f7d044ea2 100644
--- a/api/core/plugin/entities/request.py
+++ b/api/core/plugin/entities/request.py
@@ -230,6 +230,7 @@ class RequestRequestUploadFile(BaseModel):
 
     filename: str
     mimetype: str
+    conversation_id: str | None = None
 
 
 class RequestDownloadFileMapping(BaseModel):
diff --git a/api/core/plugin/plugin_service.py b/api/core/plugin/plugin_service.py
index 805c52329b3..6b306e2df86 100644
--- a/api/core/plugin/plugin_service.py
+++ b/api/core/plugin/plugin_service.py
@@ -902,7 +902,10 @@ class PluginService:
             tenant_id,
             plugin_unique_identifiers,
             PluginInstallationSource.Package,
-            [{}],
+            [
+                {"plugin_unique_identifier": plugin_unique_identifier}
+                for plugin_unique_identifier in plugin_unique_identifiers
+            ],
         )
         PluginService.invalidate_plugin_model_providers_cache(tenant_id)
         return result
diff --git a/api/core/rag/retrieval/dataset_retrieval.py b/api/core/rag/retrieval/dataset_retrieval.py
index 96c6e57e7b0..c1ba9964a3a 100644
--- a/api/core/rag/retrieval/dataset_retrieval.py
+++ b/api/core/rag/retrieval/dataset_retrieval.py
@@ -1159,6 +1159,39 @@ class DatasetRetrieval:
 
                         all_documents.extend(documents)
 
+    def _run_retriever_thread(
+        self,
+        *,
+        flask_app: Flask,
+        dataset_id: str,
+        query: str | None,
+        top_k: int,
+        all_documents: list[Document],
+        document_ids_filter: list[str] | None,
+        metadata_condition: MetadataFilteringCondition | None,
+        attachment_ids: list[str] | None,
+        cancel_event: threading.Event | None,
+        thread_exceptions: list[Exception] | None,
+    ) -> None:
+        try:
+            with session_factory.create_session() as session:
+                self._retriever(
+                    flask_app=flask_app,
+                    session=session,
+                    dataset_id=dataset_id,
+                    query=query or "",
+                    top_k=top_k,
+                    all_documents=all_documents,
+                    document_ids_filter=document_ids_filter,
+                    metadata_condition=metadata_condition,
+                    attachment_ids=attachment_ids,
+                )
+        except Exception as e:
+            if cancel_event:
+                cancel_event.set()
+            if thread_exceptions is not None:
+                thread_exceptions.append(e)
+
     def to_dataset_retriever_tool(
         self,
         session: Session,
@@ -1797,7 +1830,7 @@ class DatasetRetrieval:
                             else:
                                 continue
                     retrieval_thread = threading.Thread(
-                        target=self._retriever,
+                        target=self._run_retriever_thread,
                         kwargs={
                             "flask_app": flask_app,
                             "dataset_id": dataset.id,
@@ -1807,6 +1840,8 @@ class DatasetRetrieval:
                             "document_ids_filter": document_ids_filter,
                             "metadata_condition": metadata_condition,
                             "attachment_ids": [attachment_id] if attachment_id else None,
+                            "cancel_event": cancel_event,
+                            "thread_exceptions": thread_exceptions,
                         },
                     )
                     threads.append(retrieval_thread)
diff --git a/api/core/tools/signature.py b/api/core/tools/signature.py
index ca4756f2a4c..77d81bafbc1 100644
--- a/api/core/tools/signature.py
+++ b/api/core/tools/signature.py
@@ -64,34 +64,45 @@ def verify_tool_file_signature(file_id: str, timestamp: str, nonce: str, sign: s
     return current_time - int(timestamp) <= dify_config.FILES_ACCESS_TIMEOUT
 
 
-def get_signed_file_url_for_plugin(filename: str, mimetype: str, tenant_id: str, user_id: str) -> str:
+def get_signed_file_url_for_plugin(
+    filename: str, mimetype: str, tenant_id: str, user_id: str, conversation_id: str | None = None
+) -> str:
     """Build the signed upload URL used by the plugin-facing file upload endpoint."""
 
     base_url = dify_config.INTERNAL_FILES_URL or dify_config.FILES_URL
     upload_url = f"{base_url}/files/upload/for-plugin"
     timestamp = str(int(time.time()))
     nonce = os.urandom(16).hex()
-    data_to_sign = f"upload|{filename}|{mimetype}|{tenant_id}|{user_id}|{timestamp}|{nonce}"
+    data_to_sign = f"upload|{filename}|{mimetype}|{tenant_id}|{user_id}|{conversation_id or ''}|{timestamp}|{nonce}"
     sign = hmac.new(_secret_key(), data_to_sign.encode(), hashlib.sha256).digest()
     encoded_sign = base64.urlsafe_b64encode(sign).decode()
-    query = urllib.parse.urlencode(
-        {
-            "timestamp": timestamp,
-            "nonce": nonce,
-            "sign": encoded_sign,
-            "user_id": user_id,
-            "tenant_id": tenant_id,
-        }
-    )
+    query_params = {
+        "timestamp": timestamp,
+        "nonce": nonce,
+        "sign": encoded_sign,
+        "user_id": user_id,
+        "tenant_id": tenant_id,
+    }
+    if conversation_id:
+        query_params["conversation_id"] = conversation_id
+    query = urllib.parse.urlencode(query_params)
     return f"{upload_url}?{query}"
 
 
 def verify_plugin_file_signature(
-    *, filename: str, mimetype: str, tenant_id: str, user_id: str, timestamp: str, nonce: str, sign: str
+    *,
+    filename: str,
+    mimetype: str,
+    tenant_id: str,
+    user_id: str,
+    conversation_id: str | None = None,
+    timestamp: str,
+    nonce: str,
+    sign: str,
 ) -> bool:
     """Verify the signature used by the plugin-facing file upload endpoint."""
 
-    data_to_sign = f"upload|{filename}|{mimetype}|{tenant_id}|{user_id}|{timestamp}|{nonce}"
+    data_to_sign = f"upload|{filename}|{mimetype}|{tenant_id}|{user_id}|{conversation_id or ''}|{timestamp}|{nonce}"
     recalculated_sign = hmac.new(_secret_key(), data_to_sign.encode(), hashlib.sha256).digest()
     recalculated_encoded_sign = base64.urlsafe_b64encode(recalculated_sign).decode()
 
diff --git a/api/core/workflow/nodes/agent_v2/dify_tools_builder.py b/api/core/workflow/nodes/agent_v2/dify_tools_builder.py
index acb58a46952..fc2719a6204 100644
--- a/api/core/workflow/nodes/agent_v2/dify_tools_builder.py
+++ b/api/core/workflow/nodes/agent_v2/dify_tools_builder.py
@@ -11,6 +11,7 @@ from dify_agent.layers.dify_plugin import (
     DifyPluginToolCredentialType,
     DifyPluginToolParameter,
     DifyPluginToolParameterForm,
+    DifyPluginToolParameterType,
     DifyPluginToolsLayerConfig,
 )
 from sqlalchemy import select
@@ -351,7 +352,9 @@ class WorkflowAgentDifyToolsBuilder:
     @staticmethod
     def _tool_layer_destination(tool_config: AgentSoulDifyToolConfig) -> Literal["plugin", "core"]:
         provider_type = ToolProviderType.value_of(tool_config.provider_type)
-        if provider_type is ToolProviderType.PLUGIN:
+        if provider_type is ToolProviderType.PLUGIN or (
+            provider_type is ToolProviderType.BUILT_IN and _is_plugin_provider_id(tool_config.provider_id)
+        ):
             return "plugin"
         if provider_type in {
             ToolProviderType.BUILT_IN,
@@ -404,7 +407,7 @@ class WorkflowAgentDifyToolsBuilder:
             credentials=self._normalize_credentials(runtime.credentials, tool_name=exposed_name),
             runtime_parameters=runtime_parameters,
             parameters=parameters,
-            parameters_json_schema=tool_runtime.get_llm_parameters_json_schema(),
+            parameters_json_schema=self._plugin_parameters_json_schema(tool_runtime, parameters),
         )
 
     def _to_core_backend_tool_config(
@@ -456,6 +459,41 @@ class WorkflowAgentDifyToolsBuilder:
             description = tool_runtime.entity.description.llm
         return description
 
+    @staticmethod
+    def _plugin_parameters_json_schema(
+        tool_runtime: Tool,
+        parameters: list[DifyPluginToolParameter],
+    ) -> dict[str, Any]:
+        schema = tool_runtime.get_llm_parameters_json_schema()
+        properties = schema.setdefault("properties", {})
+        required = schema.setdefault("required", [])
+        if not isinstance(properties, dict) or not isinstance(required, list):
+            raise WorkflowAgentDifyToolsBuildError(
+                "agent_tool_declaration_invalid",
+                f"Dify Plugin Tool {tool_runtime.entity.identity.name!r} has invalid parameter schema.",
+            )
+
+        for parameter in parameters:
+            if parameter.form is not DifyPluginToolParameterForm.LLM:
+                continue
+            if parameter.type is DifyPluginToolParameterType.FILE:
+                properties[parameter.name] = _plugin_file_input_schema(parameter.llm_description or "")
+            elif parameter.type in {
+                DifyPluginToolParameterType.FILES,
+                DifyPluginToolParameterType.SYSTEM_FILES,
+            }:
+                properties[parameter.name] = {
+                    "type": "array",
+                    "items": _plugin_file_input_schema(parameter.llm_description or ""),
+                    "description": parameter.llm_description or "",
+                }
+            else:
+                continue
+
+            if parameter.required and parameter.name not in required:
+                required.append(parameter.name)
+        return schema
+
     @staticmethod
     def _runtime_parameters(
         tool_runtime: Tool,
@@ -498,3 +536,44 @@ class WorkflowAgentDifyToolsBuilder:
                 ),
             )
         return normalized
+
+
+def _is_plugin_provider_id(provider_id: str | None) -> bool:
+    if not provider_id:
+        return False
+    parts = provider_id.split("/")
+    return len(parts) == 3 and all(parts)
+
+
+def _plugin_file_input_schema(description: str) -> dict[str, Any]:
+    return {
+        "description": description,
+        "anyOf": [
+            {
+                "type": "string",
+                "minLength": 1,
+                "description": "HTTP(S) URL or sandbox-local file path.",
+            },
+            {
+                "type": "object",
+                "additionalProperties": False,
+                "required": ["transfer_method", "url"],
+                "properties": {
+                    "transfer_method": {"type": "string", "enum": ["remote_url"]},
+                    "url": {"type": "string", "minLength": 1},
+                },
+            },
+            {
+                "type": "object",
+                "additionalProperties": False,
+                "required": ["transfer_method", "reference"],
+                "properties": {
+                    "transfer_method": {
+                        "type": "string",
+                        "enum": ["local_file", "tool_file", "datasource_file"],
+                    },
+                    "reference": {"type": "string", "minLength": 1},
+                },
+            },
+        ],
+    }
diff --git a/api/core/workflow/nodes/agent_v2/output_adapter.py b/api/core/workflow/nodes/agent_v2/output_adapter.py
index 0fae0105348..b5ac93df6ae 100644
--- a/api/core/workflow/nodes/agent_v2/output_adapter.py
+++ b/api/core/workflow/nodes/agent_v2/output_adapter.py
@@ -333,6 +333,8 @@ class WorkflowAgentOutputAdapter:
         session_snapshot = None
         if isinstance(event, AgentBackendRunSucceededInternalEvent | AgentBackendDeferredToolCallInternalEvent):
             session_snapshot = event.session_snapshot
+            if event.usage is not None:
+                agent_backend["usage"] = dict(event.usage)
         if session_snapshot is not None:
             agent_backend["session_snapshot"] = {
                 "layer_count": len(session_snapshot.layers),
diff --git a/api/core/workflow/nodes/agent_v2/runtime_request_builder.py b/api/core/workflow/nodes/agent_v2/runtime_request_builder.py
index 5804fd5fdcd..1a4529b5d97 100644
--- a/api/core/workflow/nodes/agent_v2/runtime_request_builder.py
+++ b/api/core/workflow/nodes/agent_v2/runtime_request_builder.py
@@ -206,17 +206,14 @@ class WorkflowAgentRuntimeRequestBuilder:
                 "cli_tool_count": len(agent_soul.tools.cli_tools),
             }
 
-        config_layer_config: DifyConfigLayerConfig | None = None
-        soul_prompt_resolver = build_soul_mention_resolver(agent_soul)
-        if dify_config.AGENT_DRIVE_MANIFEST_ENABLED:
-            config_layer_config, config_warnings = build_config_layer_config(
-                agent_soul,
-                agent_id=context.agent.id,
-                config_version_id=context.snapshot.id,
-                config_version_kind="snapshot",
-            )
-            append_runtime_warnings(metadata, config_warnings)
-            soul_prompt_resolver = build_config_aware_soul_mention_resolver(agent_soul)
+        config_layer_config, config_warnings = build_config_layer_config(
+            agent_soul,
+            agent_id=context.agent.id,
+            config_version_id=context.snapshot.id,
+            config_version_kind="snapshot",
+        )
+        append_runtime_warnings(metadata, config_warnings)
+        soul_prompt_resolver = build_config_aware_soul_mention_resolver(agent_soul)
         soul_prompt = expand_prompt_mentions(agent_soul.prompt.system_prompt, soul_prompt_resolver).strip()
         knowledge_config = build_knowledge_layer_config(agent_soul)
 
@@ -717,7 +714,7 @@ def build_shell_layer_config(agent_soul: AgentSoulConfig) -> DifyShellLayerConfi
             for tool in (_shell_cli_tool(item) for item in agent_soul.tools.cli_tools if _cli_tool_enabled(item))
             if tool is not None
         ],
-        env=[env for env in (_shell_env_var(item) for item in agent_soul.env.variables) if env is not None],
+        env=_shell_env_vars(agent_soul.env.variables, agent_soul.env.secret_refs),
         secret_refs=[
             secret for secret in (_shell_secret_ref(item) for item in agent_soul.env.secret_refs) if secret is not None
         ],
@@ -864,8 +861,13 @@ def build_config_layer_config(
     agent_id: str | None = None,
     config_version_id: str | None = None,
     config_version_kind: Literal["snapshot", "draft", "build_draft"] = "snapshot",
-) -> tuple[DifyConfigLayerConfig | None, list[dict[str, str]]]:
-    """Derive prompt-mentioned eager-pull names from Agent Soul."""
+) -> tuple[DifyConfigLayerConfig, list[dict[str, str]]]:
+    """Build the always-present Agent config layer from Agent Soul state.
+
+    The ``dify.config`` layer must exist for every Agent v2 runtime request so
+    the backend can expose the config CLI/help surface even when the current
+    Agent Soul has no config assets, note, or prompt mentions.
+    """
 
     ordered_mentions = list(
         dict.fromkeys(
@@ -874,14 +876,6 @@ def build_config_layer_config(
             if mention.kind in {MentionKind.SKILL, MentionKind.FILE} and mention.ref_id
         )
     )
-    if (
-        not agent_soul.config_skills
-        and not agent_soul.config_files
-        and not agent_soul.config_note
-        and not ordered_mentions
-    ):
-        return None, []
-
     skill_names = {skill.name for skill in agent_soul.config_skills}
     file_names = {file_ref.name for file_ref in agent_soul.config_files}
     warnings: list[dict[str, str]] = []
@@ -968,11 +962,7 @@ def _shell_cli_tool(item: object) -> DifyShellCliToolConfig | None:
     if not commands and not isinstance(name, str):
         return None
     tool_env = data.get("env") if isinstance(data.get("env"), Mapping) else {}
-    env = [
-        env_var
-        for env_var in (_shell_env_var(item) for item in _env_entries(tool_env, "variables"))
-        if env_var is not None
-    ]
+    env = _shell_env_vars(_env_entries(tool_env, "variables"), _env_entries(tool_env, "secret_refs"))
     secret_refs = [
         secret_ref
         for secret_ref in (_shell_secret_ref(item) for item in _env_entries(tool_env, "secret_refs"))
@@ -995,6 +985,12 @@ def _env_entries(env: object, key: str) -> list[object]:
     return entries
 
 
+def _shell_env_vars(variables: Sequence[object], secret_refs: Sequence[object]) -> list[DifyShellEnvVarConfig]:
+    env_vars = [_shell_env_var(item) for item in variables]
+    secret_env_vars = [_shell_env_var(item) for item in secret_refs if _has_secret_value(item)]
+    return [env for env in [*env_vars, *secret_env_vars] if env is not None]
+
+
 def _shell_env_var(item: object) -> DifyShellEnvVarConfig | None:
     data = _plain_mapping(item)
     name = _name_from_mapping(data)
@@ -1011,13 +1007,15 @@ def _shell_secret_ref(item: object) -> DifyShellSecretRefConfig | None:
     name = _name_from_mapping(data)
     if name is None:
         return None
-    ref = (
-        data.get("ref")
-        or data.get("value")
-        or data.get("id")
-        or data.get("credential_id")
-        or data.get("provider_credential_id")
-    )
+    # Inline Composer values are passed as env vars because the agent-backend
+    # secret ref schema only accepts short backend-managed reference IDs.
+    if _has_secret_value(item):
+        return None
+    ref = data.get("ref") or data.get("credential_id") or data.get("provider_credential_id")
+    if ref is None:
+        ref = data.get("id")
+    if ref is None:
+        return None
     return DifyShellSecretRefConfig(name=name, ref=str(ref) if ref is not None else None)
 
 
@@ -1029,6 +1027,12 @@ def _plain_mapping(item: object) -> dict[str, Any]:
     return {}
 
 
+def _has_secret_value(item: object) -> bool:
+    data = _plain_mapping(item)
+    value = data.get("value")
+    return isinstance(value, str) and bool(value)
+
+
 def _name_from_mapping(item: Mapping[str, Any]) -> str | None:
     for key in ("name", "key", "env_name", "variable"):
         value = item.get(key)
diff --git a/api/models/agent_config_entities.py b/api/models/agent_config_entities.py
index 64990357acd..64fe85930c6 100644
--- a/api/models/agent_config_entities.py
+++ b/api/models/agent_config_entities.py
@@ -250,9 +250,10 @@ class AgentSecretRefConfig(AgentFlexibleConfig):
     env_name: str | None = Field(default=None, max_length=255)
     variable: str | None = Field(default=None, max_length=255)
     type: str | None = Field(default=None, max_length=64)
-    # UI-facing selected secret reference. This is a credential/ref id, not the
-    # plaintext secret value; runtime maps it to the shell-layer ``ref``.
-    value: str | None = Field(default=None, max_length=255)
+    # User-provided secret value. Long API tokens are valid here; runtime maps
+    # this field into a shell env var, while ref/id/credential_id fields keep the
+    # backend-managed secret reference path.
+    value: str | None = None
     id: str | None = Field(default=None, max_length=255)
     ref: str | None = Field(default=None, max_length=255)
     credential_id: str | None = Field(default=None, max_length=255)
@@ -515,9 +516,18 @@ class AgentTextToSpeechFeatureConfig(AgentFeatureToggleConfig):
     autoPlay: str | None = None
 
 
+class AgentSuggestedQuestionsAfterAnswerModelConfig(AgentFlexibleConfig):
+    """Legacy Chat App model config used only for follow-up question generation."""
+
+    provider: str = Field(min_length=1, max_length=255)
+    name: str = Field(min_length=1, max_length=255)
+    mode: str | None = Field(default=None, max_length=64)
+    completion_params: dict[str, Any] | None = None
+
+
 class AgentSuggestedQuestionsAfterAnswerFeatureConfig(AgentFeatureToggleConfig):
     prompt: str | None = None
-    model: AgentSoulModelConfig | None = None
+    model: AgentSuggestedQuestionsAfterAnswerModelConfig | None = None
 
 
 class AgentModerationIOConfig(AgentFlexibleConfig):
diff --git a/api/openapi/markdown/console-openapi.md b/api/openapi/markdown/console-openapi.md
index d261e338bb1..9b1341526b8 100644
--- a/api/openapi/markdown/console-openapi.md
+++ b/api/openapi/markdown/console-openapi.md
@@ -12935,7 +12935,7 @@ Default namespace
 | icon_background | string | Icon background color | No |
 | icon_type | [IconType](#icontype) | Icon type | No |
 | name | string | Agent name | Yes |
-| role | string | Agent role | Yes |
+| role | string | Agent role | No |
 
 #### AgentAppDetailWithSite
 
@@ -13062,7 +13062,7 @@ default (the config form sends the full desired feature state on save).
 | icon_type | [IconType](#icontype) | Icon type | No |
 | max_active_requests | integer | Maximum active requests | No |
 | name | string | App name | Yes |
-| role | string | Agent role | Yes |
+| role | string | Agent role | No |
 | use_icon_as_answer_icon | boolean | Use icon as answer icon | No |
 
 #### AgentAverageResponseTimeStatisticResponse
@@ -14569,9 +14569,20 @@ Soft lifecycle state for Agent records.
 | Name | Type | Description | Required |
 | ---- | ---- | ----------- | -------- |
 | enabled | boolean |  | No |
-| model | [AgentSoulModelConfig](#agentsoulmodelconfig) |  | No |
+| model | [AgentSuggestedQuestionsAfterAnswerModelConfig](#agentsuggestedquestionsafteranswermodelconfig) |  | No |
 | prompt | string |  | No |
 
+#### AgentSuggestedQuestionsAfterAnswerModelConfig
+
+Legacy Chat App model config used only for follow-up question generation.
+
+| Name | Type | Description | Required |
+| ---- | ---- | ----------- | -------- |
+| completion_params | object |  | No |
+| mode | string |  | No |
+| name | string |  | Yes |
+| provider | string |  | Yes |
+
 #### AgentTextToSpeechFeatureConfig
 
 | Name | Type | Description | Required |
diff --git a/api/services/agent/observability_service.py b/api/services/agent/observability_service.py
index 76b36abed2a..a209b8de0f0 100644
--- a/api/services/agent/observability_service.py
+++ b/api/services/agent/observability_service.py
@@ -619,9 +619,7 @@ WHERE
     @staticmethod
     def _statistics_message_scope_sql(source_filter: AgentSourceFilter) -> str:
         app_scope = "m.app_id = :app_id"
-        if source_filter.invoke_from is None:
-            app_scope += " AND m.invoke_from != :debugger"
-        else:
+        if source_filter.invoke_from is not None:
             app_scope += " AND m.invoke_from = :source"
         workflow_binding_filters = []
         if source_filter.app_id:
diff --git a/api/services/agent/skill_package_service.py b/api/services/agent/skill_package_service.py
index 3974d306a14..d5601987abc 100644
--- a/api/services/agent/skill_package_service.py
+++ b/api/services/agent/skill_package_service.py
@@ -175,7 +175,10 @@ class SkillPackageService:
                         continue
                     raise SkillPackageError(
                         "files_outside_skill_root",
-                        "skill archive contains files outside the selected skill root",
+                        (
+                            "skill package must contain exactly one skill; "
+                            "multiple skill folders in one archive are not supported"
+                        ),
                         status_code=400,
                     )
                 normalized_path = safe_path.removeprefix(skill_root_prefix)
diff --git a/api/services/agent_config_service.py b/api/services/agent_config_service.py
index af05acb9ef7..a0da08e5c2e 100644
--- a/api/services/agent_config_service.py
+++ b/api/services/agent_config_service.py
@@ -496,13 +496,16 @@ class AgentConfigService:
                 user_id=user_id,
             )
             self._require_writable(target, surface=surface)
-            skill_ref, _package = self._skill_normalizer.normalize(
-                content=content,
-                filename=filename,
-                requested_name=None,
-                tenant_id=tenant_id,
-                user_id=user_id,
-            )
+            try:
+                skill_ref, _package = self._skill_normalizer.normalize(
+                    content=content,
+                    filename=filename,
+                    requested_name=None,
+                    tenant_id=tenant_id,
+                    user_id=user_id,
+                )
+            except SkillPackageError as exc:
+                raise AgentConfigServiceError(exc.code, exc.message, status_code=exc.status_code) from exc
             agent_soul = target.agent_soul.model_copy(deep=True)
             existing = {item.name: item for item in agent_soul.config_skills}
             order = [item.name for item in agent_soul.config_skills]
diff --git a/api/tests/unit_tests/clients/agent_backend/test_event_adapter.py b/api/tests/unit_tests/clients/agent_backend/test_event_adapter.py
index ab88ba19640..f6c73fdb66d 100644
--- a/api/tests/unit_tests/clients/agent_backend/test_event_adapter.py
+++ b/api/tests/unit_tests/clients/agent_backend/test_event_adapter.py
@@ -1,6 +1,7 @@
 import pytest
 from agenton.compositor import CompositorSessionSnapshot
 from dify_agent.protocol import (
+    AgentRunUsage,
     DeferredToolCallPayload,
     PydanticAIStreamRunEvent,
     RunCancelledEvent,
@@ -59,7 +60,11 @@ def test_event_adapter_maps_run_succeeded_to_final_output():
         RunSucceededEvent(
             id="3-0",
             run_id="run-1",
-            data=RunSucceededEventData(output={"summary": "done"}, session_snapshot=snapshot),
+            data=RunSucceededEventData(
+                output={"summary": "done"},
+                session_snapshot=snapshot,
+                usage=AgentRunUsage(prompt_tokens=2, completion_tokens=3),
+            ),
         )
     )
 
@@ -69,6 +74,7 @@ def test_event_adapter_maps_run_succeeded_to_final_output():
             source_event_id="3-0",
             output={"summary": "done"},
             session_snapshot=snapshot,
+            usage={"prompt_tokens": 2, "completion_tokens": 3, "total_tokens": 5},
         )
     ]
 
diff --git a/api/tests/unit_tests/controllers/console/agent/test_agent_controllers.py b/api/tests/unit_tests/controllers/console/agent/test_agent_controllers.py
index e070a98c12a..2e5851349d4 100644
--- a/api/tests/unit_tests/controllers/console/agent/test_agent_controllers.py
+++ b/api/tests/unit_tests/controllers/console/agent/test_agent_controllers.py
@@ -318,7 +318,9 @@ def test_agent_app_list_and_create_use_agent_route(
         lambda: SimpleNamespace(webapp_auth=SimpleNamespace(enabled=False)),
     )
 
-    with app.test_request_context("/console/api/agent?page=1&limit=10&mode=workflow"):
+    with app.test_request_context(
+        "/console/api/agent?page=1&limit=10&mode=workflow&sort_by=recently_created&is_created_by_me=true"
+    ):
         listed = unwrap(AgentAppListApi.get)(AgentAppListApi(), "tenant-1", SimpleNamespace(id=account_id))
 
     assert listed["page"] == 1
@@ -343,6 +345,8 @@ def test_agent_app_list_and_create_use_agent_route(
     list_call = cast(dict[str, object], captured["list"])
     list_params = cast(Any, list_call["params"])
     assert list_params.mode == "agent"
+    assert list_params.sort_by == "recently_created"
+    assert list_params.is_created_by_me is True
     assert list_params.status == "normal"
 
     with app.test_request_context(
@@ -370,20 +374,54 @@ def test_agent_app_list_and_create_use_agent_route(
     assert create_params.agent_role == "Coordinator"
 
 
-def test_agent_app_create_requires_role(app: Flask, account_id: str) -> None:
-    with app.test_request_context(
-        "/console/api/agent",
-        json={"name": "Iris", "description": "Agent app", "icon_type": "emoji", "icon": "robot"},
-    ):
-        with pytest.raises(ValueError, match="Field required"):
-            unwrap(AgentAppListApi.post)(AgentAppListApi(), "tenant-1", SimpleNamespace(id=account_id))
+def test_agent_app_create_payload_allows_optional_role() -> None:
+    omitted = roster_controller.AgentAppCreatePayload.model_validate(
+        {"name": "Iris", "description": "Agent app", "icon_type": "emoji", "icon": "robot"}
+    )
+    blank = roster_controller.AgentAppCreatePayload.model_validate(
+        {"name": "Iris", "description": "Agent app", "role": "   ", "icon_type": "emoji", "icon": "robot"}
+    )
 
+    assert omitted.role is None
+    assert blank.role == ""
+
+
+def test_agent_app_create_omits_optional_role_as_empty_string(
+    app: Flask, monkeypatch: pytest.MonkeyPatch, account_id: str
+) -> None:
+    captured: dict[str, object] = {}
+
+    class FakeAppService:
+        def create_app(self, tenant_id: str, params: object, account: object) -> object:
+            captured["create"] = {"tenant_id": tenant_id, "params": params, "account": account}
+            return _app_detail_obj(id="app-created", bound_agent_id="agent-created")
+
+    monkeypatch.setattr(roster_controller, "AppService", FakeAppService)
+    monkeypatch.setattr(
+        roster_controller,
+        "_serialize_agent_app_detail",
+        lambda app_model, **_kwargs: {"id": "agent-created", "app_id": app_model.id},
+    )
+
+    current_user = SimpleNamespace(id=account_id)
     with app.test_request_context(
         "/console/api/agent",
-        json={"name": "Iris", "description": "Agent app", "role": "   ", "icon_type": "emoji", "icon": "robot"},
+        json={
+            "name": "No-role Iris",
+            "description": "Agent app",
+            "icon_type": "emoji",
+            "icon": "robot",
+        },
     ):
-        with pytest.raises(ValueError, match="Agent role is required"):
-            unwrap(AgentAppListApi.post)(AgentAppListApi(), "tenant-1", SimpleNamespace(id=account_id))
+        created, status = unwrap(AgentAppListApi.post)(AgentAppListApi(), "tenant-1", current_user)
+
+    assert status == 201
+    assert created == {"id": "agent-created", "app_id": "app-created"}
+    create_call = cast(dict[str, object], captured["create"])
+    create_params = cast(Any, create_call["params"])
+    assert create_call["tenant_id"] == "tenant-1"
+    assert create_call["account"] is current_user
+    assert create_params.agent_role == ""
 
 
 def test_agent_app_detail_update_delete_resolve_app_from_agent_id(
@@ -805,7 +843,7 @@ def test_agent_api_status_and_key_routes_resolve_backing_app(
     }
 
 
-def test_agent_app_update_rejects_empty_role(app: Flask, monkeypatch: pytest.MonkeyPatch) -> None:
+def test_agent_app_update_allows_empty_role(app: Flask, monkeypatch: pytest.MonkeyPatch) -> None:
     agent_id = "00000000-0000-0000-0000-000000000001"
     app_model = _app_detail_obj(id="app-1", bound_agent_id=agent_id)
     captured: dict[str, object] = {}
@@ -820,11 +858,28 @@ def test_agent_app_update_rejects_empty_role(app: Flask, monkeypatch: pytest.Mon
         "get_app_backing_agent",
         lambda _self, **kwargs: SimpleNamespace(
             id=agent_id,
+            app_id="app-1",
+            backing_app_id=None,
             role="",
             debug_conversation_id="debug-conversation-detail",
             active_config_snapshot_id=None,
         ),
     )
+    monkeypatch.setattr(
+        roster_controller.AgentRosterService,
+        "get_or_create_agent_app_debug_conversation_id",
+        lambda _self, **kwargs: "debug-conversation-detail",
+    )
+    monkeypatch.setattr(
+        roster_controller.AgentRosterService,
+        "count_agent_app_debug_conversation_messages",
+        lambda _self, **kwargs: 0,
+    )
+    monkeypatch.setattr(
+        roster_controller.AgentRosterService,
+        "active_config_is_published",
+        lambda _self, **kwargs: False,
+    )
     monkeypatch.setattr(
         roster_controller.FeatureService,
         "get_system_features",
@@ -845,8 +900,11 @@ def test_agent_app_update_rejects_empty_role(app: Flask, monkeypatch: pytest.Mon
         "/console/api/agent/00000000-0000-0000-0000-000000000001",
         json={"name": "Renamed", "description": "", "role": "", "icon_type": "emoji", "icon": "R"},
     ):
-        with pytest.raises(ValueError, match="String should have at least 1 character"):
-            unwrap(AgentAppApi.put)(AgentAppApi(), "tenant-1", SimpleNamespace(id="account-1"), agent_id)
+        updated = unwrap(AgentAppApi.put)(AgentAppApi(), "tenant-1", SimpleNamespace(id="account-1"), agent_id)
+
+    assert updated["role"] == ""
+    update_call = cast(dict[str, object], captured["update"])
+    assert cast(dict[str, object], update_call["args"])["role"] == ""
 
 
 def test_invite_options_get_parses_app_id(app: Flask, monkeypatch: pytest.MonkeyPatch) -> None:
@@ -1369,6 +1427,56 @@ def test_agent_chat_generate_and_stop_routes_resolve_app_from_agent_id(
     assert stop_call == {"current_user_id": account_id, "app_model": app_model, "task_id": "task-1"}
 
 
+def test_agent_chat_stream_preflight_raises_first_error_event() -> None:
+    class ClosableStream:
+        def __init__(self) -> None:
+            self.closed = False
+            self._chunks = iter(
+                [
+                    "event: ping\n\n",
+                    (
+                        'data: {"event":"error","message":"Incorrect API key provided",'
+                        '"code":"completion_request_error","status":400}\n\n'
+                    ),
+                ]
+            )
+
+        def __iter__(self):
+            return self
+
+        def __next__(self) -> str:
+            return next(self._chunks)
+
+        def close(self) -> None:
+            self.closed = True
+
+    stream = ClosableStream()
+
+    with pytest.raises(CompletionRequestError) as exc_info:
+        completion_controller._raise_agent_stream_error_before_response(stream)
+
+    assert "Incorrect API key provided" in exc_info.value.description
+    assert stream.closed is True
+
+
+def test_agent_chat_stream_preflight_preserves_first_normal_event() -> None:
+    stream = iter(
+        [
+            "event: ping\n\n",
+            'data: {"event":"message","answer":"hello"}\n\n',
+            'data: {"event":"message_end"}\n\n',
+        ]
+    )
+
+    wrapped = completion_controller._raise_agent_stream_error_before_response(stream)
+
+    assert list(wrapped) == [
+        "event: ping\n\n",
+        'data: {"event":"message","answer":"hello"}\n\n',
+        'data: {"event":"message_end"}\n\n',
+    ]
+
+
 def test_agent_build_chat_finalize_route_resolves_app_from_agent_id(
     app: Flask, monkeypatch: pytest.MonkeyPatch, account_id: str
 ) -> None:
diff --git a/api/tests/unit_tests/controllers/files/test_upload.py b/api/tests/unit_tests/controllers/files/test_upload.py
index 5281dcf0645..4d5522c0c6c 100644
--- a/api/tests/unit_tests/controllers/files/test_upload.py
+++ b/api/tests/unit_tests/controllers/files/test_upload.py
@@ -65,6 +65,7 @@ class TestPluginUploadFileApi:
                 "sign": "sig",
                 "tenant_id": "tenant-1",
                 "user_id": "user-1",
+                "conversation_id": "conversation-1",
             },
             file=dummy_file,
         )
@@ -83,6 +84,10 @@ class TestPluginUploadFileApi:
         assert result["id"] == "file-id"
         assert result["reference"] == build_file_reference(record_id="file-id")
         assert result["preview_url"] == "signed-url"
+        mock_verify_signature.assert_called_once()
+        assert mock_verify_signature.call_args.kwargs["conversation_id"] == "conversation-1"
+        tool_file_manager_instance.create_file_by_raw.assert_called_once()
+        assert tool_file_manager_instance.create_file_by_raw.call_args.kwargs["conversation_id"] == "conversation-1"
 
     def test_missing_file(self):
         module.request = fake_request(
diff --git a/api/tests/unit_tests/controllers/inner_api/plugin/test_plugin.py b/api/tests/unit_tests/controllers/inner_api/plugin/test_plugin.py
index 45f6ad31034..1b0a82ed7c5 100644
--- a/api/tests/unit_tests/controllers/inner_api/plugin/test_plugin.py
+++ b/api/tests/unit_tests/controllers/inner_api/plugin/test_plugin.py
@@ -272,6 +272,7 @@ class TestPluginUploadFileRequestApi:
         mock_payload = MagicMock()
         mock_payload.filename = "test.pdf"
         mock_payload.mimetype = "application/pdf"
+        mock_payload.conversation_id = "conversation-id"
 
         # Act
         raw_post = _extract_raw_post(PluginUploadFileRequestApi)
@@ -279,7 +280,11 @@ class TestPluginUploadFileRequestApi:
 
         # Assert
         mock_get_url.assert_called_once_with(
-            filename="test.pdf", mimetype="application/pdf", tenant_id="tenant-id", user_id="user-id"
+            filename="test.pdf",
+            mimetype="application/pdf",
+            tenant_id="tenant-id",
+            user_id="user-id",
+            conversation_id="conversation-id",
         )
         assert result["data"]["url"] == "https://storage.example.com/signed-upload-url"
 
diff --git a/api/tests/unit_tests/core/app/apps/agent_app/test_app_runner.py b/api/tests/unit_tests/core/app/apps/agent_app/test_app_runner.py
index a1f6512625b..8ebf68c7061 100644
--- a/api/tests/unit_tests/core/app/apps/agent_app/test_app_runner.py
+++ b/api/tests/unit_tests/core/app/apps/agent_app/test_app_runner.py
@@ -14,6 +14,7 @@ import pytest
 from agenton.compositor import CompositorSessionSnapshot
 from dify_agent.layers.ask_human import AskHumanToolResult
 from dify_agent.protocol import (
+    AgentRunUsage,
     CancelRunRequest,
     CancelRunResponse,
     PydanticAIStreamRunEvent,
@@ -59,13 +60,6 @@ class _FakeCredentialsProvider:
         return {"openai_api_key": "sk-test"}
 
 
-@pytest.fixture(autouse=True)
-def _disable_drive_manifest_by_default(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setattr(
-        "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", False
-    )
-
-
 class _NoToolsBuilder:
     def build_layers(self, **kwargs):
         del kwargs
@@ -75,12 +69,16 @@ class _NoToolsBuilder:
 class _FakeQueueManager:
     def __init__(self) -> None:
         self.events: list[Any] = []
+        self._stop_requested = False
 
     def publish(self, event: Any, _from: Any) -> None:
         self.events.append(event)
 
     def is_stopped(self) -> bool:
-        return False
+        return self._stop_requested
+
+    def request_stop(self) -> None:
+        self._stop_requested = True
 
 
 class _StoppedQueueManager(_FakeQueueManager):
@@ -143,10 +141,65 @@ class _StreamingFakeAgentBackendRunClient(FakeAgentBackendRunClient):
             data=RunSucceededEventData(
                 output={"text": "hello agent"},
                 session_snapshot=CompositorSessionSnapshot(layers=[]),
+                usage=AgentRunUsage(prompt_tokens=3, completion_tokens=5),
             ),
         )
 
 
+class _StreamingRecordingFakeAgentBackendRunClient(_RecordingFakeAgentBackendRunClient):
+    @override
+    def stream_events(self, run_id: str, *, after: str | None = None) -> Iterator[RunEvent]:
+        del after
+        created_at = datetime(2026, 1, 1, tzinfo=UTC)
+        yield RunStartedEvent(id="1-0", run_id=run_id, created_at=created_at)
+        yield PydanticAIStreamRunEvent(
+            id="2-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=PartDeltaEvent(index=0, delta=TextPartDelta(content_delta="hello ")),
+        )
+        yield PydanticAIStreamRunEvent(
+            id="3-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=PartDeltaEvent(index=0, delta=TextPartDelta(content_delta="agent")),
+        )
+        yield RunSucceededEvent(
+            id="4-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=RunSucceededEventData(
+                output={"text": "hello agent"},
+                session_snapshot=CompositorSessionSnapshot(layers=[]),
+            ),
+        )
+
+
+class _StreamingStopAfterFirstDeltaFakeAgentBackendRunClient(_RecordingFakeAgentBackendRunClient):
+    def __init__(self, *, queue_manager: _FakeQueueManager, **kwargs: Any) -> None:
+        super().__init__(**kwargs)
+        self._queue_manager = queue_manager
+
+    @override
+    def stream_events(self, run_id: str, *, after: str | None = None) -> Iterator[RunEvent]:
+        del after
+        created_at = datetime(2026, 1, 1, tzinfo=UTC)
+        yield RunStartedEvent(id="1-0", run_id=run_id, created_at=created_at)
+        yield PydanticAIStreamRunEvent(
+            id="2-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=PartDeltaEvent(index=0, delta=TextPartDelta(content_delta="hello ")),
+        )
+        self._queue_manager.request_stop()
+        yield PydanticAIStreamRunEvent(
+            id="3-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=PartDeltaEvent(index=0, delta=TextPartDelta(content_delta="agent")),
+        )
+
+
 class _StreamingPartStartFakeAgentBackendRunClient(FakeAgentBackendRunClient):
     @override
     def stream_events(self, run_id: str, *, after: str | None = None) -> Iterator[RunEvent]:
@@ -170,6 +223,46 @@ class _StreamingPartStartFakeAgentBackendRunClient(FakeAgentBackendRunClient):
         )
 
 
+class _NullOutputFakeAgentBackendRunClient(FakeAgentBackendRunClient):
+    @override
+    def stream_events(self, run_id: str, *, after: str | None = None) -> Iterator[RunEvent]:
+        del after
+        created_at = datetime(2026, 1, 1, tzinfo=UTC)
+        yield RunStartedEvent(id="1-0", run_id=run_id, created_at=created_at)
+        yield RunSucceededEvent(
+            id="2-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=RunSucceededEventData(
+                output=None,
+                session_snapshot=CompositorSessionSnapshot(layers=[]),
+            ),
+        )
+
+
+class _StreamingTextNullOutputFakeAgentBackendRunClient(FakeAgentBackendRunClient):
+    @override
+    def stream_events(self, run_id: str, *, after: str | None = None) -> Iterator[RunEvent]:
+        del after
+        created_at = datetime(2026, 1, 1, tzinfo=UTC)
+        yield RunStartedEvent(id="1-0", run_id=run_id, created_at=created_at)
+        yield PydanticAIStreamRunEvent(
+            id="2-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=PartDeltaEvent(index=0, delta=TextPartDelta(content_delta="streamed answer")),
+        )
+        yield RunSucceededEvent(
+            id="3-0",
+            run_id=run_id,
+            created_at=created_at,
+            data=RunSucceededEventData(
+                output=None,
+                session_snapshot=CompositorSessionSnapshot(layers=[]),
+            ),
+        )
+
+
 class _ProcessStreamingFakeAgentBackendRunClient(FakeAgentBackendRunClient):
     @override
     def stream_events(self, run_id: str, *, after: str | None = None) -> Iterator[RunEvent]:
@@ -276,6 +369,19 @@ class _FakeSessionStore:
         )
 
 
+class _MonotonicClock:
+    def __init__(self, *values: float) -> None:
+        self._values = list(values)
+        self._index = 0
+
+    def __call__(self) -> float:
+        if self._index >= len(self._values):
+            return self._values[-1]
+        value = self._values[self._index]
+        self._index += 1
+        return value
+
+
 def _soul() -> AgentSoulConfig:
     return AgentSoulConfig.model_validate(
         {
@@ -299,7 +405,12 @@ def _dify_ctx() -> Any:
     )
 
 
-def _runner(client: FakeAgentBackendRunClient, store: _FakeSessionStore) -> AgentAppRunner:
+def _runner(
+    client: FakeAgentBackendRunClient,
+    store: _FakeSessionStore,
+    *,
+    text_delta_debounce_seconds: float | None = 0,
+) -> AgentAppRunner:
     return AgentAppRunner(
         request_builder=AgentAppRuntimeRequestBuilder(
             credentials_provider=_FakeCredentialsProvider(),
@@ -308,6 +419,7 @@ def _runner(client: FakeAgentBackendRunClient, store: _FakeSessionStore) -> Agen
         agent_backend_client=client,
         event_adapter=AgentBackendRunEventAdapter(),
         session_store=store,  # type: ignore[arg-type]
+        text_delta_debounce_seconds=text_delta_debounce_seconds,
     )
 
 
@@ -374,15 +486,10 @@ def test_successful_turn_publishes_chunk_and_message_end_and_saves_session():
     assert saved_scope.agent_config_snapshot_id == "snap-1"
     assert saved_run_id == "fake-run-1"
     assert saved_snapshot is not None
+    assert saved_specs
     # A successful turn carries no ask_human pause correlation.
     assert pending_form_id is None
     assert pending_tool_call_id is None
-    assert [spec.name for spec in saved_specs] == [
-        "agent_soul_prompt",
-        "agent_app_user_prompt",
-        "execution_context",
-        "history",
-    ]
 
 
 def test_successful_turn_forwards_agent_backend_stream_text_deltas_without_duplicate_terminal_chunk():
@@ -390,13 +497,16 @@ def test_successful_turn_forwards_agent_backend_stream_text_deltas_without_dupli
     store = _FakeSessionStore()
     qm = _FakeQueueManager()
 
-    _run(_runner(client, store), qm)
+    _run(_runner(client, store, text_delta_debounce_seconds=0), qm)
 
     chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
     end_events = [e for e in qm.events if isinstance(e, QueueMessageEndEvent)]
     assert [event.chunk.delta.message.content for event in chunk_events] == ["hello ", "agent"]
     assert len(end_events) == 1
     assert end_events[0].llm_result.message.content == "hello agent"
+    assert end_events[0].llm_result.usage.prompt_tokens == 3
+    assert end_events[0].llm_result.usage.completion_tokens == 5
+    assert end_events[0].llm_result.usage.total_tokens == 8
     assert store.saved
 
 
@@ -405,7 +515,7 @@ def test_successful_turn_forwards_part_start_text_and_publishes_missing_terminal
     store = _FakeSessionStore()
     qm = _FakeQueueManager()
 
-    _run(_runner(client, store), qm)
+    _run(_runner(client, store, text_delta_debounce_seconds=0), qm)
 
     chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
     end_events = [e for e in qm.events if isinstance(e, QueueMessageEndEvent)]
@@ -414,6 +524,34 @@ def test_successful_turn_forwards_part_start_text_and_publishes_missing_terminal
     assert end_events[0].llm_result.message.content == "hello agent"
 
 
+def test_successful_turn_with_null_terminal_output_publishes_empty_answer_not_literal_null():
+    client = _NullOutputFakeAgentBackendRunClient()
+    store = _FakeSessionStore()
+    qm = _FakeQueueManager()
+
+    _run(_runner(client, store), qm)
+
+    chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
+    end_events = [e for e in qm.events if isinstance(e, QueueMessageEndEvent)]
+    assert chunk_events == []
+    assert len(end_events) == 1
+    assert end_events[0].llm_result.message.content == ""
+
+
+def test_successful_turn_with_streamed_text_and_null_terminal_output_keeps_streamed_answer():
+    client = _StreamingTextNullOutputFakeAgentBackendRunClient()
+    store = _FakeSessionStore()
+    qm = _FakeQueueManager()
+
+    _run(_runner(client, store, text_delta_debounce_seconds=0), qm)
+
+    chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
+    end_events = [e for e in qm.events if isinstance(e, QueueMessageEndEvent)]
+    assert [event.chunk.delta.message.content for event in chunk_events] == ["streamed answer"]
+    assert len(end_events) == 1
+    assert end_events[0].llm_result.message.content == "streamed answer"
+
+
 def test_successful_turn_persists_thinking_and_tool_process_events(monkeypatch):
     fake_session = _FakeDbSession()
     monkeypatch.setattr(app_runner_module.db, "session", fake_session)
@@ -421,7 +559,7 @@ def test_successful_turn_persists_thinking_and_tool_process_events(monkeypatch):
     store = _FakeSessionStore()
     qm = _FakeQueueManager()
 
-    _run(_runner(client, store), qm)
+    _run(_runner(client, store, text_delta_debounce_seconds=0), qm)
 
     chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
     assert [event.chunk.delta.message.content for event in chunk_events] == ["final answer"]
@@ -436,6 +574,50 @@ def test_successful_turn_persists_thinking_and_tool_process_events(monkeypatch):
     assert rows[1].observation == "ok"
 
 
+def test_streaming_turn_batches_text_deltas_within_debounce_window(monkeypatch):
+    monkeypatch.setattr(app_runner_module.time, "monotonic", _MonotonicClock(0.0, 0.2))
+    client = _StreamingFakeAgentBackendRunClient()
+    store = _FakeSessionStore()
+    qm = _FakeQueueManager()
+
+    _run(_runner(client, store, text_delta_debounce_seconds=0.5), qm)
+
+    chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
+    end_events = [e for e in qm.events if isinstance(e, QueueMessageEndEvent)]
+    assert [event.chunk.delta.message.content for event in chunk_events] == ["hello agent"]
+    assert len(end_events) == 1
+    assert end_events[0].llm_result.message.content == "hello agent"
+
+
+def test_streaming_turn_flushes_pending_text_before_terminal_success(monkeypatch):
+    monkeypatch.setattr(app_runner_module.time, "monotonic", _MonotonicClock(0.0))
+    client = _StreamingPartStartFakeAgentBackendRunClient()
+    store = _FakeSessionStore()
+    qm = _FakeQueueManager()
+
+    _run(_runner(client, store, text_delta_debounce_seconds=0.5), qm)
+
+    chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
+    end_events = [e for e in qm.events if isinstance(e, QueueMessageEndEvent)]
+    assert [event.chunk.delta.message.content for event in chunk_events] == ["hello", " agent"]
+    assert len(end_events) == 1
+    assert end_events[0].llm_result.message.content == "hello agent"
+
+
+def test_streaming_turn_flushes_pending_text_before_stop_and_cancel(monkeypatch):
+    monkeypatch.setattr(app_runner_module.time, "monotonic", _MonotonicClock(0.0))
+    store = _FakeSessionStore()
+    qm = _FakeQueueManager()
+    client = _StreamingStopAfterFirstDeltaFakeAgentBackendRunClient(queue_manager=qm)
+
+    with pytest.raises(GenerateTaskStoppedError):
+        _run(_runner(client, store, text_delta_debounce_seconds=0.5), qm)
+
+    chunk_events = [e for e in qm.events if isinstance(e, QueueLLMChunkEvent)]
+    assert [event.chunk.delta.message.content for event in chunk_events] == ["hello "]
+    assert client.cancelled_run_ids == ["fake-run-1"]
+
+
 def test_tool_result_without_identity_does_not_attach_to_previous_tool(monkeypatch):
     fake_session = _FakeDbSession()
     monkeypatch.setattr(app_runner_module.db, "session", fake_session)
@@ -611,6 +793,7 @@ def test_stopped_task_cancels_agent_backend_run_and_skips_session_save():
 
 
 def test_extract_answer_handles_plain_string_and_dict():
+    assert AgentAppRunner._extract_answer(None) == ""
     assert AgentAppRunner._extract_answer("plain text") == "plain text"
     assert AgentAppRunner._extract_answer({"text": "hi"}) == "hi"
     assert AgentAppRunner._extract_answer({"a": 1}) == '{"a": 1}'
diff --git a/api/tests/unit_tests/core/app/apps/agent_app/test_runtime_request_builder.py b/api/tests/unit_tests/core/app/apps/agent_app/test_runtime_request_builder.py
index 5267d1833e3..9828216f3b8 100644
--- a/api/tests/unit_tests/core/app/apps/agent_app/test_runtime_request_builder.py
+++ b/api/tests/unit_tests/core/app/apps/agent_app/test_runtime_request_builder.py
@@ -191,6 +191,8 @@ class TestAgentAppRuntimeRequestBuilder:
             "agent_soul_prompt",
             "agent_app_user_prompt",
             "execution_context",
+            DIFY_SHELL_LAYER_ID,
+            DIFY_CONFIG_LAYER_ID,
             "history",
             "llm",
         ]
@@ -394,10 +396,7 @@ def _soul_with_model_and_skill() -> AgentSoulConfig:
 
 
 class TestAgentAppConfigLayer:
-    def test_config_layer_injected_when_flag_enabled(self, monkeypatch: pytest.MonkeyPatch):
-        monkeypatch.setattr(
-            "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-        )
+    def test_config_layer_injected(self):
         builder = AgentAppRuntimeRequestBuilder(
             credentials_provider=_FakeCredentialsProvider(),
             dify_tools_builder=_NoToolsBuilder(),  # type: ignore[arg-type]
@@ -423,10 +422,7 @@ class TestAgentAppConfigLayer:
         assert names.index(DIFY_SHELL_LAYER_ID) == names.index("execution_context") + 1
         assert names.index(DIFY_CONFIG_LAYER_ID) == names.index(DIFY_SHELL_LAYER_ID) + 1
 
-    def test_no_config_layer_when_agent_soul_has_no_config_assets(self, monkeypatch: pytest.MonkeyPatch):
-        monkeypatch.setattr(
-            "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-        )
+    def test_config_layer_present_when_agent_soul_has_no_config_assets(self, monkeypatch: pytest.MonkeyPatch):
         monkeypatch.setattr("core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_SHELL_ENABLED", True)
         builder = AgentAppRuntimeRequestBuilder(
             credentials_provider=_FakeCredentialsProvider(),
@@ -436,14 +432,20 @@ class TestAgentAppConfigLayer:
         result = builder.build(_ctx(_soul_with_model()))
 
         layers = {layer.name: layer for layer in result.request.composition.layers}
-        assert DIFY_CONFIG_LAYER_ID not in layers
+        assert layers[DIFY_CONFIG_LAYER_ID].config.model_dump(mode="json") == {
+            "agent_id": "agent-1",
+            "config_version": {"id": "snap-1", "kind": "snapshot", "writable": False},
+            "skills": [],
+            "files": [],
+            "env_keys": [],
+            "note": "",
+            "mentioned_skill_names": [],
+            "mentioned_file_names": [],
+        }
         assert layers[DIFY_SHELL_LAYER_ID].deps == {"execution_context": "execution_context"}
         assert layers[DIFY_SHELL_LAYER_ID].config.agent_stub_drive_ref is None
 
-    def test_config_layer_for_build_draft_marks_config_writable(self, monkeypatch: pytest.MonkeyPatch):
-        monkeypatch.setattr(
-            "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-        )
+    def test_config_layer_for_build_draft_marks_config_writable(self):
         builder = AgentAppRuntimeRequestBuilder(
             credentials_provider=_FakeCredentialsProvider(),
             dify_tools_builder=_NoToolsBuilder(),  # type: ignore[arg-type]
@@ -470,17 +472,6 @@ class TestAgentAppConfigLayer:
             "mentioned_file_names": [],
         }
 
-    def test_no_config_layer_when_flag_disabled(self, monkeypatch: pytest.MonkeyPatch):
-        monkeypatch.setattr(
-            "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", False
-        )
-        builder = AgentAppRuntimeRequestBuilder(
-            credentials_provider=_FakeCredentialsProvider(),
-            dify_tools_builder=_NoToolsBuilder(),  # type: ignore[arg-type]
-        )
-        result = builder.build(_ctx(_soul_with_model_and_skill()))
-        assert all(layer.name != DIFY_CONFIG_LAYER_ID for layer in result.request.composition.layers)
-
     @pytest.mark.parametrize(
         ("system_prompt", "expected_prefix"),
         [
@@ -496,13 +487,9 @@ class TestAgentAppConfigLayer:
     )
     def test_agent_app_runtime_expands_config_mentions_in_agent_soul_prompt(
         self,
-        monkeypatch: pytest.MonkeyPatch,
         system_prompt: str,
         expected_prefix: str,
     ):
-        monkeypatch.setattr(
-            "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-        )
         soul = _soul_with_model_and_skill()
         soul.prompt.system_prompt = system_prompt
         builder = AgentAppRuntimeRequestBuilder(
@@ -518,11 +505,7 @@ class TestAgentAppConfigLayer:
 
     def test_agent_app_runtime_missing_config_mentions_fall_back_without_marker_leak(
         self,
-        monkeypatch: pytest.MonkeyPatch,
     ):
-        monkeypatch.setattr(
-            "core.app.apps.agent_app.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-        )
         soul = _soul_with_model()
         soul.prompt.system_prompt = (
             "Use [§skill:ghost-skill:Ghost Skill§], [§file:ghost.txt:Ghost File§], and [§file:no-label.txt§]."
@@ -537,3 +520,8 @@ class TestAgentAppConfigLayer:
         prompt_layer = next(layer for layer in result.request.composition.layers if layer.name == "agent_soul_prompt")
         assert prompt_layer.config.prefix == "Use Ghost Skill, Ghost File, and no-label.txt."
         assert "[§" not in prompt_layer.config.prefix
+        assert [warning["code"] for warning in result.metadata["runtime_support"]["unsupported_runtime_warnings"]] == [
+            "mention_target_missing",
+            "mention_target_missing",
+            "mention_target_missing",
+        ]
diff --git a/api/tests/unit_tests/core/helper/test_marketplace.py b/api/tests/unit_tests/core/helper/test_marketplace.py
index bd561b16373..eba1e4e5442 100644
--- a/api/tests/unit_tests/core/helper/test_marketplace.py
+++ b/api/tests/unit_tests/core/helper/test_marketplace.py
@@ -14,22 +14,22 @@ from core.helper.marketplace import (
 
 
 def test_get_plugin_pkg_url_contains_unique_identifier() -> None:
-    url = get_plugin_pkg_url("plugin@1.0.0")
+    url = get_plugin_pkg_url("langgenius/openai:0.4.2@checksum")
 
     assert "api/v1/plugins/download" in url
-    assert "unique_identifier=plugin@1.0.0" in url
+    assert "unique_identifier=langgenius%2Fopenai%3A0.4.2%40checksum" in url
 
 
 def test_download_plugin_pkg_delegates_with_configured_size(mocker: MockerFixture) -> None:
     mocked_download = mocker.patch("core.helper.marketplace.download_with_size_limit", return_value=b"pkg")
     mocker.patch("core.helper.marketplace.dify_config.PLUGIN_MAX_PACKAGE_SIZE", 1234)
 
-    result = download_plugin_pkg("plugin.a.b")
+    result = download_plugin_pkg("langgenius/openai:0.4.2@checksum")
 
     assert result == b"pkg"
     mocked_download.assert_called_once()
     called_url, called_limit = mocked_download.call_args.args
-    assert "unique_identifier=plugin.a.b" in called_url
+    assert "unique_identifier=langgenius%2Fopenai%3A0.4.2%40checksum" in called_url
     assert called_limit == 1234
 
 
diff --git a/api/tests/unit_tests/core/rag/retrieval/test_dataset_retrieval.py b/api/tests/unit_tests/core/rag/retrieval/test_dataset_retrieval.py
index b53520f1eaf..b121fe261c2 100644
--- a/api/tests/unit_tests/core/rag/retrieval/test_dataset_retrieval.py
+++ b/api/tests/unit_tests/core/rag/retrieval/test_dataset_retrieval.py
@@ -95,6 +95,16 @@ def _metadata_condition() -> AppMetadataFilteringCondition:
     return AppMetadataFilteringCondition(logical_operator="and", conditions=[])
 
 
+@contextmanager
+def _patched_retriever_session():
+    session = MagicMock()
+    session_ctx = MagicMock()
+    session_ctx.__enter__.return_value = session
+    session_ctx.__exit__.return_value = None
+    with patch("core.rag.retrieval.dataset_retrieval.session_factory.create_session", return_value=session_ctx):
+        yield session
+
+
 def create_side_effect_for_search(documents: list[Document]):
     """
     Create a side effect function for mocking search methods.
@@ -1682,7 +1692,15 @@ class TestRetrievalService:
 
         # Mock _retriever to return documents
         def side_effect_retriever(
-            flask_app, dataset_id, query, top_k, all_documents, document_ids_filter, metadata_condition, attachment_ids
+            flask_app,
+            session,
+            dataset_id,
+            query,
+            top_k,
+            all_documents,
+            document_ids_filter,
+            metadata_condition,
+            attachment_ids,
         ):
             all_documents.extend([doc1, doc2])
 
@@ -1694,23 +1712,24 @@ class TestRetrievalService:
         all_documents = []
 
         # Act - Call with dataset_count = 1
-        dataset_retrieval._multiple_retrieve_thread(
-            flask_app=mock_flask_app,
-            available_datasets=[mock_dataset],
-            metadata_condition=None,
-            metadata_filter_document_ids=None,
-            all_documents=all_documents,
-            tenant_id=tenant_id,
-            reranking_enable=True,
-            reranking_mode="reranking_model",
-            reranking_model={"reranking_provider_name": "cohere", "reranking_model_name": "rerank-v2"},
-            weights=None,
-            top_k=5,
-            score_threshold=0.5,
-            query="test query",
-            attachment_id=None,
-            dataset_count=1,  # Single dataset - should skip second reranking
-        )
+        with _patched_retriever_session():
+            dataset_retrieval._multiple_retrieve_thread(
+                flask_app=mock_flask_app,
+                available_datasets=[mock_dataset],
+                metadata_condition=None,
+                metadata_filter_document_ids=None,
+                all_documents=all_documents,
+                tenant_id=tenant_id,
+                reranking_enable=True,
+                reranking_mode="reranking_model",
+                reranking_model={"reranking_provider_name": "cohere", "reranking_model_name": "rerank-v2"},
+                weights=None,
+                top_k=5,
+                score_threshold=0.5,
+                query="test query",
+                attachment_id=None,
+                dataset_count=1,  # Single dataset - should skip second reranking
+            )
 
         # Assert
         # DataPostProcessor should NOT be called (second reranking skipped)
@@ -1757,7 +1776,15 @@ class TestRetrievalService:
 
         # Mock _retriever to return documents
         def side_effect_retriever(
-            flask_app, dataset_id, query, top_k, all_documents, document_ids_filter, metadata_condition, attachment_ids
+            flask_app,
+            session,
+            dataset_id,
+            query,
+            top_k,
+            all_documents,
+            document_ids_filter,
+            metadata_condition,
+            attachment_ids,
         ):
             all_documents.extend([doc1, doc2])
 
@@ -1793,23 +1820,24 @@ class TestRetrievalService:
         mock_dataset2.provider = "dify"
 
         # Act - Call with dataset_count = 2
-        dataset_retrieval._multiple_retrieve_thread(
-            flask_app=mock_flask_app,
-            available_datasets=[mock_dataset, mock_dataset2],
-            metadata_condition=None,
-            metadata_filter_document_ids=None,
-            all_documents=all_documents,
-            tenant_id=tenant_id,
-            reranking_enable=True,
-            reranking_mode="reranking_model",
-            reranking_model={"reranking_provider_name": "cohere", "reranking_model_name": "rerank-v2"},
-            weights=None,
-            top_k=5,
-            score_threshold=0.5,
-            query="test query",
-            attachment_id=None,
-            dataset_count=2,  # Multiple datasets - should perform second reranking
-        )
+        with _patched_retriever_session():
+            dataset_retrieval._multiple_retrieve_thread(
+                flask_app=mock_flask_app,
+                available_datasets=[mock_dataset, mock_dataset2],
+                metadata_condition=None,
+                metadata_filter_document_ids=None,
+                all_documents=all_documents,
+                tenant_id=tenant_id,
+                reranking_enable=True,
+                reranking_mode="reranking_model",
+                reranking_model={"reranking_provider_name": "cohere", "reranking_model_name": "rerank-v2"},
+                weights=None,
+                top_k=5,
+                score_threshold=0.5,
+                query="test query",
+                attachment_id=None,
+                dataset_count=2,  # Multiple datasets - should perform second reranking
+            )
 
         # Assert
         # DataPostProcessor SHOULD be called (second reranking performed)
@@ -1867,7 +1895,15 @@ class TestRetrievalService:
 
         # Mock _retriever to return documents
         def side_effect_retriever(
-            flask_app, dataset_id, query, top_k, all_documents, document_ids_filter, metadata_condition, attachment_ids
+            flask_app,
+            session,
+            dataset_id,
+            query,
+            top_k,
+            all_documents,
+            document_ids_filter,
+            metadata_condition,
+            attachment_ids,
         ):
             all_documents.extend([doc1, doc2])
 
@@ -1889,23 +1925,24 @@ class TestRetrievalService:
         all_documents = []
 
         # Act - Call with dataset_count = 1
-        dataset_retrieval._multiple_retrieve_thread(
-            flask_app=mock_flask_app,
-            available_datasets=[mock_dataset],
-            metadata_condition=None,
-            metadata_filter_document_ids=None,
-            all_documents=all_documents,
-            tenant_id=tenant_id,
-            reranking_enable=True,  # Reranking enabled but should be skipped for single dataset
-            reranking_mode="reranking_model",
-            reranking_model={"reranking_provider_name": "cohere", "reranking_model_name": "rerank-v2"},
-            weights=None,
-            top_k=5,
-            score_threshold=0.5,
-            query="test query",
-            attachment_id=None,
-            dataset_count=1,
-        )
+        with _patched_retriever_session():
+            dataset_retrieval._multiple_retrieve_thread(
+                flask_app=mock_flask_app,
+                available_datasets=[mock_dataset],
+                metadata_condition=None,
+                metadata_filter_document_ids=None,
+                all_documents=all_documents,
+                tenant_id=tenant_id,
+                reranking_enable=True,  # Reranking enabled but should be skipped for single dataset
+                reranking_mode="reranking_model",
+                reranking_model={"reranking_provider_name": "cohere", "reranking_model_name": "rerank-v2"},
+                weights=None,
+                top_k=5,
+                score_threshold=0.5,
+                query="test query",
+                attachment_id=None,
+                dataset_count=1,
+            )
 
         # Assert
         # DataPostProcessor should NOT be called
@@ -3720,7 +3757,15 @@ class TestKnowledgeRetrievalRegression:
         )
 
         def fake_retriever(
-            flask_app, dataset_id, query, top_k, all_documents, document_ids_filter, metadata_condition, attachment_ids
+            flask_app,
+            session,
+            dataset_id,
+            query,
+            top_k,
+            all_documents,
+            document_ids_filter,
+            metadata_condition,
+            attachment_ids,
         ):
             all_documents.append(document)
 
@@ -3744,32 +3789,35 @@ class TestKnowledgeRetrievalRegression:
         thread_exceptions: list[Exception] = []
 
         def target():
-            with patch.object(dataset_retrieval, "_retriever", side_effect=fake_retriever):
-                with patch(
+            with (
+                patch.object(dataset_retrieval, "_retriever", side_effect=fake_retriever),
+                patch(
                     "core.rag.retrieval.dataset_retrieval.DataPostProcessor",
                     ContextRequiredPostProcessor,
-                ):
-                    dataset_retrieval._multiple_retrieve_thread(
-                        flask_app=flask_app,
-                        available_datasets=[mock_dataset, secondary_dataset],
-                        metadata_condition=None,
-                        metadata_filter_document_ids=None,
-                        all_documents=all_documents,
-                        tenant_id=tenant_id,
-                        reranking_enable=True,
-                        reranking_mode="reranking_model",
-                        reranking_model={
-                            "reranking_provider_name": "cohere",
-                            "reranking_model_name": "rerank-v2",
-                        },
-                        weights=None,
-                        top_k=3,
-                        score_threshold=0.0,
-                        query="test query",
-                        attachment_id=None,
-                        dataset_count=2,  # force reranking branch
-                        thread_exceptions=thread_exceptions,  # ✅ key
-                    )
+                ),
+                _patched_retriever_session(),
+            ):
+                dataset_retrieval._multiple_retrieve_thread(
+                    flask_app=flask_app,
+                    available_datasets=[mock_dataset, secondary_dataset],
+                    metadata_condition=None,
+                    metadata_filter_document_ids=None,
+                    all_documents=all_documents,
+                    tenant_id=tenant_id,
+                    reranking_enable=True,
+                    reranking_mode="reranking_model",
+                    reranking_model={
+                        "reranking_provider_name": "cohere",
+                        "reranking_model_name": "rerank-v2",
+                    },
+                    weights=None,
+                    top_k=3,
+                    score_threshold=0.0,
+                    query="test query",
+                    attachment_id=None,
+                    dataset_count=2,  # force reranking branch
+                    thread_exceptions=thread_exceptions,
+                )
 
         t = threading.Thread(target=target)
         t.start()
@@ -3781,6 +3829,53 @@ class TestKnowledgeRetrievalRegression:
         # Current buggy code should record an exception (not raise it)
         assert not thread_exceptions, thread_exceptions
 
+    def test_run_retriever_thread_provides_session_to_retriever(self):
+        dataset_retrieval = DatasetRetrieval()
+        all_documents: list[Document] = []
+
+        with _patched_retriever_session() as session:
+            with patch.object(dataset_retrieval, "_retriever") as mock_retriever:
+                dataset_retrieval._run_retriever_thread(
+                    flask_app=_FakeFlaskApp(),
+                    dataset_id="dataset-1",
+                    query="test query",
+                    top_k=3,
+                    all_documents=all_documents,
+                    document_ids_filter=None,
+                    metadata_condition=None,
+                    attachment_ids=None,
+                    cancel_event=None,
+                    thread_exceptions=[],
+                )
+
+        mock_retriever.assert_called_once()
+        assert mock_retriever.call_args.kwargs["session"] is session
+
+    def test_run_retriever_thread_records_retriever_exception(self):
+        dataset_retrieval = DatasetRetrieval()
+        all_documents: list[Document] = []
+        cancel_event = threading.Event()
+        thread_exceptions: list[Exception] = []
+        expected_error = RuntimeError("retrieval failed")
+
+        with _patched_retriever_session():
+            with patch.object(dataset_retrieval, "_retriever", side_effect=expected_error):
+                dataset_retrieval._run_retriever_thread(
+                    flask_app=_FakeFlaskApp(),
+                    dataset_id="dataset-1",
+                    query="test query",
+                    top_k=3,
+                    all_documents=all_documents,
+                    document_ids_filter=None,
+                    metadata_condition=None,
+                    attachment_ids=None,
+                    cancel_event=cancel_event,
+                    thread_exceptions=thread_exceptions,
+                )
+
+        assert cancel_event.is_set()
+        assert thread_exceptions == [expected_error]
+
 
 class _FakeFlaskApp:
     def app_context(self):
diff --git a/api/tests/unit_tests/core/tools/test_signature.py b/api/tests/unit_tests/core/tools/test_signature.py
index a75fdee9085..ee80cedd499 100644
--- a/api/tests/unit_tests/core/tools/test_signature.py
+++ b/api/tests/unit_tests/core/tools/test_signature.py
@@ -138,6 +138,7 @@ def test_get_signed_file_url_for_plugin_and_verify_roundtrip(monkeypatch: pytest
         mimetype="application/pdf",
         tenant_id="tenant-id",
         user_id="user-id",
+        conversation_id="conversation-id",
     )
     parsed = urlparse(url)
     query = parse_qs(parsed.query)
@@ -146,12 +147,14 @@ def test_get_signed_file_url_for_plugin_and_verify_roundtrip(monkeypatch: pytest
     assert parsed.path == "/files/upload/for-plugin"
     assert query["tenant_id"] == ["tenant-id"]
     assert query["user_id"] == ["user-id"]
+    assert query["conversation_id"] == ["conversation-id"]
     assert (
         verify_plugin_file_signature(
             filename="report.pdf",
             mimetype="application/pdf",
             tenant_id="tenant-id",
             user_id="user-id",
+            conversation_id="conversation-id",
             timestamp=query["timestamp"][0],
             nonce=query["nonce"][0],
             sign=query["sign"][0],
diff --git a/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_agent_node.py b/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_agent_node.py
index f24f02a130f..99448f7f5a9 100644
--- a/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_agent_node.py
+++ b/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_agent_node.py
@@ -2,7 +2,6 @@ from types import SimpleNamespace
 from typing import cast
 from unittest.mock import MagicMock, patch
 
-import pytest
 from agenton.compositor import CompositorSessionSnapshot
 from dify_agent.layers.ask_human import AskHumanToolResult
 from dify_agent.protocol import RunStartedEvent, RunSucceededEvent, RunSucceededEventData
@@ -51,13 +50,6 @@ class FakeCredentialsProvider:
         return {"api_key": "secret-key"}
 
 
-@pytest.fixture(autouse=True)
-def _disable_drive_manifest_by_default(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", False
-    )
-
-
 def _restored_file(*, transfer_method: FileTransferMethod, reference: str) -> File:
     return File(
         type=FileType.DOCUMENT,
@@ -281,7 +273,8 @@ def test_agent_node_run_maps_successful_agent_backend_run_to_node_result():
     assert agent_log["agent_backend"]["run_id"] == "fake-run-1"
     assert agent_log["agent_backend"]["status"] == "succeeded"
     assert result.process_data["agent_id"] == "agent-1"
-    assert result.inputs["agent_backend_request"]["composition"]["layers"][5]["config"]["credentials"] == "[REDACTED]"
+    layers = {layer["name"]: layer for layer in result.inputs["agent_backend_request"]["composition"]["layers"]}
+    assert layers["llm"]["config"]["credentials"] == "[REDACTED]"
 
 
 def test_agent_node_run_normalizes_declared_file_output_with_canonical_mapping():
diff --git a/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_dify_tools_builder.py b/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_dify_tools_builder.py
index e9af5e7707a..9a040926edf 100644
--- a/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_dify_tools_builder.py
+++ b/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_dify_tools_builder.py
@@ -148,6 +148,31 @@ def _file_tool() -> FakeTool:
     return FakeTool(entity=entity, runtime=runtime)
 
 
+def _files_tool() -> FakeTool:
+    parameters = [
+        ToolParameter(
+            name="documents",
+            label=I18nObject(en_US="Documents"),
+            type=ToolParameter.ToolParameterType.FILES,
+            form=ToolParameter.ToolParameterForm.LLM,
+            required=True,
+            llm_description="The documents to inspect.",
+        )
+    ]
+    entity = ToolEntity(
+        identity=ToolIdentity(
+            author="langgenius",
+            name="inspect",
+            label=I18nObject(en_US="Inspect"),
+            provider="documents",
+        ),
+        description=ToolDescription(human=I18nObject(en_US="Inspect"), llm="Inspect documents."),
+        parameters=parameters,
+    )
+    runtime = ToolRuntime(tenant_id="tenant-1", user_id="user-1", credentials={}, runtime_parameters={})
+    return FakeTool(entity=entity, runtime=runtime)
+
+
 def _tts_tool() -> FakeTool:
     parameters = [
         ToolParameter(
@@ -274,6 +299,103 @@ def test_builds_core_tool_with_file_llm_parameter():
     assert runtime_provider.last_use_default_for_missing_form_parameters is True
 
 
+def test_builds_plugin_tool_with_file_llm_parameter_schema():
+    runtime_provider = FakeRuntimeProvider(_file_tool())
+    builder = WorkflowAgentDifyToolsBuilder(tool_runtime_provider=runtime_provider)
+    tools = AgentSoulToolsConfig.model_validate(
+        {
+            "dify_tools": [
+                {
+                    "provider_id": "langgenius/audio/audio",
+                    "provider_type": "plugin",
+                    "tool_name": "asr",
+                    "credential_type": "unauthorized",
+                }
+            ]
+        }
+    )
+
+    result = _build(builder, tools)
+
+    assert result is not None
+    schema = result.tools[0].parameters_json_schema
+    file_schema = schema["properties"]["audio_file"]
+    assert file_schema["anyOf"][0]["type"] == "string"
+    assert file_schema["anyOf"][1]["properties"]["transfer_method"]["enum"] == ["remote_url"]
+    assert file_schema["anyOf"][2]["properties"]["transfer_method"]["enum"] == [
+        "local_file",
+        "tool_file",
+        "datasource_file",
+    ]
+    assert schema["required"] == ["audio_file"]
+
+
+def test_builds_plugin_tool_with_files_llm_parameter_schema():
+    runtime_provider = FakeRuntimeProvider(_files_tool())
+    builder = WorkflowAgentDifyToolsBuilder(tool_runtime_provider=runtime_provider)
+    tools = AgentSoulToolsConfig.model_validate(
+        {
+            "dify_tools": [
+                {
+                    "provider_id": "langgenius/documents/documents",
+                    "provider_type": "plugin",
+                    "tool_name": "inspect",
+                    "credential_type": "unauthorized",
+                }
+            ]
+        }
+    )
+
+    result = _build(builder, tools)
+
+    assert result is not None
+    schema = result.tools[0].parameters_json_schema
+    files_schema = schema["properties"]["documents"]
+    assert files_schema["type"] == "array"
+    assert files_schema["items"]["anyOf"][0]["description"] == "HTTP(S) URL or sandbox-local file path."
+    assert schema["required"] == ["documents"]
+
+
+def test_builds_builtin_compat_plugin_tool_with_files_llm_parameter_schema():
+    runtime_provider = FakeRuntimeProvider(_files_tool())
+    builder = WorkflowAgentDifyToolsBuilder(tool_runtime_provider=runtime_provider)
+    tools = AgentSoulToolsConfig.model_validate(
+        {
+            "dify_tools": [
+                {
+                    "provider_id": "langgenius/dify-gmail/dify-gmail",
+                    "provider_type": "builtin",
+                    "provider": "langgenius/dify-gmail/dify-gmail",
+                    "tool_name": "add_attachment_to_draft",
+                    "credential_type": "api-key",
+                    "credential_id": "credential-1",
+                }
+            ]
+        }
+    )
+
+    result = builder.build_layers(
+        tenant_id="tenant-1",
+        app_id="app-1",
+        user_id="user-1",
+        tools=tools,
+        invoke_from=InvokeFrom.DEBUGGER,
+    )
+
+    assert result.plugin_tools is not None
+    assert result.core_tools is None
+    prepared = result.plugin_tools.tools[0]
+    assert prepared.plugin_id == "langgenius/dify-gmail"
+    assert prepared.provider == "dify-gmail"
+    assert prepared.tool_name == "add_attachment_to_draft"
+    files_schema = prepared.parameters_json_schema["properties"]["documents"]
+    assert files_schema["type"] == "array"
+    assert files_schema["items"]["anyOf"][0]["description"] == "HTTP(S) URL or sandbox-local file path."
+    assert prepared.parameters_json_schema["required"] == ["documents"]
+    assert runtime_provider.last_agent_tool is not None
+    assert runtime_provider.last_agent_tool.provider_type.value == "builtin"
+
+
 def test_build_layers_routes_plugin_direct_and_builtin_via_core() -> None:
     runtime_provider = FakeRuntimeProvider(_tool())
     builder = WorkflowAgentDifyToolsBuilder(tool_runtime_provider=runtime_provider)
diff --git a/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_runtime_request_builder.py b/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_runtime_request_builder.py
index 3ed5a688c8b..10f321d3a2a 100644
--- a/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_runtime_request_builder.py
+++ b/api/tests/unit_tests/core/workflow/nodes/agent_v2/test_runtime_request_builder.py
@@ -47,13 +47,6 @@ class FakeCredentialsProvider:
         return {"api_key": "secret-key"}
 
 
-@pytest.fixture(autouse=True)
-def _disable_drive_manifest_by_default(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", False
-    )
-
-
 class CapturingCredentialsProvider:
     def __init__(self) -> None:
         self.provider_name: str | None = None
@@ -491,7 +484,8 @@ def test_build_shell_layer_config_accepts_legacy_fallback_keys():
                 "secret_refs": [
                     {"variable": "TOKEN", "credential_id": "credential-1"},
                     {"name": "API_KEY", "provider_credential_id": "credential-2"},
-                    {"name": "EDITABLE_TOKEN", "value": "credential-3"},
+                    {"name": "EDITABLE_TOKEN", "value": "inline-secret-value"},
+                    {"name": "LEGACY_SECRET_REF", "id": "credential-3"},
                     {"ref": "missing-name"},
                 ],
             },
@@ -508,11 +502,12 @@ def test_build_shell_layer_config_accepts_legacy_fallback_keys():
     assert config["env"] == [
         {"name": "PROJECT_NAME", "value": "demo"},
         {"name": "RETRY_COUNT", "value": "3"},
+        {"name": "EDITABLE_TOKEN", "value": "inline-secret-value"},
     ]
     assert config["secret_refs"] == [
         {"name": "TOKEN", "ref": "credential-1"},
         {"name": "API_KEY", "ref": "credential-2"},
-        {"name": "EDITABLE_TOKEN", "ref": "credential-3"},
+        {"name": "LEGACY_SECRET_REF", "ref": "credential-3"},
     ]
     assert config["sandbox"] is None
 
@@ -613,7 +608,37 @@ def test_build_shell_layer_config_maps_cli_tool_scoped_env():
     ]
 
 
-def test_builds_workflow_run_request_with_dify_plugin_tools_layer():
+def test_build_shell_layer_config_maps_cli_tool_inline_secret_value_to_env():
+    agent_soul = AgentSoulConfig.model_validate(
+        {
+            "tools": {
+                "cli_tools": [
+                    {
+                        "name": "github",
+                        "command": "apt-get install -y gh",
+                        "env": {
+                            "secret_refs": [{"name": "GITHUB_TOKEN", "value": "ghp_" + "x" * 300}],
+                        },
+                    }
+                ]
+            }
+        }
+    )
+
+    config = build_shell_layer_config(agent_soul).model_dump(mode="json")
+
+    assert config["cli_tools"] == [
+        {
+            "name": "github",
+            "install_commands": ["apt-get install -y gh"],
+            "env": [{"name": "GITHUB_TOKEN", "value": "ghp_" + "x" * 300}],
+            "secret_refs": [],
+        }
+    ]
+
+
+def test_builds_workflow_run_request_with_dify_plugin_tools_layer(monkeypatch: pytest.MonkeyPatch):
+    monkeypatch.setattr("core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_SHELL_ENABLED", True)
     context = _context()
     snapshot = AgentConfigSnapshot(
         id="snapshot-1",
@@ -649,7 +674,10 @@ def test_builds_workflow_run_request_with_dify_plugin_tools_layer():
     dumped = result.request.model_dump(mode="json")
     layers = {layer["name"]: layer for layer in dumped["composition"]["layers"]}
     assert layers[DIFY_PLUGIN_TOOLS_LAYER_ID]["type"] == "dify.plugin.tools"
-    assert layers[DIFY_PLUGIN_TOOLS_LAYER_ID]["deps"] == {"execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID}
+    assert layers[DIFY_PLUGIN_TOOLS_LAYER_ID]["deps"] == {
+        "execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID,
+        "shell": DIFY_SHELL_LAYER_ID,
+    }
     assert layers[DIFY_PLUGIN_TOOLS_LAYER_ID]["config"]["tools"][0]["tool_name"] == "current_time"
     assert result.metadata["agent_tools"] == {
         "dify_tool_count": 1,
@@ -1300,7 +1328,7 @@ def test_build_config_layer_config_includes_soul_context_and_mentions():
     assert warnings == []
 
 
-def test_build_config_layer_config_returns_none_for_empty_agent_soul():
+def test_build_config_layer_config_returns_empty_config_for_empty_agent_soul():
     from core.workflow.nodes.agent_v2.runtime_request_builder import build_config_layer_config
 
     soul = AgentSoulConfig(
@@ -1308,30 +1336,43 @@ def test_build_config_layer_config_returns_none_for_empty_agent_soul():
     )
     config, warnings = build_config_layer_config(soul)
 
-    assert config is None
+    assert config is not None
+    assert config.model_dump(mode="json") == {
+        "agent_id": None,
+        "config_version": {"id": None, "kind": "snapshot", "writable": False},
+        "skills": [],
+        "files": [],
+        "env_keys": [],
+        "note": "",
+        "mentioned_skill_names": [],
+        "mentioned_file_names": [],
+    }
     assert warnings == []
 
 
-def test_workflow_run_request_has_no_config_layer_with_empty_agent_soul(monkeypatch: pytest.MonkeyPatch):
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-    )
+def test_workflow_run_request_has_config_layer_with_empty_agent_soul(monkeypatch: pytest.MonkeyPatch):
     monkeypatch.setattr("core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_SHELL_ENABLED", True)
 
     result = WorkflowAgentRuntimeRequestBuilder(credentials_provider=FakeCredentialsProvider()).build(_context())
 
     dumped = result.request.model_dump(mode="json")
     layers = {layer["name"]: layer for layer in dumped["composition"]["layers"]}
-    assert DIFY_CONFIG_LAYER_ID not in layers
+    assert layers[DIFY_CONFIG_LAYER_ID]["config"] == {
+        "agent_id": "agent-1",
+        "config_version": {"id": "snapshot-1", "kind": "snapshot", "writable": False},
+        "skills": [],
+        "files": [],
+        "env_keys": [],
+        "note": "",
+        "mentioned_skill_names": [],
+        "mentioned_file_names": [],
+    }
     assert layers[DIFY_SHELL_LAYER_ID]["deps"] == {"execution_context": DIFY_EXECUTION_CONTEXT_LAYER_ID}
     assert layers[DIFY_SHELL_LAYER_ID]["config"]["agent_stub_drive_ref"] is None
 
 
-def test_workflow_run_request_contains_config_layer_when_flag_enabled(monkeypatch: pytest.MonkeyPatch):
+def test_workflow_run_request_contains_config_layer():
     """Contract test: locks the dify.config composition shape against cross-package drift."""
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-    )
     context = _context()
     context.snapshot.config_snapshot = _soul_with_config_assets()
 
@@ -1372,10 +1413,7 @@ def test_workflow_run_request_contains_config_layer_when_flag_enabled(monkeypatc
     assert any(spec.name == DIFY_CONFIG_LAYER_ID and spec.type == "dify.config" for spec in specs)
 
 
-def test_workflow_runtime_expands_config_mentions_in_agent_soul_prompt(monkeypatch: pytest.MonkeyPatch):
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-    )
+def test_workflow_runtime_expands_config_mentions_in_agent_soul_prompt():
     context = _context()
     context.snapshot.config_snapshot = _soul_with_config_assets()
 
@@ -1386,10 +1424,7 @@ def test_workflow_runtime_expands_config_mentions_in_agent_soul_prompt(monkeypat
     assert "[§" not in soul_prompt.config.prefix
 
 
-def test_workflow_runtime_missing_config_mentions_fall_back_to_label_then_name(monkeypatch: pytest.MonkeyPatch):
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", True
-    )
+def test_workflow_runtime_missing_config_mentions_fall_back_to_label_then_name():
     context = _context()
     context.snapshot.config_snapshot = AgentSoulConfig(
         prompt={
@@ -1405,20 +1440,11 @@ def test_workflow_runtime_missing_config_mentions_fall_back_to_label_then_name(m
     soul_prompt = next(layer for layer in result.request.composition.layers if layer.name == "agent_soul_prompt")
     assert soul_prompt.config.prefix == "Use Ghost Skill, Ghost File, and no-label.txt."
     assert "[§" not in soul_prompt.config.prefix
-
-
-def test_workflow_run_request_has_no_config_layer_when_flag_disabled(monkeypatch: pytest.MonkeyPatch):
-    monkeypatch.setattr(
-        "core.workflow.nodes.agent_v2.runtime_request_builder.dify_config.AGENT_DRIVE_MANIFEST_ENABLED", False
-    )
-    context = _context()
-    context.snapshot.config_snapshot = _soul_with_config_assets()
-
-    result = WorkflowAgentRuntimeRequestBuilder(credentials_provider=FakeCredentialsProvider()).build(context)
-
-    dumped = result.request.model_dump(mode="json")
-    assert all(layer["name"] != DIFY_CONFIG_LAYER_ID for layer in dumped["composition"]["layers"])
-    assert result.metadata["runtime_support"]["unsupported_runtime_warnings"] == []
+    assert [warning["code"] for warning in result.metadata["runtime_support"]["unsupported_runtime_warnings"]] == [
+        "mention_target_missing",
+        "mention_target_missing",
+        "mention_target_missing",
+    ]
 
 
 def test_build_config_layer_config_missing_mentions_warn_without_catalog():
diff --git a/api/tests/unit_tests/services/agent/test_agent_composer_entities.py b/api/tests/unit_tests/services/agent/test_agent_composer_entities.py
index 43323ca49b4..07f993c1b8e 100644
--- a/api/tests/unit_tests/services/agent/test_agent_composer_entities.py
+++ b/api/tests/unit_tests/services/agent/test_agent_composer_entities.py
@@ -124,6 +124,38 @@ def test_agent_app_soul_allows_app_features_and_variables():
     assert payload.agent_soul.app_variables[0].name == "company_name"
 
 
+def test_agent_app_soul_accepts_legacy_follow_up_model_config():
+    payload = ComposerSavePayload.model_validate(
+        {
+            "variant": ComposerVariant.AGENT_APP,
+            "save_strategy": ComposerSaveStrategy.SAVE_TO_CURRENT_VERSION,
+            "agent_soul": {
+                "app_features": {
+                    "suggested_questions_after_answer": {
+                        "enabled": True,
+                        "prompt": "Suggest useful follow-up questions.",
+                        "model": {
+                            "provider": "openai",
+                            "name": "gpt-4o-mini",
+                            "mode": "chat",
+                            "completion_params": {"temperature": 0.7, "max_tokens": 128},
+                        },
+                    },
+                },
+            },
+        }
+    )
+
+    assert payload.agent_soul is not None
+    follow_up = payload.agent_soul.app_features.suggested_questions_after_answer
+    assert follow_up is not None
+    assert follow_up.enabled is True
+    assert follow_up.model is not None
+    assert follow_up.model.provider == "openai"
+    assert follow_up.model.name == "gpt-4o-mini"
+    assert follow_up.model.completion_params == {"temperature": 0.7, "max_tokens": 128}
+
+
 def test_composer_save_payload_accepts_new_roster_metadata():
     payload = ComposerSavePayload.model_validate(
         {
diff --git a/api/tests/unit_tests/services/agent/test_agent_observability_service.py b/api/tests/unit_tests/services/agent/test_agent_observability_service.py
index f98c548d176..1b735a18a62 100644
--- a/api/tests/unit_tests/services/agent/test_agent_observability_service.py
+++ b/api/tests/unit_tests/services/agent/test_agent_observability_service.py
@@ -50,6 +50,23 @@ def test_resolve_source_filters_accepts_multiple_structured_sources() -> None:
     assert AgentObservabilityService.resolve_source_filters(("all", "webapp:app-1"))[0].kind == "all"
 
 
+def test_statistics_all_source_includes_debugger_messages() -> None:
+    source_filter = AgentObservabilityService.resolve_source_filter("all")
+
+    scope_sql = AgentObservabilityService._statistics_message_scope_sql(source_filter)
+
+    assert "m.app_id = :app_id" in scope_sql
+    assert "m.invoke_from != :debugger" not in scope_sql
+
+
+def test_statistics_explicit_source_filters_invoke_from() -> None:
+    source_filter = AgentObservabilityService.resolve_source_filter("debugger")
+
+    scope_sql = AgentObservabilityService._statistics_message_scope_sql(source_filter)
+
+    assert "m.invoke_from = :source" in scope_sql
+
+
 def test_apply_status_filter_accepts_multiple_statuses() -> None:
     class FakeStmt:
         def __init__(self):
diff --git a/api/tests/unit_tests/services/agent/test_agent_services.py b/api/tests/unit_tests/services/agent/test_agent_services.py
index 2ff68108c81..cebbbf4f94f 100644
--- a/api/tests/unit_tests/services/agent/test_agent_services.py
+++ b/api/tests/unit_tests/services/agent/test_agent_services.py
@@ -2826,11 +2826,12 @@ def test_composer_validator_rejects_unauthorized_secret_and_cli_tool():
 
 def test_composer_validator_accepts_valid_shell_env_and_cli():
     """Valid shell identifiers + a disabled empty CLI tool pass validation."""
+    long_secret = "sk-" + "x" * 512
     config = ComposerConfigValidator.validate_agent_soul_dict(
         {
             "env": {
                 "variables": [{"name": "MY_VAR", "value": "v"}],
-                "secret_refs": [{"name": "API_TOKEN", "value": "credential-1"}],
+                "secret_refs": [{"name": "API_TOKEN", "value": long_secret}],
             },
             "tools": {
                 "cli_tools": [
@@ -2855,7 +2856,7 @@ def test_composer_validator_accepts_valid_shell_env_and_cli():
     )
     assert {variable.name for variable in config.env.variables} == {"MY_VAR"}
     assert {secret.name for secret in config.env.secret_refs} == {"API_TOKEN"}
-    assert config.env.secret_refs[0].value == "credential-1"
+    assert config.env.secret_refs[0].value == long_secret
     assert config.tools.cli_tools[0].env.variables[0].name == "JQ_COLOR"
     assert config.tools.cli_tools[0].env.secret_refs[0].name == "JQ_TOKEN"
     assert config.tools.cli_tools[0].env.secret_refs[0].value == "credential-2"
diff --git a/api/tests/unit_tests/services/plugin/test_plugin_service.py b/api/tests/unit_tests/services/plugin/test_plugin_service.py
index bd8d84834fb..a8922154a95 100644
--- a/api/tests/unit_tests/services/plugin/test_plugin_service.py
+++ b/api/tests/unit_tests/services/plugin/test_plugin_service.py
@@ -8,6 +8,7 @@ import zstandard
 from pydantic import TypeAdapter
 from redis import RedisError
 
+from core.plugin.entities.plugin import PluginInstallationSource
 from core.plugin.entities.plugin_daemon import PluginInstallTask, PluginInstallTaskStatus, PluginModelProviderEntity
 from graphon.model_runtime.entities.common_entities import I18nObject
 from graphon.model_runtime.entities.provider_entities import ConfigurateMethod, ProviderEntity
@@ -888,9 +889,27 @@ class TestPluginModelProviderCacheInvalidation:
 
             from core.plugin.plugin_service import PluginService
 
-            result = PluginService.install_from_local_pkg("tenant-1", ["langgenius/openai:1.0.0"])
+            result = PluginService.install_from_local_pkg(
+                "tenant-1",
+                [
+                    "langgenius/openai:1.0.0",
+                    "langgenius/tavily:1.0.0",
+                ],
+            )
 
         assert result == "task-id"
+        installer.install_from_identifiers.assert_called_once_with(
+            "tenant-1",
+            [
+                "langgenius/openai:1.0.0",
+                "langgenius/tavily:1.0.0",
+            ],
+            PluginInstallationSource.Package,
+            [
+                {"plugin_unique_identifier": "langgenius/openai:1.0.0"},
+                {"plugin_unique_identifier": "langgenius/tavily:1.0.0"},
+            ],
+        )
         invalidate_cache.assert_called_once_with("tenant-1")
 
     def test_upgrade_plugin_with_github_invalidates_model_provider_cache_for_tenant(self) -> None:
diff --git a/api/tests/unit_tests/services/test_agent_config_service.py b/api/tests/unit_tests/services/test_agent_config_service.py
index 57b55f0a49f..f9a649a13ab 100644
--- a/api/tests/unit_tests/services/test_agent_config_service.py
+++ b/api/tests/unit_tests/services/test_agent_config_service.py
@@ -342,6 +342,38 @@ def test_push_file_for_console_uses_service_owned_upload_lookup_and_naming() ->
     session.commit.assert_called_once()
 
 
+def test_upload_skill_for_console_maps_package_validation_failures() -> None:
+    session = MagicMock()
+    service = AgentConfigService()
+    target = _target(kind=AgentConfigVersionKind.DRAFT, writable=False)
+    message = "skill package must contain exactly one skill; multiple skill folders in one archive are not supported"
+
+    with (
+        patch(f"{MODULE}.session_factory.create_session", return_value=_session_cm(session)),
+        patch.object(service, "_resolve_target_in_session", return_value=target),
+        patch.object(
+            service._skill_normalizer,
+            "normalize",
+            side_effect=SkillPackageError("files_outside_skill_root", message, status_code=400),
+        ),
+    ):
+        with pytest.raises(AgentConfigServiceError, match="exactly one skill") as exc_info:
+            service.upload_skill_for_console(
+                tenant_id=TENANT,
+                agent_id=AGENT,
+                user_id=USER,
+                config_version_id="draft-1",
+                config_version_kind=AgentConfigVersionKind.DRAFT,
+                content=b"bad-archive",
+                filename="skills.zip",
+            )
+
+    assert exc_info.value.code == "files_outside_skill_root"
+    assert exc_info.value.message == message
+    assert exc_info.value.status_code == 400
+    session.commit.assert_not_called()
+
+
 def test_apply_skill_updates_rejects_non_tool_file_refs() -> None:
     service = AgentConfigService()
 
diff --git a/api/tests/unit_tests/services/test_file_request_service.py b/api/tests/unit_tests/services/test_file_request_service.py
index 3bab033dfec..f5d1a59c17e 100644
--- a/api/tests/unit_tests/services/test_file_request_service.py
+++ b/api/tests/unit_tests/services/test_file_request_service.py
@@ -5,7 +5,6 @@ import pytest
 
 from core.app.entities.app_invoke_entities import InvokeFrom, UserFrom
 from core.app.file_access import FileAccessScope
-from core.workflow.file_reference import build_file_reference
 from services.file_request_service import FileRequestService
 
 
@@ -25,7 +24,7 @@ def test_request_download_url_builds_file_under_bound_scope(
     fake_file = MagicMock(filename="report.pdf", mime_type="application/pdf", size=123)
     access_controller = MagicMock()
     service = FileRequestService(access_controller=access_controller)
-    reference = build_file_reference(record_id="tool-file-1")
+    reference = "dify-file-ref:tool-file-1"
 
     with (
         patch("services.file_request_service.bind_file_access_scope", return_value=nullcontext()) as bind_scope,
diff --git a/api/uv.lock b/api/uv.lock
index 6e7e1811064..30df243450a 100644
--- a/api/uv.lock
+++ b/api/uv.lock
@@ -1306,7 +1306,7 @@ requires-dist = [
     { name = "pydantic-ai-slim", extras = ["anthropic", "google", "openai"], marker = "extra == 'server'", specifier = ">=1.85.1,<2.0.0" },
     { name = "pydantic-settings", marker = "extra == 'server'", specifier = ">=2.12.0,<3.0.0" },
     { name = "redis", marker = "extra == 'server'", specifier = ">=7.4.0,<8.0.0" },
-    { name = "shell-session-manager", marker = "extra == 'server'", specifier = "==2.3.0" },
+    { name = "shell-session-manager", marker = "extra == 'server'", specifier = "==2.4.0" },
     { name = "typer", specifier = ">=0.16.1,<0.17" },
     { name = "typing-extensions", specifier = ">=4.12.2,<5.0.0" },
     { name = "uvicorn", extras = ["standard"], marker = "extra == 'server'", specifier = "==0.46.0" },
diff --git a/dify-agent/docker/local-sandbox/Dockerfile b/dify-agent/docker/local-sandbox/Dockerfile
index 955ed68a599..382e43fb0a6 100644
--- a/dify-agent/docker/local-sandbox/Dockerfile
+++ b/dify-agent/docker/local-sandbox/Dockerfile
@@ -4,7 +4,7 @@
 #   docker build -f docker/local-sandbox/Dockerfile -t dify-agent-local-sandbox:local .
 #
 # This image merges the former shellctl-only image with the sandbox-visible
-# Agent Stub client CLI. It runs shellctl by default, and shellctl-managed jobs
+# Agent Stub client CLI. It runs `shellctl serve` by default, and shellctl-managed jobs
 # can call `dify-agent ...` without installing extra packages at runtime.
 
 FROM python:3.12-slim-bookworm AS base
@@ -13,7 +13,7 @@ ARG NODE_VERSION=22.22.1
 ARG PNPM_VERSION=11.9.0
 ARG UV_VERSION=0.8.9
 ARG DIFY_AGENT_TOOL_SPEC=.[grpc]
-ARG SHELL_SESSION_MANAGER_TOOL_SPEC=shell-session-manager
+ARG SHELL_SESSION_MANAGER_TOOL_SPEC=shell-session-manager==2.4.0
 
 ENV PYTHONDONTWRITEBYTECODE=1 \
     PYTHONUNBUFFERED=1 \
@@ -83,6 +83,7 @@ COPY --from=tools ${UV_TOOL_BIN_DIR} ${UV_TOOL_BIN_DIR}
 
 RUN useradd --create-home --shell /bin/sh dify \
     && mkdir -p /mnt/drive \
+    && chown dify:dify /home \
     && chown -R dify:dify /home/dify /mnt/drive
 
 USER dify
diff --git a/dify-agent/docs/dify-agent/user-manual/shell-layer/index.md b/dify-agent/docs/dify-agent/user-manual/shell-layer/index.md
index a915031fe1d..23efec73205 100644
--- a/dify-agent/docs/dify-agent/user-manual/shell-layer/index.md
+++ b/dify-agent/docs/dify-agent/user-manual/shell-layer/index.md
@@ -230,7 +230,7 @@ The provided `docker/local-sandbox/Dockerfile` installs:
 - `tmux`, required by `shellctl` to manage shell jobs;
 - common shell workspace tools: `git`, `openssh-client`, `jq`, `ripgrep`,
   `unzip`, `zip`, `file`, `procps`, and `less`;
-- `shell-session-manager==2.3.0` as a standalone uv tool, which provides the
+- `shell-session-manager==2.3.1` as a standalone uv tool, which provides the
   `shellctl` CLI/server;
 - `uv`, so uv shebang scripts with PEP 723 metadata can run inside the shell
   workspace and Python CLI tools can be installed with isolated tool
diff --git a/dify-agent/pyproject.toml b/dify-agent/pyproject.toml
index 1c8842dccc5..5bc84091a24 100644
--- a/dify-agent/pyproject.toml
+++ b/dify-agent/pyproject.toml
@@ -27,7 +27,7 @@ server = [
     "pydantic-ai-slim[anthropic,google,openai]>=1.85.1,<2.0.0",
     "pydantic-settings>=2.12.0,<3.0.0",
     "redis>=7.4.0,<8.0.0",
-    "shell-session-manager==2.3.0",
+    "shell-session-manager==2.4.0",
     "uvicorn[standard]==0.46.0",
 ]
 
diff --git a/dify-agent/src/dify_agent/adapters/llm/model.py b/dify-agent/src/dify_agent/adapters/llm/model.py
index 1863dec1388..26982cc2c29 100644
--- a/dify-agent/src/dify_agent/adapters/llm/model.py
+++ b/dify-agent/src/dify_agent/adapters/llm/model.py
@@ -565,13 +565,7 @@ def _chunk_to_stream_events(
     elif isinstance(message.content, list):
         for part in _map_assistant_content_to_response_parts(message.content):
             if isinstance(part, TextPart):
-                events.extend(
-                    parts_manager.handle_text_delta(
-                        vendor_part_id=None,
-                        content=part.content,
-                        provider_name=provider_name,
-                    )
-                )
+                events.extend(embedded_thinking_parser.parse(parts_manager, part.content, provider_name))
             else:
                 events.append(parts_manager.handle_part(vendor_part_id=None, part=part))
 
diff --git a/dify-agent/src/dify_agent/adapters/shell/shellctl.py b/dify-agent/src/dify_agent/adapters/shell/shellctl.py
index 31f2e13675f..6cf334245fb 100644
--- a/dify-agent/src/dify_agent/adapters/shell/shellctl.py
+++ b/dify-agent/src/dify_agent/adapters/shell/shellctl.py
@@ -1,3 +1,11 @@
+"""Shellctl-backed shell provider adapter for dify-agent.
+
+The shell-session-manager SDK owns the HTTP timeout policy for long-polling
+shellctl requests. This adapter stays narrowly focused on translating SDK and
+transport failures into ``ShellProviderError`` so the shell layer can return
+tool observations instead of aborting the agent loop.
+"""
+
 from __future__ import annotations
 
 import base64
@@ -10,6 +18,8 @@ from collections.abc import Callable
 from dataclasses import dataclass
 from typing import Protocol, TypeVar
 
+import httpx
+
 from dify_agent.adapters.shell.protocols import (
     ShellCommandProtocol,
     ShellCommandResult,
@@ -168,11 +178,11 @@ class ShellctlCommands(ShellCommandProtocol):
         grace_seconds: float | None = None,
     ) -> None:
         try:
-            _ = await self.client.delete(job_id, force=force, grace_seconds=grace_seconds)
-        except RuntimeError as exc:
-            if getattr(exc, "code", None) == "job_not_found":
+            _ = await _run_client_call(self.client.delete(job_id, force=force, grace_seconds=grace_seconds))
+        except ShellProviderError as exc:
+            if exc.code == "job_not_found":
                 return
-            raise _map_error(exc) from exc
+            raise
 
 
 @dataclass(slots=True)
@@ -274,8 +284,14 @@ class _CompletedShellctlJob:
 
 
 async def _run_client_call(awaitable: Awaitable[ResultT]) -> ResultT:
+    """Map shellctl client boundary failures into provider-layer errors."""
+
     try:
         return await awaitable
+    except httpx.TimeoutException as exc:
+        raise ShellProviderError(str(exc), code="timeout") from exc
+    except httpx.RequestError as exc:
+        raise ShellProviderError(str(exc), code="request_error") from exc
     except RuntimeError as exc:
         raise _map_error(exc) from exc
 
@@ -338,7 +354,7 @@ async def _run_to_completion(
     finally:
         if job_id is not None:
             try:
-                await client.delete(job_id, force=True)
+                await _run_client_call(client.delete(job_id, force=True))
             except RuntimeError as exc:
                 logger.warning("Failed to delete shellctl job %s: %s", job_id, exc)
 
diff --git a/dify-agent/src/dify_agent/agent_stub/cli/_config.py b/dify-agent/src/dify_agent/agent_stub/cli/_config.py
index 0aabc9b1e25..7d20354ac7d 100644
--- a/dify-agent/src/dify_agent/agent_stub/cli/_config.py
+++ b/dify-agent/src/dify_agent/agent_stub/cli/_config.py
@@ -2,9 +2,7 @@
 
 from __future__ import annotations
 
-import json
 import os
-import re
 from dataclasses import dataclass
 from pathlib import Path
 from tempfile import TemporaryDirectory
@@ -36,7 +34,6 @@ from dify_agent.agent_stub.protocol.agent_stub import (
 
 _DEFAULT_CONFIG_BASE = Path("./.dify_conf")
 _SKILL_MD_FILENAME = "SKILL.md"
-_SAFE_ENV_VALUE = re.compile(r"^[A-Za-z0-9_./:@+-]+$")
 
 
 class ConfigSkillPullResult(BaseModel):
@@ -141,19 +138,6 @@ def pull_config_files_from_environment(
     return ConfigFilePullResult(items=items)
 
 
-def pull_config_env_from_environment(local_path: str | None = None) -> Path:
-    manifest = manifest_from_environment()
-    target_path = Path(local_path or (_DEFAULT_CONFIG_BASE / ".env")).expanduser().resolve()
-    target_path.parent.mkdir(parents=True, exist_ok=True)
-    lines: list[str] = []
-    for key in manifest.env_keys:
-        if key not in os.environ:
-            raise AgentStubValidationError(f"config env key is not present in the current environment: {key}")
-        lines.append(f"{key}={_format_env_value(os.environ[key])}")
-    target_path.write_text("\n".join(lines) + ("\n" if lines else ""), encoding="utf-8")
-    return target_path
-
-
 def pull_config_note_from_environment(local_path: str | None = None) -> Path:
     manifest = manifest_from_environment()
     target_path = Path(local_path or (_DEFAULT_CONFIG_BASE / "note.md")).expanduser().resolve()
@@ -167,22 +151,19 @@ def push_config_note_from_environment(local_path: str | None) -> AgentStubConfig
     return _push_config_from_environment(note=note)
 
 
-def push_config_env_from_environment(local_path: str | None) -> AgentStubConfigManifestResponse:
-    env_text = _read_text_input(local_path, _DEFAULT_CONFIG_BASE / ".env")
+def push_config_env_from_environment(local_path: str) -> AgentStubConfigManifestResponse:
+    env_text = _read_text_input(local_path, default_path=None)
     return _push_config_from_environment(env_text=env_text)
 
 
-def push_config_files_from_environment(paths: list[str], name: str | None) -> AgentStubConfigManifestResponse:
+def push_config_files_from_environment(paths: list[str]) -> AgentStubConfigManifestResponse:
     _require_non_empty_inputs(paths, kind="file path")
-    override_name = None if name is None else _require_config_entry_name(name, kind="file")
-    if override_name is not None and len(paths) != 1:
-        raise AgentStubValidationError("--name requires exactly one PATH")
     items = [
         _PreparedPushItem(
-            name=override_name if index == 0 and override_name is not None else _infer_name_from_path(source_path),
+            name=_infer_name_from_path(source_path),
             path=source_path,
         )
-        for index, source_path in enumerate(_resolve_input_paths(paths))
+        for source_path in _resolve_input_paths(paths)
     ]
     return _push_config_from_environment(files=[_build_file_push_item(item=item) for item in items])
 
@@ -236,10 +217,15 @@ def _push_config_from_environment(
     )
 
 
-def _read_text_input(path_value: str | None, default_path: Path) -> str:
+def _read_text_input(path_value: str | None, default_path: Path | None) -> str:
     if path_value == "-":
         return os.fdopen(os.dup(0), encoding="utf-8").read()
-    source_path = Path(path_value or default_path).expanduser().resolve()
+    if path_value is None:
+        if default_path is None:
+            raise AgentStubValidationError("local file path or '-' is required")
+        source_path = default_path.expanduser().resolve()
+    else:
+        source_path = Path(path_value).expanduser().resolve()
     if not source_path.is_file():
         raise AgentStubValidationError(f"local file not found: {source_path}")
     return source_path.read_text(encoding="utf-8")
@@ -297,17 +283,10 @@ def _require_non_empty_inputs(values: list[str], *, kind: str) -> None:
         raise AgentStubValidationError(f"at least one {kind} is required")
 
 
-def _format_env_value(value: str) -> str:
-    if _SAFE_ENV_VALUE.fullmatch(value):
-        return value
-    return json.dumps(value)
-
-
 __all__ = [
     "ConfigFilePullResult",
     "ConfigSkillPullResult",
     "manifest_from_environment",
-    "pull_config_env_from_environment",
     "pull_config_files_from_environment",
     "pull_config_note_from_environment",
     "pull_config_skills_from_environment",
diff --git a/dify-agent/src/dify_agent/agent_stub/cli/main.py b/dify-agent/src/dify_agent/agent_stub/cli/main.py
index 1c1e328e436..6ccc9865989 100644
--- a/dify-agent/src/dify_agent/agent_stub/cli/main.py
+++ b/dify-agent/src/dify_agent/agent_stub/cli/main.py
@@ -22,7 +22,6 @@ from dify_agent.agent_stub.cli._config import (
     delete_config_files_from_environment,
     delete_config_skills_from_environment,
     manifest_from_environment,
-    pull_config_env_from_environment,
     pull_config_files_from_environment,
     pull_config_note_from_environment,
     pull_config_skills_from_environment,
@@ -68,9 +67,7 @@ config_skills_app = typer.Typer(help="Pull or update config skills through the A
 config_files_app = typer.Typer(help="Pull or update config files through the Agent Stub.", rich_markup_mode=None)
 config_skill_pull_alias_app = typer.Typer(help="Pull config skills through the Agent Stub.", rich_markup_mode=None)
 config_file_pull_alias_app = typer.Typer(help="Pull config files through the Agent Stub.", rich_markup_mode=None)
-config_env_app = typer.Typer(
-    help="Pull or update config env variables visible to the current run.", rich_markup_mode=None
-)
+config_env_app = typer.Typer(help="Update config env variables visible to the current run.", rich_markup_mode=None)
 config_note_app = typer.Typer(help="Pull or update the current config note.", rich_markup_mode=None)
 drive_app = typer.Typer(help="List, pull, or push agent drive files through the Agent Stub.", rich_markup_mode=None)
 app.add_typer(file_app, name="file")
@@ -147,9 +144,13 @@ def config_skill_pull_alias(
 
 @config_skills_app.command("push")
 def config_skills_push(
-    paths: list[str] = typer.Argument(..., metavar="PATH"),
+    paths: list[str] = typer.Argument(..., metavar="PATH..."),
 ) -> None:
-    """Upload one or more local skill directories into the current config manifest."""
+    """Upload one or more local skill directories into the current config manifest.
+
+    Pass a directory such as ./skills/researcher that contains SKILL.md. Other files in that directory are
+    archived with the skill. Pushing a skill with an existing name replaces that config skill.
+    """
     _run_config_skills_push(paths=paths)
 
 
@@ -183,15 +184,15 @@ def config_file_pull_alias(
 
 @config_files_app.command("push")
 def config_files_push(
-    paths: list[str] = typer.Argument(..., metavar="PATH"),
-    name: str | None = typer.Option(
-        None,
-        "--name",
-        help="Override the target config file name when uploading exactly one local file.",
-    ),
+    paths: list[str] = typer.Argument(..., metavar="PATH..."),
 ) -> None:
-    """Upload one or more local files into the current config manifest."""
-    _run_config_files_push(paths=paths, name=name)
+    """Upload one or more local files into the current config manifest.
+
+    Pass one or more regular local files such as ./guide.md and ./policy.txt. To upload a folder, compress it
+    first and pass the archive file. The config file name is the local basename. Pushing a file with an existing config
+    name replaces that config file.
+    """
+    _run_config_files_push(paths=paths)
 
 
 @config_files_app.command("delete")
@@ -202,19 +203,15 @@ def config_files_delete(
     _run_config_files_delete(names=names)
 
 
-@config_env_app.command("pull")
-def config_env_pull(
-    local_path: str | None = typer.Option(None, "--to", help="Local dotenv file path."),
-) -> None:
-    """Export visible config env values into ./.dify_conf/.env by default."""
-    _run_config_env_pull(local_path=local_path)
-
-
 @config_env_app.command("push")
 def config_env_push(
-    local_path: str | None = typer.Argument(None, metavar="PATH|-"),
+    local_path: str = typer.Argument(..., metavar="PATH|-"),
 ) -> None:
-    """Replace visible config env values from one local dotenv file or stdin."""
+    """Set or delete config env entries from one local dotenv file or stdin.
+
+    Pass dotenv text with lines like API_KEY=value and OLD_KEY=. KEY=value sets or updates an entry, KEY= deletes
+    it, and omitted keys are unchanged.
+    """
     _run_config_env_push(local_path=local_path)
 
 
@@ -228,9 +225,13 @@ def config_note_pull(
 
 @config_note_app.command("push")
 def config_note_push(
-    local_path: str | None = typer.Argument(None, metavar="PATH|-"),
+    local_path: str | None = typer.Argument(None, metavar="[PATH|-]"),
 ) -> None:
-    """Replace the current config note from one local text file or stdin."""
+    """Replace the current config note from one local text file or stdin.
+
+    Pass a plain text or Markdown note file, or use - to read the note from stdin. When PATH is omitted, the
+    command reads ./.dify_conf/note.md.
+    """
     _run_config_note_push(local_path=local_path)
 
 
@@ -448,18 +449,6 @@ def _run_config_file_pull(*, names: list[str] | None, local_dir: str | None, jso
         typer.echo(item.path)
 
 
-def _run_config_env_pull(*, local_path: str | None) -> None:
-    try:
-        path = pull_config_env_from_environment(local_path=local_path)
-    except MissingAgentStubEnvironmentError as exc:
-        typer.echo(str(exc), err=True)
-        raise SystemExit(2) from exc
-    except AgentStubClientError as exc:
-        typer.echo(str(exc), err=True)
-        raise SystemExit(1) from exc
-    typer.echo(str(path))
-
-
 def _run_config_note_pull(*, local_path: str | None) -> None:
     try:
         path = pull_config_note_from_environment(local_path=local_path)
@@ -484,7 +473,7 @@ def _run_config_note_push(*, local_path: str | None) -> None:
     typer.echo(response.model_dump_json())
 
 
-def _run_config_env_push(*, local_path: str | None) -> None:
+def _run_config_env_push(*, local_path: str) -> None:
     try:
         response = push_config_env_from_environment(local_path=local_path)
     except MissingAgentStubEnvironmentError as exc:
@@ -496,9 +485,9 @@ def _run_config_env_push(*, local_path: str | None) -> None:
     typer.echo(response.model_dump_json())
 
 
-def _run_config_files_push(*, paths: list[str], name: str | None) -> None:
+def _run_config_files_push(*, paths: list[str]) -> None:
     try:
-        response = push_config_files_from_environment(paths=paths, name=name)
+        response = push_config_files_from_environment(paths=paths)
     except MissingAgentStubEnvironmentError as exc:
         typer.echo(str(exc), err=True)
         raise SystemExit(2) from exc
diff --git a/dify-agent/src/dify_agent/agent_stub/client/_agent_stub.py b/dify-agent/src/dify_agent/agent_stub/client/_agent_stub.py
index b76e74cd354..704b099a4de 100644
--- a/dify-agent/src/dify_agent/agent_stub/client/_agent_stub.py
+++ b/dify-agent/src/dify_agent/agent_stub/client/_agent_stub.py
@@ -289,7 +289,7 @@ def request_agent_stub_config_env_update_sync(
     timeout: float | httpx.Timeout = 30.0,
     sync_http_client: httpx.Client | None = None,
 ):
-    """Replace or delete config env entries through the HTTP transport."""
+    """Set or delete config env entries through the HTTP transport."""
     endpoint = _parse_endpoint(url)
     if endpoint.is_grpc:
         raise AgentStubValidationError("Agent Stub config operations require an HTTP Agent Stub URL")
diff --git a/dify-agent/src/dify_agent/agent_stub/client/_agent_stub_http.py b/dify-agent/src/dify_agent/agent_stub/client/_agent_stub_http.py
index 49c7c83ae94..beb777c58bd 100644
--- a/dify-agent/src/dify_agent/agent_stub/client/_agent_stub_http.py
+++ b/dify-agent/src/dify_agent/agent_stub/client/_agent_stub_http.py
@@ -338,7 +338,7 @@ def request_agent_stub_config_env_update_http_sync(
     timeout: float | httpx.Timeout = 30.0,
     sync_http_client: httpx.Client | None = None,
 ) -> dict[str, object]:
-    """Replace or delete config env entries through the HTTP Agent Stub endpoint."""
+    """Set or delete config env entries through the HTTP Agent Stub endpoint."""
     request_model = AgentStubConfigEnvUpdateRequest(env_text=env_text)
     response = _send_agent_stub_json(
         method="PATCH",
diff --git a/dify-agent/src/dify_agent/agent_stub/server/agent_stub_files.py b/dify-agent/src/dify_agent/agent_stub/server/agent_stub_files.py
index cedbb6e862d..06cb78d32fa 100644
--- a/dify-agent/src/dify_agent/agent_stub/server/agent_stub_files.py
+++ b/dify-agent/src/dify_agent/agent_stub/server/agent_stub_files.py
@@ -84,8 +84,8 @@ class DifyApiAgentStubFileRequestHandler:
     """Call Dify API inner file request endpoints on behalf of the sandbox.
 
     The upload path calls ``/inner/api/upload/file/request`` and injects the
-    authenticated execution context's ``tenant_id`` and ``user_id`` along with
-    the requested filename and mimetype. The download path calls
+    authenticated execution context's ``tenant_id``, ``user_id``, and optional
+    ``conversation_id`` along with the requested filename and mimetype. The download path calls
     ``/inner/api/download/file/request`` and injects ``tenant_id``,
     ``user_id``, ``user_from``, and ``invoke_from`` plus the validated public
     file mapping.
@@ -126,6 +126,7 @@ class DifyApiAgentStubFileRequestHandler:
             "user_id": execution_context.user_id,
             "filename": request.filename,
             "mimetype": request.mimetype,
+            "conversation_id": execution_context.conversation_id,
         }
         data = await self._post_inner_api("/inner/api/upload/file/request", payload)
         upload_url = data.get("url")
diff --git a/dify-agent/src/dify_agent/layers/config/layer.py b/dify-agent/src/dify_agent/layers/config/layer.py
index 5bec83d6173..d914f599efb 100644
--- a/dify-agent/src/dify_agent/layers/config/layer.py
+++ b/dify-agent/src/dify_agent/layers/config/layer.py
@@ -19,6 +19,7 @@ from dify_agent.layers.config.configs import (
 from dify_agent.layers.shell.layer import DifyShellLayer
 
 _CONFIG_CONTEXT_HEADING = "Agent config context from the current Agent Soul:"
+_CONFIG_CONTEXT_COMMAND = "$ dify-agent config manifest"
 _CONFIG_CLI_USAGE_PROMPT = """Agent config CLI usage is available inside shell jobs. The command help below is generated
 from the same `dify-agent` CLI definitions available in shell jobs.
 
@@ -30,7 +31,6 @@ _CONFIG_CLI_HELP_COMMANDS: dict[str, tuple[str, ...]] = {
     "dify-agent config manifest --help": ("config", "manifest"),
     "dify-agent config skills pull --help": ("config", "skills", "pull"),
     "dify-agent config files pull --help": ("config", "files", "pull"),
-    "dify-agent config env pull --help": ("config", "env", "pull"),
     "dify-agent config note pull --help": ("config", "note", "pull"),
 }
 _CONFIG_CLI_MUTATION_HELP_COMMANDS: dict[str, tuple[str, ...]] = {
@@ -124,7 +124,9 @@ class DifyConfigLayer(PlainLayer[DifyConfigDeps, DifyConfigLayerConfig, DifyConf
     def build_suffix_prompt(self) -> str:
         sections: list[str] = []
         if self.runtime_state.config_context_json:
-            sections.append(f"{_CONFIG_CONTEXT_HEADING}\n{self.runtime_state.config_context_json}")
+            sections.append(
+                f"{_CONFIG_CONTEXT_COMMAND}\n{_CONFIG_CONTEXT_HEADING}\n{self.runtime_state.config_context_json}"
+            )
         usage_lines = [_CONFIG_CLI_USAGE_PROMPT]
         if cli_help := self._format_config_cli_help():
             usage_lines.append(cli_help)
diff --git a/dify-agent/src/dify_agent/layers/dify_plugin/tools_layer.py b/dify-agent/src/dify_agent/layers/dify_plugin/tools_layer.py
index dd6c0c3cfb2..6e43566e589 100644
--- a/dify-agent/src/dify_agent/layers/dify_plugin/tools_layer.py
+++ b/dify-agent/src/dify_agent/layers/dify_plugin/tools_layer.py
@@ -7,19 +7,22 @@ only validates hidden/manual inputs, prepares invocation arguments, and maps
 daemon responses into agent-friendly observations.
 
 Like the LLM layer, this layer never owns live HTTP clients. The runtime passes
-the FastAPI lifespan-owned shared client into ``get_tools`` so the layer can
-build Pydantic AI tool adapters on demand.
+the FastAPI lifespan-owned shared plugin-daemon and Dify API clients into
+``get_tools`` so the layer can build Pydantic AI tool adapters on demand.
 """
 
 from __future__ import annotations
 
 from copy import deepcopy
 import json
+import mimetypes
 from collections.abc import Mapping, Sequence
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import ClassVar
+from urllib.parse import urlparse
 
 import httpx
+from pydantic import BaseModel, ConfigDict, JsonValue, ValidationError
 from pydantic_ai import RunContext, Tool
 from pydantic_ai.tools import ToolDefinition
 from typing_extensions import Self, override
@@ -38,7 +41,9 @@ from dify_agent.layers.dify_plugin.tool_client import (
     DifyPluginToolClientError,
     DifyPluginToolInvokeMessage,
 )
+from dify_agent.layers.execution_context.configs import DifyExecutionContextLayerConfig
 from dify_agent.layers.execution_context.layer import DifyExecutionContextLayer
+from dify_agent.layers.shell.layer import DifyShellLayer
 
 
 # Plugin tools intentionally do not expose a per-tool strictness override in the
@@ -46,12 +51,107 @@ from dify_agent.layers.execution_context.layer import DifyExecutionContextLayer
 # registers those tools in loose mode so daemon tool invocation stays tolerant of
 # plugin schema differences and older API-prepared payloads.
 PLUGIN_TOOL_STRICT = False
+PLUGIN_FILE_DEFAULT_TYPE = "custom"
+_FILE_UPLOAD_BEGIN = "<<<DIFY_PLUGIN_TOOL_FILE_UPLOAD_BEGIN>>>"
+_FILE_UPLOAD_END = "<<<DIFY_PLUGIN_TOOL_FILE_UPLOAD_END>>>"
+_FILE_UPLOAD_TIMEOUT_SECONDS = 60.0
+_SUPPORTED_REMOTE_URL_PREFIXES = ("http://", "https://")
 
 
 class DifyPluginToolsDeps(LayerDeps):
     """Dependencies required by ``DifyPluginToolsLayer``."""
 
     execution_context: DifyExecutionContextLayer  # pyright: ignore[reportUninitializedInstanceVariable]
+    shell: DifyShellLayer | None  # pyright: ignore[reportUninitializedInstanceVariable]
+
+
+class DifyPluginToolsClientConfigurationError(ValueError):
+    """Raised when local plugin-tool file conversion preconditions are missing."""
+
+
+class _BackwardsInvocationEnvelope(BaseModel):
+    data: JsonValue = None
+    error: str | None = None
+
+    model_config: ClassVar[ConfigDict] = ConfigDict(extra="ignore")
+
+
+class _DownloadFileResponse(BaseModel):
+    filename: str
+    mime_type: str | None = None
+    size: int
+    download_url: str
+
+    model_config: ClassVar[ConfigDict] = ConfigDict(extra="forbid")
+
+
+@dataclass(slots=True)
+class _DifyPluginToolFileClient:
+    """Resolve Dify file mappings into signed URLs for plugin file parameters."""
+
+    base_url: str
+    api_key: str = field(repr=False)
+    http_client: httpx.AsyncClient = field(repr=False)
+
+    def __post_init__(self) -> None:
+        self.base_url = self.base_url.rstrip("/")
+
+    async def request_download(
+        self,
+        *,
+        execution_context: DifyExecutionContextLayerConfig,
+        file_mapping: Mapping[str, object],
+    ) -> _DownloadFileResponse:
+        missing_fields = []
+        if execution_context.user_id is None:
+            missing_fields.append("user_id")
+        if execution_context.user_from is None:
+            missing_fields.append("user_from")
+        if missing_fields:
+            missing = ", ".join(missing_fields)
+            raise DifyPluginToolsClientConfigurationError(
+                f"Missing required execution context fields for file parameters: {missing}."
+            )
+
+        payload = {
+            "tenant_id": execution_context.tenant_id,
+            "user_id": execution_context.user_id,
+            "user_from": execution_context.user_from,
+            "invoke_from": execution_context.invoke_from,
+            "file": dict(file_mapping),
+        }
+        try:
+            response = await self.http_client.post(
+                f"{self.base_url}/inner/api/download/file/request",
+                headers={
+                    "X-Inner-Api-Key": self.api_key,
+                    "Content-Type": "application/json",
+                },
+                json=payload,
+            )
+        except (httpx.InvalidURL, httpx.UnsupportedProtocol) as exc:
+            raise DifyPluginToolClientError(f"Dify API file download is misconfigured: {exc}") from exc
+        except httpx.TimeoutException as exc:
+            raise DifyPluginToolClientError("Dify API file download request timed out.") from exc
+        except httpx.RequestError as exc:
+            raise DifyPluginToolClientError(f"Dify API file download request failed: {exc}") from exc
+
+        if response.status_code >= 400:
+            raise DifyPluginToolClientError(
+                response.text or f"HTTP {response.status_code}", status_code=response.status_code
+            )
+        try:
+            envelope = _BackwardsInvocationEnvelope.model_validate_json(response.text)
+        except ValidationError as exc:
+            raise DifyPluginToolClientError("Invalid Dify API file download response.") from exc
+        if envelope.error:
+            raise DifyPluginToolClientError(envelope.error, status_code=response.status_code)
+        if not isinstance(envelope.data, dict):
+            raise DifyPluginToolClientError("Dify API file download response is missing data.")
+        try:
+            return _DownloadFileResponse.model_validate(envelope.data)
+        except ValidationError as exc:
+            raise DifyPluginToolClientError("Invalid Dify API file download data.") from exc
 
 
 @dataclass(slots=True)
@@ -61,17 +161,52 @@ class DifyPluginToolsLayer(PlainLayer[DifyPluginToolsDeps, DifyPluginToolsLayerC
     type_id: ClassVar[str | None] = DIFY_PLUGIN_TOOLS_LAYER_TYPE_ID
 
     config: DifyPluginToolsLayerConfig
+    inner_api_url: str
+    inner_api_key: str
 
     @classmethod
     @override
     def from_config(cls, config: DifyPluginToolsLayerConfig) -> Self:
-        """Create the tools layer from validated public config."""
-        return cls(config=DifyPluginToolsLayerConfig.model_validate(config))
+        """Reject construction without server-injected Dify API settings."""
+        del config
+        raise TypeError("DifyPluginToolsLayer requires server-side Dify API settings and must use a provider factory.")
 
-    async def get_tools(self, *, http_client: httpx.AsyncClient) -> list[Tool[object]]:
+    @classmethod
+    def from_config_with_settings(
+        cls,
+        config: DifyPluginToolsLayerConfig,
+        *,
+        inner_api_url: str,
+        inner_api_key: str,
+    ) -> Self:
+        return cls(
+            config=DifyPluginToolsLayerConfig.model_validate(config),
+            inner_api_url=inner_api_url,
+            inner_api_key=inner_api_key,
+        )
+
+    async def get_tools(
+        self,
+        *,
+        http_client: httpx.AsyncClient,
+        dify_api_http_client: httpx.AsyncClient,
+    ) -> list[Tool[object]]:
         """Build Pydantic AI tool adapters from prepared plugin tool config."""
+        if dify_api_http_client.is_closed:
+            raise RuntimeError("DifyPluginToolsLayer.get_tools() requires an open shared Dify API HTTP client.")
+
         tool_clients: dict[str, DifyPluginDaemonToolClient] = {}
         tools: list[Tool[object]] = []
+        file_client = _DifyPluginToolFileClient(
+            base_url=self.inner_api_url,
+            api_key=self.inner_api_key,
+            http_client=dify_api_http_client,
+        )
+        file_context = _PluginToolFileContext(
+            file_client=file_client,
+            execution_context=self.deps.execution_context.config,
+            shell=self.deps.shell,
+        )
 
         for tool_config in self.config.tools:
             client = tool_clients.get(tool_config.plugin_id)
@@ -89,6 +224,7 @@ class DifyPluginToolsLayer(PlainLayer[DifyPluginToolsDeps, DifyPluginToolsLayerC
                     client=client,
                     tool_config=tool_config,
                     effective_parameters=effective_parameters,
+                    file_context=file_context,
                 )
             )
 
@@ -117,6 +253,7 @@ def _build_pydantic_ai_tool(
     client: DifyPluginDaemonToolClient,
     tool_config: DifyPluginToolConfig,
     effective_parameters: Sequence[DifyPluginToolParameter],
+    file_context: "_PluginToolFileContext",
 ) -> Tool[object]:
     tool_name = tool_config.name or tool_config.tool_name
     tool_description = tool_config.description or tool_name
@@ -124,7 +261,12 @@ def _build_pydantic_ai_tool(
 
     async def invoke_tool(_ctx: RunContext[object], **tool_arguments: object) -> str:
         try:
-            merged_arguments = _prepare_tool_arguments(effective_parameters, tool_config, tool_arguments)
+            merged_arguments = await _prepare_tool_arguments(
+                effective_parameters,
+                tool_config,
+                tool_arguments,
+                file_context=file_context,
+            )
             messages = await client.invoke(
                 provider=tool_config.provider,
                 tool_name=tool_config.tool_name,
@@ -162,10 +304,12 @@ def _build_pydantic_ai_tool(
     )
 
 
-def _prepare_tool_arguments(
+async def _prepare_tool_arguments(
     effective_parameters: Sequence[DifyPluginToolParameter],
     tool_config: DifyPluginToolConfig,
     tool_arguments: Mapping[str, object],
+    *,
+    file_context: "_PluginToolFileContext",
 ) -> dict[str, object]:
     """Build the daemon invocation payload from prepared config + model args.
 
@@ -193,14 +337,103 @@ def _prepare_tool_arguments(
             raise ValueError(f"tool parameter {parameter.name} not found in tool config")
         else:
             continue
-        prepared_arguments[parameter.name] = _cast_tool_parameter_value(parameter.type, value)
+        prepared_arguments[parameter.name] = await _cast_tool_parameter_value(
+            parameter.type,
+            value,
+            file_context=file_context,
+        )
 
     for key, value in merged_arguments.items():
         prepared_arguments.setdefault(key, value)
     return prepared_arguments
 
 
-def _cast_tool_parameter_value(parameter_type: DifyPluginToolParameterType, value: object) -> object:
+@dataclass(slots=True)
+class _PluginToolFileContext:
+    file_client: _DifyPluginToolFileClient
+    execution_context: DifyExecutionContextLayerConfig
+    shell: DifyShellLayer | None
+
+    async def to_plugin_file_parameter(self, value: object) -> dict[str, object]:
+        if isinstance(value, str):
+            if _is_remote_url(value):
+                return _plugin_file_parameter_from_url(value)
+            mapping = await self._upload_sandbox_path(value)
+            return await self._plugin_file_parameter_from_mapping(mapping)
+
+        if isinstance(value, Mapping):
+            if _is_plugin_file_parameter(value):
+                return _normalize_plugin_file_parameter(value)
+            return await self._plugin_file_parameter_from_mapping(value)
+
+        raise ValueError("file parameter must be an HTTP(S) URL, sandbox path, or file mapping.")
+
+    async def _upload_sandbox_path(self, path: str) -> dict[str, object]:
+        if self.shell is None:
+            raise ValueError("sandbox file path parameters require an active shell layer.")
+        script = (
+            "python - <<'PY'\n"
+            "import json\n"
+            "import subprocess\n"
+            "import sys\n\n"
+            f"command = ['dify-agent', 'file', 'upload', {json.dumps(path)}]\n"
+            "completed = subprocess.run(command, capture_output=True, text=True, check=False)\n"
+            "if completed.returncode != 0:\n"
+            "    print(completed.stderr or completed.stdout or f'upload exited with code {completed.returncode}', file=sys.stderr)\n"
+            "    sys.exit(completed.returncode or 1)\n"
+            "try:\n"
+            "    payload = json.loads(completed.stdout)\n"
+            "except ValueError as exc:\n"
+            "    print(f'upload returned invalid JSON: {exc}', file=sys.stderr)\n"
+            "    sys.exit(1)\n"
+            f"print('{_FILE_UPLOAD_BEGIN}' + json.dumps(payload, separators=(',', ':')) + '{_FILE_UPLOAD_END}')\n"
+            "PY"
+        )
+        result = await self.shell.run_remote_script_complete(
+            script,
+            timeout=_FILE_UPLOAD_TIMEOUT_SECONDS,
+            inject_agent_stub_env=True,
+        )
+        if result.exit_code != 0:
+            raise ValueError(
+                "sandbox file upload failed: "
+                + f"{result.status} exit_code={result.exit_code} output_complete={result.output_complete} "
+                + result.output.strip()
+            )
+        return _parse_uploaded_file_mapping(result.output)
+
+    async def _plugin_file_parameter_from_mapping(self, mapping: Mapping[str, object]) -> dict[str, object]:
+        transfer_method = mapping.get("transfer_method")
+        if transfer_method == "remote_url":
+            url = mapping.get("url")
+            if not isinstance(url, str) or not url:
+                raise ValueError("remote_url file mapping requires url.")
+            return _plugin_file_parameter_from_url(url)
+
+        if transfer_method not in {"local_file", "tool_file", "datasource_file"}:
+            raise ValueError(
+                "file mapping transfer_method must be local_file, tool_file, datasource_file, or remote_url."
+            )
+        reference = mapping.get("reference")
+        if not isinstance(reference, str) or not reference:
+            raise ValueError(f"{transfer_method} file mapping requires reference.")
+
+        download = await self.file_client.request_download(
+            execution_context=self.execution_context,
+            file_mapping={
+                "transfer_method": transfer_method,
+                "reference": reference,
+            },
+        )
+        return _plugin_file_parameter_from_download(download)
+
+
+async def _cast_tool_parameter_value(
+    parameter_type: DifyPluginToolParameterType,
+    value: object,
+    *,
+    file_context: _PluginToolFileContext,
+) -> object:
     """Cast prepared tool argument values into daemon-facing wire shapes.
 
     The API side prepares declaration metadata, but the actual invocation payload
@@ -235,13 +468,14 @@ def _cast_tool_parameter_value(parameter_type: DifyPluginToolParameterType, valu
                 return float(value) if "." in value else int(value)
             return value
         case DifyPluginToolParameterType.SYSTEM_FILES | DifyPluginToolParameterType.FILES:
-            return value if isinstance(value, list) else [value]
+            values = value if isinstance(value, list) else [value]
+            return [await file_context.to_plugin_file_parameter(item) for item in values]
         case DifyPluginToolParameterType.FILE:
             if isinstance(value, list):
                 if len(value) != 1:
                     raise ValueError("This parameter only accepts one file but got multiple files while invoking.")
-                return value[0]
-            return value
+                return await file_context.to_plugin_file_parameter(value[0])
+            return await file_context.to_plugin_file_parameter(value)
         case DifyPluginToolParameterType.MODEL_SELECTOR | DifyPluginToolParameterType.APP_SELECTOR:
             if not isinstance(value, dict):
                 raise ValueError("The selector must be a dictionary.")
@@ -276,6 +510,112 @@ def _cast_tool_parameter_value(parameter_type: DifyPluginToolParameterType, valu
     raise AssertionError(f"Unsupported tool parameter type: {parameter_type}")
 
 
+def _is_remote_url(value: str) -> bool:
+    return value.lower().startswith(_SUPPORTED_REMOTE_URL_PREFIXES)
+
+
+def _is_plugin_file_parameter(value: Mapping[str, object]) -> bool:
+    url = value.get("url")
+    return isinstance(url, str) and bool(url) and "transfer_method" not in value
+
+
+def _normalize_plugin_file_parameter(value: Mapping[str, object]) -> dict[str, object]:
+    url = value.get("url")
+    if not isinstance(url, str) or not url:
+        raise ValueError("plugin file parameter requires url.")
+    return _plugin_file_parameter(
+        url=url,
+        filename=value.get("filename"),
+        mime_type=value.get("mime_type"),
+        extension=value.get("extension"),
+        size=value.get("size"),
+        file_type=value.get("type"),
+    )
+
+
+def _plugin_file_parameter_from_url(url: str) -> dict[str, object]:
+    if not _is_remote_url(url):
+        raise ValueError("remote file URL must start with http:// or https://.")
+    filename = _filename_from_url(url)
+    mime_type = mimetypes.guess_type(filename or url)[0] if filename else None
+    return _plugin_file_parameter(
+        url=url,
+        filename=filename,
+        mime_type=mime_type,
+        extension=_extension_from_filename(filename),
+        size=-1,
+        file_type=PLUGIN_FILE_DEFAULT_TYPE,
+    )
+
+
+def _plugin_file_parameter_from_download(download: _DownloadFileResponse) -> dict[str, object]:
+    return _plugin_file_parameter(
+        url=download.download_url,
+        filename=download.filename,
+        mime_type=download.mime_type,
+        extension=_extension_from_filename(download.filename),
+        size=download.size,
+        file_type=PLUGIN_FILE_DEFAULT_TYPE,
+    )
+
+
+def _plugin_file_parameter(
+    *,
+    url: str,
+    filename: object,
+    mime_type: object,
+    extension: object,
+    size: object,
+    file_type: object,
+) -> dict[str, object]:
+    normalized: dict[str, object] = {
+        "dify_model_identity": "__dify__file__",
+        "type": file_type if isinstance(file_type, str) and file_type else PLUGIN_FILE_DEFAULT_TYPE,
+        "url": url,
+    }
+    if isinstance(filename, str) and filename:
+        normalized["filename"] = filename
+    if isinstance(mime_type, str) and mime_type:
+        normalized["mime_type"] = mime_type
+    if isinstance(extension, str) and extension:
+        normalized["extension"] = extension
+    if isinstance(size, int):
+        normalized["size"] = size
+    return normalized
+
+
+def _filename_from_url(url: str) -> str | None:
+    parsed = urlparse(url)
+    filename = parsed.path.rsplit("/", 1)[-1]
+    return filename or None
+
+
+def _extension_from_filename(filename: str | None) -> str | None:
+    if not filename or "." not in filename:
+        return None
+    suffix = "." + filename.rsplit(".", 1)[-1]
+    return suffix if len(suffix) > 1 else None
+
+
+def _parse_uploaded_file_mapping(output: str) -> dict[str, object]:
+    begin_index = output.find(_FILE_UPLOAD_BEGIN)
+    end_index = output.find(_FILE_UPLOAD_END, begin_index + len(_FILE_UPLOAD_BEGIN))
+    if begin_index < 0 or end_index < 0:
+        raise ValueError("sandbox file upload did not return a framed file mapping.")
+    raw_payload = output[begin_index + len(_FILE_UPLOAD_BEGIN) : end_index]
+    try:
+        payload = json.loads(raw_payload)
+    except json.JSONDecodeError as exc:
+        raise ValueError("sandbox file upload returned invalid file mapping JSON.") from exc
+    if not isinstance(payload, dict):
+        raise ValueError("sandbox file upload returned a non-object file mapping.")
+    transfer_method = payload.get("transfer_method")
+    reference = payload.get("reference")
+    if transfer_method != "tool_file" or not isinstance(reference, str) or not reference:
+        raise ValueError("sandbox file upload returned an invalid tool_file mapping.")
+    return {"transfer_method": "tool_file", "reference": reference}
+
+
 def _tool_error_text(*, tool_name: str, error: DifyPluginToolClientError) -> str:
     """Map expected daemon/tool failures into agent-visible observation text.
 
diff --git a/dify-agent/src/dify_agent/layers/shell/layer.py b/dify-agent/src/dify_agent/layers/shell/layer.py
index d514025aa45..7346acc7b33 100644
--- a/dify-agent/src/dify_agent/layers/shell/layer.py
+++ b/dify-agent/src/dify_agent/layers/shell/layer.py
@@ -1,4 +1,15 @@
-"""Shell runtime layer backed by a live shell provider resource."""
+"""Shell runtime layer backed by a live shell provider resource.
+
+Shell command execution requires a bound execution-context layer with a safe
+``agent_id``. The layer uses the current bound execution context to run
+commands with ``HOME=/home/<agent_id>`` and a home-rooted workspace path. The
+persisted runtime state intentionally keeps the historical
+``~/workspace/<session>`` identity so existing session snapshots stay
+compatible while live command execution no longer depends on the sandbox user's
+ambient home directory. Entering or re-entering the layer re-ensures the live
+home/workspace directories for the currently bound ``agent_id`` before user
+commands are sent.
+"""
 
 from __future__ import annotations
 
@@ -10,13 +21,22 @@ import re
 import secrets
 import time
 from dataclasses import dataclass
-from typing import ClassVar, Literal, NotRequired, TypedDict
+from typing import ClassVar, Literal, NotRequired, Protocol, TypedDict, runtime_checkable
 
 from pydantic import BaseModel, ConfigDict, Field, NonNegativeInt, field_validator, model_validator
 from pydantic_ai import Tool
 from typing_extensions import Self, override
 
-from agenton.layers import LayerDeps, PydanticAILayer, PydanticAIPrompt, PydanticAITool
+from agenton.layers import (
+    EmptyLayerConfig,
+    EmptyRuntimeState,
+    LayerDeps,
+    NoLayerDeps,
+    PlainLayer,
+    PydanticAILayer,
+    PydanticAIPrompt,
+    PydanticAITool,
+)
 from dify_agent.adapters.shell.protocols import (
     CompleteShellCommandResult,
     ShellCommandProtocol,
@@ -26,20 +46,35 @@ from dify_agent.adapters.shell.protocols import (
     ShellResourceProtocol,
 )
 from dify_agent.agent_stub.server.shell_agent_stub_env import ShellAgentStubTokenFactory, build_shell_agent_stub_env
-from dify_agent.layers.execution_context.layer import DifyExecutionContextLayer
 from dify_agent.layers.shell.configs import DIFY_SHELL_LAYER_TYPE_ID, DifyShellLayerConfig
 from dify_agent.layers.shell.output_text import normalized_output_text, utf8_prefix, utf8_suffix
 
+try:
+    from dify_agent.layers.execution_context.layer import DifyExecutionContextLayer
+except ModuleNotFoundError:
+
+    class DifyExecutionContextLayer(PlainLayer[NoLayerDeps, EmptyLayerConfig, EmptyRuntimeState]):
+        """Minimal fallback for shell-only imports without server extras installed."""
+
+
 logger = logging.getLogger(__name__)
 
+
+@runtime_checkable
+class _HasErrorCode(Protocol):
+    code: object
+
+
 DEFAULT_TIMEOUT_SECONDS = 30.0
 DEFAULT_TERMINATE_GRACE_SECONDS = 10.0
 _WORKSPACE_ROOT = "~/workspace"
+_WORKSPACE_DIR_NAME = "workspace"
 _WORKSPACE_COLLISION_EXIT_CODE = 17
 _SESSION_TIME_HEX_MASK = 0xFFFFF
 _SESSION_RANDOM_HEX_LENGTH = 2
 _SESSION_ID_ATTEMPT_LIMIT = 256
 _SESSION_ID_PATTERN = re.compile(r"^[0-9a-f]{7}$")
+_AGENT_HOME_SEGMENT_PATTERN = re.compile(r"^[A-Za-z0-9._-]+$")
 _SHELL_OUTPUT_PROMPT_EDGE_BYTES = 8 * 1024
 _SHELLCTL_OUTPUT_LIMIT_BYTES = 2 * _SHELL_OUTPUT_PROMPT_EDGE_BYTES
 _REMOTE_COMPLETE_OUTPUT_MAX_BYTES = 1024 * 1024
@@ -123,6 +158,8 @@ from rich import print
 
 response = httpx.get("https://example.com", timeout=10)
 print(f"[green]status:[/green] {response.status_code}")"""
+_SHELL_LAYER_SUFFIX_PROMPT = """Environment variables may contain API keys, tokens, or credentials.
+You may refer to environment variable names when needed."""
 
 
 class ShellToolErrorObservation(TypedDict):
@@ -219,6 +256,11 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
     def prefix_prompts(self) -> Sequence[PydanticAIPrompt[object]]:
         return [_shell_layer_prefix_prompt]
 
+    @property
+    @override
+    def suffix_prompts(self) -> Sequence[PydanticAIPrompt[object]]:
+        return [_shell_layer_suffix_prompt]
+
     @property
     @override
     def tools(self) -> Sequence[PydanticAITool[object]]:
@@ -248,7 +290,7 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
         session_id: str | None = None
         try:
             session_id, workspace_cwd = await self._allocate_workspace()
-            await self._bootstrap_workspace(workspace_cwd)
+            await self._bootstrap_workspace(session_id)
         except BaseException:
             if session_id is not None:
                 await self._cleanup_workspace_best_effort(session_id)
@@ -264,7 +306,8 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
     @override
     async def on_context_resume(self) -> None:
         _ = self._require_resource()
-        _ = self._require_session_identity()
+        session_id, _workspace_cwd = self._require_session_identity()
+        await self._ensure_live_workspace_exists(session_id)
 
     @override
     async def on_context_suspend(self) -> None:
@@ -295,7 +338,7 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
             result = await self._require_resource().commands.run(
                 _wrap_user_script(script, self.config),
                 cwd=self._require_workspace_cwd(),
-                env=self._build_user_shell_run_env(),
+                env=self._build_shell_command_env(include_agent_stub_env=True),
                 timeout=timeout,
             )
             observation = await render_prompt_observation_from_result(
@@ -316,7 +359,9 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
                 observation.text,
             )
         except (RuntimeError, ValueError) as exc:
-            return _tool_error(str(exc))
+            return _tool_error_from_exception(exc)
+        except Exception as exc:
+            return _tool_unexpected_error("shell_run", exc, session_id=self.runtime_state.session_id)
 
     async def _tool_wait(self, job_id: str, timeout: float = DEFAULT_TIMEOUT_SECONDS) -> ShellRunToolResult:
         try:
@@ -340,7 +385,9 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
                 observation.text,
             )
         except (RuntimeError, ValueError) as exc:
-            return _tool_error(str(exc), job_id=job_id)
+            return _tool_error_from_exception(exc, job_id=job_id)
+        except Exception as exc:
+            return _tool_unexpected_error("shell_wait", exc, session_id=self.runtime_state.session_id, job_id=job_id)
 
     async def _tool_input(self, job_id: str, text: str, timeout: float = DEFAULT_TIMEOUT_SECONDS) -> ShellRunToolResult:
         try:
@@ -364,7 +411,9 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
                 observation.text,
             )
         except (RuntimeError, ValueError) as exc:
-            return _tool_error(str(exc), job_id=job_id)
+            return _tool_error_from_exception(exc, job_id=job_id)
+        except Exception as exc:
+            return _tool_unexpected_error("shell_input", exc, session_id=self.runtime_state.session_id, job_id=job_id)
 
     async def _tool_interrupt(
         self,
@@ -378,6 +427,9 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
             self._remember_job_offset(result.job_id, result.offset)
             output_path: str | None = None
             try:
+                # Once the interrupt itself succeeds, resolving the output path is
+                # best-effort metadata enrichment and must not turn the interrupt
+                # into a failed tool result.
                 output_path = (await self._require_resource().commands.tail(job_id)).output_path
             except (RuntimeError, ValueError) as exc:
                 logger.warning(
@@ -386,6 +438,12 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
                     self.runtime_state.session_id,
                     exc,
                 )
+            except Exception:
+                logger.exception(
+                    "Failed to fetch output path for interrupted shell job %s in session %s",
+                    job_id,
+                    self.runtime_state.session_id,
+                )
             return _tagged_shell_observation(
                 _metadata_dict(
                     job_id=result.job_id,
@@ -397,7 +455,11 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
                 "Job was interrupted.",
             )
         except (RuntimeError, ValueError) as exc:
-            return _tool_error(str(exc), job_id=job_id)
+            return _tool_error_from_exception(exc, job_id=job_id)
+        except Exception as exc:
+            return _tool_unexpected_error(
+                "shell_interrupt", exc, session_id=self.runtime_state.session_id, job_id=job_id
+            )
 
     async def run_remote_script_complete(
         self,
@@ -407,16 +469,14 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
         max_output_bytes: int = _REMOTE_COMPLETE_OUTPUT_MAX_BYTES,
         inject_agent_stub_env: bool = False,
     ) -> CompleteRemoteCommandResult:
-        env = None
-        if inject_agent_stub_env:
-            env = self._build_user_shell_run_env()
-            if env is None:
-                raise RuntimeError("Agent Stub environment injection is not available for this shell session.")
         return await execute_complete_with_commands(
             self._require_resource().commands,
             script,
             cwd=self._require_workspace_cwd(),
-            env=env,
+            env=self._build_shell_command_env(
+                include_agent_stub_env=inject_agent_stub_env,
+                require_agent_stub_env=inject_agent_stub_env,
+            ),
             timeout=timeout,
             max_output_bytes=max_output_bytes,
         )
@@ -448,10 +508,11 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
             return session_id, _workspace_cwd(session_id)
         raise RuntimeError("Failed to allocate a unique shell workspace session id after 256 attempts.")
 
-    async def _bootstrap_workspace(self, workspace_cwd: str) -> None:
+    async def _bootstrap_workspace(self, session_id: str) -> None:
         bootstrap_script = _workspace_bootstrap_script(self.config)
         if not bootstrap_script:
             return
+        workspace_cwd = _workspace_cwd_for_home(self._shell_home_dir(), session_id)
         result = await self._run_internal_script_complete(bootstrap_script, cwd=workspace_cwd)
         if result.exit_code != 0 or not result.output_complete:
             raise RuntimeError(
@@ -464,6 +525,14 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
         except (RuntimeError, ValueError) as exc:
             logger.warning("Failed to remove shell workspace for session %s after create failure: %s", session_id, exc)
 
+    async def _ensure_live_workspace_exists(self, session_id: str) -> None:
+        result = await self._run_internal_script_complete(_workspace_ensure_script(session_id=session_id), cwd=None)
+        if result.exit_code != 0 or not result.output_complete:
+            raise RuntimeError(
+                f"Failed to ensure shell workspace {_workspace_cwd_for_home(self._shell_home_dir(), session_id)} "
+                + f"exists: {result.status} exit_code={result.exit_code}"
+            )
+
     async def _run_internal_script_complete(
         self,
         script: str,
@@ -474,7 +543,7 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
             self._require_resource().commands,
             script,
             cwd=cwd,
-            env=None,
+            env=self._build_shell_command_env(include_agent_stub_env=False),
             timeout=DEFAULT_TIMEOUT_SECONDS,
             max_output_bytes=_REMOTE_COMPLETE_OUTPUT_MAX_BYTES,
         )
@@ -488,8 +557,8 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
         return self._shell_resource
 
     def _require_workspace_cwd(self) -> str:
-        _session_id, workspace_cwd = self._require_session_identity()
-        return workspace_cwd
+        session_id, _workspace_cwd = self._require_session_identity()
+        return _workspace_cwd_for_home(self._shell_home_dir(), session_id)
 
     def _require_session_identity(self) -> tuple[str, str]:
         identity = self._try_session_identity()
@@ -545,16 +614,45 @@ class DifyShellLayer(PydanticAILayer[DifyShellLayerDeps, object, DifyShellLayerC
         self.runtime_state.job_offsets = {}
         self.runtime_state.job_ids = []
 
-    def _build_user_shell_run_env(self) -> dict[str, str] | None:
+    def _shell_home_dir(self) -> str:
+        return _shell_home_dir_for_agent_id(self._require_current_execution_agent_id())
+
+    def _current_execution_agent_id(self) -> str | None:
         execution_context_layer = self.deps.execution_context
         execution_context = execution_context_layer.config if execution_context_layer is not None else None
-        return build_shell_agent_stub_env(
+        return execution_context.agent_id if execution_context is not None else None
+
+    def _require_current_execution_agent_id(self) -> str:
+        agent_id = self._current_execution_agent_id()
+        if agent_id is None:
+            raise ValueError("ShellLayer command execution requires execution_context.agent_id.")
+        return _validated_agent_home_segment(agent_id)
+
+    def _build_shell_command_env(
+        self,
+        *,
+        include_agent_stub_env: bool,
+        require_agent_stub_env: bool = False,
+    ) -> dict[str, str]:
+        env = _shell_config_env(self.config)
+        env["HOME"] = self._shell_home_dir()
+        if not include_agent_stub_env:
+            return env
+        execution_context_layer = self.deps.execution_context
+        execution_context = execution_context_layer.config if execution_context_layer is not None else None
+        agent_stub_env = build_shell_agent_stub_env(
             agent_stub_api_base_url=self.agent_stub_api_base_url,
             agent_stub_drive_ref=self.config.agent_stub_drive_ref,
             execution_context=execution_context,
             token_factory=self.agent_stub_token_factory,
             session_id=self.runtime_state.session_id,
         )
+        if agent_stub_env is None:
+            if not require_agent_stub_env:
+                return env
+            raise RuntimeError("Agent Stub environment injection is not available for this shell session.")
+        env.update(agent_stub_env)
+        return env
 
 
 async def execute_complete_with_commands(
@@ -663,6 +761,10 @@ def _shell_layer_prefix_prompt() -> str:
     return _SHELL_LAYER_PREFIX_PROMPT
 
 
+def _shell_layer_suffix_prompt() -> str:
+    return _SHELL_LAYER_SUFFIX_PROMPT
+
+
 def _metadata_dict(
     *,
     job_id: str,
@@ -689,6 +791,35 @@ def _tool_error(message: str, *, job_id: str | None = None) -> ShellToolErrorObs
     return result
 
 
+def _tool_error_from_exception(exc: Exception, *, job_id: str | None = None) -> ShellToolErrorObservation:
+    # Expected provider/runtime failures stay inside the tool contract and are
+    # returned to the model as observations. The broader Exception fallback
+    # below handles unexpected failures; BaseException, including cancellation,
+    # is intentionally left uncaught at the tool boundary.
+    if isinstance(exc, _HasErrorCode) and isinstance(exc.code, str) and exc.code:
+        return _tool_error(f"{exc.code}: {exc}", job_id=job_id)
+    return _tool_error(str(exc), job_id=job_id)
+
+
+def _tool_unexpected_error(
+    tool_name: str,
+    exc: Exception,
+    *,
+    session_id: str | None,
+    job_id: str | None = None,
+) -> ShellToolErrorObservation:
+    # Unexpected Exception still becomes a tool observation so one shell tool
+    # failure does not abort the agent loop, but it is logged with traceback for
+    # debugging. BaseException is intentionally not caught by callers.
+    logger.exception(
+        "Unexpected shell tool failure: tool=%s session_id=%s job_id=%s",
+        tool_name,
+        session_id,
+        job_id,
+    )
+    return _tool_error_from_exception(exc, job_id=job_id)
+
+
 def _generate_session_id() -> str:
     time_component = int(time.time()) & _SESSION_TIME_HEX_MASK
     random_component = secrets.token_hex(1)
@@ -701,6 +832,22 @@ def _workspace_cwd(session_id: str) -> str:
     return f"{_WORKSPACE_ROOT}/{_validated_session_id(session_id)}"
 
 
+def _shell_home_dir_for_agent_id(agent_id: str | None) -> str:
+    if agent_id is None:
+        raise ValueError("ShellLayer command execution requires execution_context.agent_id.")
+    return f"/home/{_validated_agent_home_segment(agent_id)}"
+
+
+def _validated_agent_home_segment(agent_id: str) -> str:
+    if agent_id in {".", ".."} or not _AGENT_HOME_SEGMENT_PATTERN.fullmatch(agent_id):
+        raise ValueError("execution_context.agent_id must be a safe single path segment for shell HOME.")
+    return agent_id
+
+
+def _workspace_cwd_for_home(home_dir: str, session_id: str) -> str:
+    return f"{home_dir}/{_WORKSPACE_DIR_NAME}/{_validated_session_id(session_id)}"
+
+
 def _workspace_bootstrap_script(config: DifyShellLayerConfig) -> str:
     install_commands = [command for tool in config.cli_tools for command in tool.install_commands]
     if not install_commands:
@@ -711,13 +858,11 @@ def _workspace_bootstrap_script(config: DifyShellLayerConfig) -> str:
 
 def _shell_config_export_lines(config: DifyShellLayerConfig) -> list[str]:
     lines: list[str] = []
-    for env_var in config.env:
-        lines.append(f"export {env_var.name}={_shquote(env_var.value)}")
+    # Plain env values travel through ShellCommandProtocol.run(env=...) so inline
+    # secrets are not rendered into generated shell scripts.
     for secret_ref in config.secret_refs:
         lines.append(f'export {secret_ref.name}="${{{secret_ref.name}:-}}"')
     for tool in config.cli_tools:
-        for env_var in tool.env:
-            lines.append(f"export {env_var.name}={_shquote(env_var.value)}")
         for secret_ref in tool.secret_refs:
             lines.append(f'export {secret_ref.name}="${{{secret_ref.name}:-}}"')
     if config.sandbox is not None:
@@ -729,6 +874,16 @@ def _shell_config_export_lines(config: DifyShellLayerConfig) -> list[str]:
     return lines
 
 
+def _shell_config_env(config: DifyShellLayerConfig) -> dict[str, str]:
+    env: dict[str, str] = {}
+    for env_var in config.env:
+        env[env_var.name] = env_var.value
+    for tool in config.cli_tools:
+        for env_var in tool.env:
+            env[env_var.name] = env_var.value
+    return env
+
+
 def _wrap_user_script(script: str, config: DifyShellLayerConfig) -> str:
     lines = _shell_config_export_lines(config)
     if not lines:
@@ -751,6 +906,10 @@ def _workspace_cleanup_script(*, session_id: str) -> str:
     return f'rm -rf -- "$HOME/workspace/{_validated_session_id(session_id)}"'
 
 
+def _workspace_ensure_script(*, session_id: str) -> str:
+    return f'mkdir -p "$HOME/workspace/{_validated_session_id(session_id)}"'
+
+
 def _shquote(value: str) -> str:
     return "'" + value.replace("'", "'\\''") + "'"
 
diff --git a/dify-agent/src/dify_agent/protocol/__init__.py b/dify-agent/src/dify_agent/protocol/__init__.py
index 46dc9c0ea10..a43b6cd4812 100644
--- a/dify-agent/src/dify_agent/protocol/__init__.py
+++ b/dify-agent/src/dify_agent/protocol/__init__.py
@@ -9,6 +9,7 @@ from .schemas import (
     DIFY_AGENT_MODEL_LAYER_ID,
     DIFY_AGENT_OUTPUT_LAYER_ID,
     RUN_EVENT_ADAPTER,
+    AgentRunUsage,
     BaseRunEvent,
     CancelRunRequest,
     CancelRunResponse,
@@ -55,6 +56,7 @@ from .sandbox import (
 
 __all__ = [
     "BaseRunEvent",
+    "AgentRunUsage",
     "CancelRunRequest",
     "CancelRunResponse",
     "CreateRunRequest",
diff --git a/dify-agent/src/dify_agent/protocol/schemas.py b/dify-agent/src/dify_agent/protocol/schemas.py
index 77501942666..1a02c7232a4 100644
--- a/dify-agent/src/dify_agent/protocol/schemas.py
+++ b/dify-agent/src/dify_agent/protocol/schemas.py
@@ -252,12 +252,29 @@ class DeferredToolCallPayload(BaseModel):
     model_config: ClassVar[ConfigDict] = ConfigDict(extra="forbid")
 
 
+class AgentRunUsage(BaseModel):
+    """Token usage reported by the model request behind one Agent run."""
+
+    prompt_tokens: int = 0
+    completion_tokens: int = 0
+    total_tokens: int = 0
+
+    model_config: ClassVar[ConfigDict] = ConfigDict(extra="forbid")
+
+    @model_validator(mode="after")
+    def _derive_total_tokens(self) -> AgentRunUsage:
+        if self.total_tokens == 0 and (self.prompt_tokens > 0 or self.completion_tokens > 0):
+            self.total_tokens = self.prompt_tokens + self.completion_tokens
+        return self
+
+
 class RunSucceededEventData(BaseModel):
     """Terminal success payload for final output or deferred tool continuation."""
 
     output: JsonValue | None = None
     deferred_tool_call: DeferredToolCallPayload | None = None
     session_snapshot: CompositorSessionSnapshot
+    usage: AgentRunUsage | None = None
 
     model_config: ClassVar[ConfigDict] = ConfigDict(extra="forbid")
 
@@ -276,6 +293,8 @@ class RunSucceededEventData(BaseModel):
             data["output"] = self.output
         if "deferred_tool_call" in self.model_fields_set:
             data["deferred_tool_call"] = self.deferred_tool_call
+        if self.usage is not None:
+            data["usage"] = self.usage
         return data
 
 
@@ -361,6 +380,7 @@ class RunEventsResponse(BaseModel):
 
 __all__ = [
     "BaseRunEvent",
+    "AgentRunUsage",
     "CancelRunRequest",
     "CancelRunResponse",
     "CreateRunRequest",
diff --git a/dify-agent/src/dify_agent/runtime/compositor_factory.py b/dify-agent/src/dify_agent/runtime/compositor_factory.py
index 9772663c020..3d1e2afc488 100644
--- a/dify-agent/src/dify_agent/runtime/compositor_factory.py
+++ b/dify-agent/src/dify_agent/runtime/compositor_factory.py
@@ -42,6 +42,7 @@ from dify_agent.layers.config.layer import DifyConfigLayer
 from dify_agent.layers.dify_core_tools.configs import DifyCoreToolsLayerConfig
 from dify_agent.layers.dify_core_tools.layer import DifyCoreToolsLayer
 from dify_agent.layers.dify_plugin.llm_layer import DifyPluginLLMLayer
+from dify_agent.layers.dify_plugin.configs import DifyPluginToolsLayerConfig
 from dify_agent.layers.dify_plugin.tools_layer import DifyPluginToolsLayer
 from dify_agent.layers.drive.layer import DifyDriveLayer
 from dify_agent.layers.execution_context.configs import DifyExecutionContextLayerConfig
@@ -128,7 +129,14 @@ def create_default_layer_providers(
             ),
         ),
         LayerProvider.from_layer_type(DifyPluginLLMLayer),
-        LayerProvider.from_layer_type(DifyPluginToolsLayer),
+        LayerProvider.from_factory(
+            layer_type=DifyPluginToolsLayer,
+            create=lambda config: DifyPluginToolsLayer.from_config_with_settings(
+                DifyPluginToolsLayerConfig.model_validate(config),
+                inner_api_url=inner_api_url,
+                inner_api_key=inner_api_key,
+            ),
+        ),
         LayerProvider.from_factory(
             layer_type=DifyCoreToolsLayer,
             create=lambda config: DifyCoreToolsLayer.from_config_with_settings(
diff --git a/dify-agent/src/dify_agent/runtime/event_sink.py b/dify-agent/src/dify_agent/runtime/event_sink.py
index 80cbf76cbd9..5dbe9fce62d 100644
--- a/dify-agent/src/dify_agent/runtime/event_sink.py
+++ b/dify-agent/src/dify_agent/runtime/event_sink.py
@@ -17,6 +17,7 @@ from pydantic_ai.messages import AgentStreamEvent
 
 from agenton.compositor import CompositorSessionSnapshot
 from dify_agent.protocol.schemas import (
+    AgentRunUsage,
     DeferredToolCallPayload,
     EmptyRunEventData,
     PydanticAIStreamRunEvent,
@@ -103,6 +104,7 @@ async def emit_run_succeeded(
     output: JsonValue | None | object = _UNSET,
     deferred_tool_call: DeferredToolCallPayload | object = _UNSET,
     session_snapshot: CompositorSessionSnapshot,
+    usage: AgentRunUsage | None = None,
 ) -> str:
     """Emit the terminal success event with output or deferred continuation.
 
@@ -112,13 +114,15 @@ async def emit_run_succeeded(
     Without that sentinel, ``output=None`` would be indistinguishable from
     “output field absent”, which would break nullable-success payloads.
     """
-    data: dict[str, JsonValue | DeferredToolCallPayload | CompositorSessionSnapshot | None] = {
+    data: dict[str, JsonValue | DeferredToolCallPayload | CompositorSessionSnapshot | AgentRunUsage | None] = {
         "session_snapshot": session_snapshot,
     }
     if output is not _UNSET:
         data["output"] = cast(JsonValue | None, output)
     if deferred_tool_call is not _UNSET:
         data["deferred_tool_call"] = cast(DeferredToolCallPayload, deferred_tool_call)
+    if usage is not None:
+        data["usage"] = usage
 
     return await emit_run_event(
         sink,
diff --git a/dify-agent/src/dify_agent/runtime/runner.py b/dify-agent/src/dify_agent/runtime/runner.py
index 65946736f02..97fbfe4ba4c 100644
--- a/dify-agent/src/dify_agent/runtime/runner.py
+++ b/dify-agent/src/dify_agent/runtime/runner.py
@@ -24,10 +24,10 @@ both the JSON-safe final output or deferred tool call and the session snapshot;
 there are no separate output or snapshot events to correlate.
 """
 
-from collections.abc import AsyncIterable
+from collections.abc import AsyncIterable, Callable
 from collections import Counter
 from dataclasses import dataclass
-from typing import Any, Literal, cast
+from typing import Any, Literal, Protocol, cast, runtime_checkable
 
 import httpx
 from pydantic import JsonValue, TypeAdapter
@@ -43,6 +43,7 @@ from dify_agent.layers.dify_plugin.llm_layer import DifyPluginLLMLayer
 from dify_agent.layers.dify_plugin.tools_layer import DifyPluginToolsLayer
 from dify_agent.layers.knowledge.layer import DifyKnowledgeBaseLayer
 from dify_agent.protocol.schemas import (
+    AgentRunUsage,
     CreateRunRequest,
     DIFY_AGENT_MODEL_LAYER_ID,
     DeferredToolCallPayload,
@@ -72,6 +73,26 @@ from dify_agent.runtime.user_prompt_validation import EMPTY_USER_PROMPTS_ERROR,
 _AGENT_OUTPUT_ADAPTER = TypeAdapter(object)
 
 
+@runtime_checkable
+class _HasUsage(Protocol):
+    usage: object
+
+
+@runtime_checkable
+class _HasInputTokens(Protocol):
+    input_tokens: object
+
+
+@runtime_checkable
+class _HasOutputTokens(Protocol):
+    output_tokens: object
+
+
+@runtime_checkable
+class _HasTotalTokens(Protocol):
+    total_tokens: object
+
+
 class AgentRunValidationError(ValueError):
     """Raised when a run request is valid JSON but cannot execute."""
 
@@ -84,6 +105,7 @@ class RunSuccessOutcome:
     output: JsonValue | None
     deferred_tool_call: DeferredToolCallPayload | None
     session_snapshot: CompositorSessionSnapshot
+    usage: AgentRunUsage | None
 
 
 class AgentRunRunner:
@@ -136,6 +158,7 @@ class AgentRunRunner:
                 else {"deferred_tool_call": outcome.deferred_tool_call}
             ),
             session_snapshot=outcome.session_snapshot,
+            usage=outcome.usage,
         )
         await self.sink.update_status(self.run_id, "succeeded")
 
@@ -171,6 +194,7 @@ class AgentRunRunner:
         output: JsonValue | None = None
         deferred_tool_call: DeferredToolCallPayload | None = None
         result_kind: Literal["output", "deferred_tool_call"] | None = None
+        usage: AgentRunUsage | None = None
         try:
             async with compositor.enter(configs=layer_configs, session_snapshot=self.request.session_snapshot) as run:
                 entered_run = True
@@ -218,6 +242,7 @@ class AgentRunRunner:
                     deferred_tool_results=deferred_tool_results,
                     event_stream_handler=handle_events,
                 )
+                usage = _serialize_agent_usage(_result_usage(result))
                 append_successful_run_history(history_layer, result.new_messages())
                 if isinstance(result.output, DeferredToolRequests):
                     if ask_human_layer is None:
@@ -252,6 +277,7 @@ class AgentRunRunner:
             output=output,
             deferred_tool_call=deferred_tool_call,
             session_snapshot=run.session_snapshot,
+            usage=usage,
         )
 
 
@@ -260,6 +286,34 @@ def _serialize_agent_output(output: object) -> JsonValue:
     return cast(JsonValue, _AGENT_OUTPUT_ADAPTER.dump_python(output, mode="json"))
 
 
+def _result_usage(result: object) -> object | None:
+    """Return pydantic-ai result usage across method/property API variants."""
+    if not isinstance(result, _HasUsage):
+        return None
+
+    usage = result.usage
+    if isinstance(usage, _HasInputTokens) or isinstance(usage, _HasOutputTokens):
+        return usage
+    if callable(usage):
+        usage_getter = cast(Callable[[], object], usage)
+        return usage_getter()
+    return usage
+
+
+def _serialize_agent_usage(usage: object | None) -> AgentRunUsage | None:
+    """Convert pydantic-ai request usage into the public Agent run usage shape."""
+    if usage is None:
+        return None
+    input_tokens = int(usage.input_tokens or 0) if isinstance(usage, _HasInputTokens) else 0
+    output_tokens = int(usage.output_tokens or 0) if isinstance(usage, _HasOutputTokens) else 0
+    total_tokens = int(usage.total_tokens or 0) if isinstance(usage, _HasTotalTokens) else 0
+    return AgentRunUsage(
+        prompt_tokens=input_tokens,
+        completion_tokens=output_tokens,
+        total_tokens=total_tokens,
+    )
+
+
 def _resolve_agent_output_type(output_type: OutputSpec[object], allow_deferred_tools: bool) -> OutputSpec[object]:
     """Return the run output type, optionally augmented with deferred-tool support."""
     if not allow_deferred_tools:
@@ -285,7 +339,12 @@ async def _resolve_run_tools(
     for slot in run.slots.values():
         layer = slot.layer
         if isinstance(layer, DifyPluginToolsLayer):
-            resolved_tools.extend(await layer.get_tools(http_client=plugin_daemon_http_client))
+            resolved_tools.extend(
+                await layer.get_tools(
+                    http_client=plugin_daemon_http_client,
+                    dify_api_http_client=dify_api_http_client,
+                )
+            )
         if isinstance(layer, DifyCoreToolsLayer):
             resolved_tools.extend(await layer.get_tools(http_client=dify_api_http_client))
         if isinstance(layer, DifyKnowledgeBaseLayer):
diff --git a/dify-agent/src/dify_agent/server/sandbox_files.py b/dify-agent/src/dify_agent/server/sandbox_files.py
index 78f5210b6aa..9d712f2db3e 100644
--- a/dify-agent/src/dify_agent/server/sandbox_files.py
+++ b/dify-agent/src/dify_agent/server/sandbox_files.py
@@ -9,7 +9,11 @@ minimal compositor from ``SandboxLocator``, enters the saved
 The scripts still frame their structured payloads with a PTY-safe
 base64-between-sentinels envelope. shellctl jobs are tmux-backed, so raw JSON can
 be wrapped or surrounded by prompt noise; the framing keeps list/read/upload
-responses parseable without falling back to direct shellctl file access.
+responses parseable without falling back to direct shellctl file access. Path
+arguments resolve from the saved shell workspace cwd, and ``~`` resolves through
+the shell layer's injected sandbox ``HOME``. The scripts do not re-impose a
+workspace-root boundary, so callers can use ``../`` or ``~/...`` when the sandbox
+filesystem layout expects it.
 """
 
 from __future__ import annotations
@@ -63,21 +67,9 @@ def emit(payload):
     print(BEGIN + blob + END)
 
 
-def resolve_target(root: Path, raw_path: str) -> tuple[Path | None, str]:
-    target = (root / raw_path).resolve()
-    if target != root and root not in target.parents:
-        emit({"error": "invalid_sandbox_path", "message": "path escapes the sandbox workspace"})
-        return None, raw_path
-    return target, raw_path
-
-
-root = Path.cwd().resolve()
 raw_path = sys.argv[1]
 limit = int(sys.argv[2])
-target_info = resolve_target(root, raw_path)
-if target_info[0] is None:
-    sys.exit(0)
-target, normalized_path = target_info
+target = Path(raw_path).expanduser().resolve()
 
 if not target.exists():
     emit({"error": "sandbox_path_not_found", "message": "path not found in sandbox"})
@@ -109,7 +101,7 @@ for child in sorted(target.iterdir(), key=lambda item: item.name)[:limit]:
 
 emit(
     {
-        "path": normalized_path,
+        "path": raw_path,
         "entries": entries,
         "truncated": len(list(target.iterdir())) > limit,
     }
@@ -132,13 +124,9 @@ def emit(payload):
     print(BEGIN + blob + END)
 
 
-root = Path.cwd().resolve()
 raw_path = sys.argv[1]
 max_bytes = int(sys.argv[2])
-target = (root / raw_path).resolve()
-if target != root and root not in target.parents:
-    emit({"error": "invalid_sandbox_path", "message": "path escapes the sandbox workspace"})
-    sys.exit(0)
+target = Path(raw_path).expanduser().resolve()
 if not target.exists():
     emit({"error": "sandbox_path_not_found", "message": "path not found in sandbox"})
     sys.exit(0)
@@ -194,12 +182,8 @@ def emit(payload):
     print(BEGIN + blob + END)
 
 
-root = Path.cwd().resolve()
 raw_path = sys.argv[1]
-target = (root / raw_path).resolve()
-if target != root and root not in target.parents:
-    emit({"error": "invalid_sandbox_path", "message": "path escapes the sandbox workspace"})
-    sys.exit(0)
+target = Path(raw_path).expanduser().resolve()
 if not target.exists():
     emit({"error": "sandbox_path_not_found", "message": "path not found in sandbox"})
     sys.exit(0)
@@ -310,15 +294,25 @@ class SandboxFileService:
 
 
 def _normalize_sandbox_path(path: str, *, allow_current_directory: bool) -> str:
+    """Reject only syntactically unsafe paths and preserve relative traversal.
+
+    The remote scripts run with the saved workspace cwd, so ``../`` remains a
+    valid sandbox-relative path when callers need to reach sibling directories.
+    ``~`` and ``~/...`` are also accepted and resolved by the embedded scripts
+    against the shell layer's sandbox ``HOME``.
+    """
+
     normalized = (path or "").strip()
     if normalized in {"", ".", "./"}:
         if allow_current_directory:
             return "."
         raise SandboxFileError("invalid_sandbox_path", "path must not be blank", status_code=400)
-    if normalized.startswith("/") or normalized.startswith("~"):
+    if normalized.startswith("/"):
         raise SandboxFileError(
             "invalid_sandbox_path", "path must be relative to the sandbox workspace", status_code=400
         )
+    if normalized.startswith("~") and normalized != "~" and not normalized.startswith("~/"):
+        raise SandboxFileError("invalid_sandbox_path", "path must use ~ or ~/ for the sandbox home", status_code=400)
     if "\x00" in normalized or any(ord(ch) < 0x20 for ch in normalized):
         raise SandboxFileError("invalid_sandbox_path", "path contains unsupported control characters", status_code=400)
     return normalized
diff --git a/dify-agent/tests/local/dify_agent/adapters/llm/test_model.py b/dify-agent/tests/local/dify_agent/adapters/llm/test_model.py
index c68142ce577..da541ca8cb0 100644
--- a/dify-agent/tests/local/dify_agent/adapters/llm/test_model.py
+++ b/dify-agent/tests/local/dify_agent/adapters/llm/test_model.py
@@ -5,6 +5,7 @@ from typing import cast
 from unittest.mock import patch
 
 import httpx
+from graphon.model_runtime.entities.message_entities import TextPromptMessageContent
 from pydantic_ai.exceptions import ModelHTTPError, UserError
 from pydantic_ai.messages import (
     InstructionPart,
@@ -311,6 +312,62 @@ class DifyLLMAdapterModelTests(unittest.IsolatedAsyncioTestCase):
         self.assertEqual(response.usage.input_tokens, 11)
         self.assertEqual(response.usage.output_tokens, 7)
 
+    async def test_request_stream_splits_embedded_thinking_tags_from_text_content_parts(self) -> None:
+        def handler(_request: httpx.Request) -> httpx.Response:
+            return build_stream_response(
+                LLMResultChunk(
+                    model="demo-model",
+                    delta=LLMResultChunkDelta(
+                        index=0,
+                        message=AssistantPromptMessage(
+                            content=[TextPromptMessageContent(data="before<think>reasoning")],
+                            tool_calls=[],
+                        ),
+                    ),
+                ),
+                LLMResultChunk(
+                    model="demo-model",
+                    delta=LLMResultChunkDelta(
+                        index=1,
+                        message=AssistantPromptMessage(
+                            content=[TextPromptMessageContent(data=" continues</think>after")],
+                            tool_calls=[],
+                        ),
+                    ),
+                ),
+                LLMResultChunk(
+                    model="demo-model",
+                    delta=LLMResultChunkDelta(
+                        index=2,
+                        message=AssistantPromptMessage(content="", tool_calls=[]),
+                        usage=make_usage(prompt_tokens=6, completion_tokens=4),
+                        finish_reason="stop",
+                    ),
+                ),
+            )
+
+        async with self.mock_daemon_stream(httpx.MockTransport(handler)):
+            adapter = DifyLLMAdapterModel(
+                "demo-model",
+                self.make_provider(),
+                model_provider="openai",
+                credentials={"api_key": "secret"},
+            )
+
+            async with adapter.request_stream(
+                [ModelRequest(parts=[UserPromptPart("hello")])],
+                model_settings=None,
+                model_request_parameters=ModelRequestParameters(),
+            ) as stream:
+                events = [event async for event in stream]
+                response = stream.get()
+
+        self.assertTrue(events)
+        self.assertEqual([part.part_kind for part in response.parts], ["text", "thinking", "text"])
+        self.assertEqual(cast(TextPart, response.parts[0]).content, "before")
+        self.assertEqual(cast(ThinkingPart, response.parts[1]).content, "reasoning continues")
+        self.assertEqual(cast(TextPart, response.parts[2]).content, "after")
+
     async def test_request_stream_yields_response_parts_and_usage(self) -> None:
         def handler(_request: httpx.Request) -> httpx.Response:
             return build_stream_response(
diff --git a/dify-agent/tests/local/dify_agent/adapters/shell/test_shellctl.py b/dify-agent/tests/local/dify_agent/adapters/shell/test_shellctl.py
index dca0a9e7376..4fcb407555b 100644
--- a/dify-agent/tests/local/dify_agent/adapters/shell/test_shellctl.py
+++ b/dify-agent/tests/local/dify_agent/adapters/shell/test_shellctl.py
@@ -7,13 +7,14 @@ import base64
 from collections.abc import Callable
 from dataclasses import dataclass, field
 
+import httpx
 import pytest
 
 from dify_agent.adapters.shell import shellctl
 from dify_agent.adapters.shell.config import ShellAdapterSettings
 from dify_agent.adapters.shell.factory import create_shell_provider
-from dify_agent.adapters.shell.protocols import ShellCommandResult
-from dify_agent.adapters.shell.shellctl import ShellFileTransferError, ShellProviderError, ShellctlProvider
+from dify_agent.adapters.shell.protocols import ShellCommandResult, ShellProviderError
+from dify_agent.adapters.shell.shellctl import ShellFileTransferError, ShellctlProvider
 
 
 @dataclass(slots=True)
@@ -228,6 +229,82 @@ def test_commands_forward_parameters_and_map_metadata() -> None:
     assert client.delete_calls == [("run-job", True, 2.0)]
 
 
+def test_commands_map_http_timeout_to_shell_provider_error() -> None:
+    request = httpx.Request("POST", "http://shellctl.example/v1/jobs")
+    client = FakeShellctlClient(
+        run_handler=lambda script, cwd, env, timeout: (_ for _ in ()).throw(
+            httpx.ReadTimeout("timed out", request=request)
+        )
+    )
+
+    async def scenario() -> None:
+        resource = await _provider(client).create()
+        with pytest.raises(ShellProviderError, match="timed out") as exc_info:
+            await resource.commands.run("pwd", timeout=2.5)
+        assert exc_info.value.code == "timeout"
+
+    asyncio.run(scenario())
+
+
+def test_commands_map_http_request_error_to_shell_provider_error() -> None:
+    request = httpx.Request("POST", "http://shellctl.example/v1/jobs/run")
+    client = FakeShellctlClient(
+        wait_handler=lambda job_id, offset, timeout: (_ for _ in ()).throw(
+            httpx.ConnectError("connection failed", request=request)
+        )
+    )
+
+    async def scenario() -> None:
+        resource = await _provider(client).create()
+        with pytest.raises(ShellProviderError, match="connection failed") as exc_info:
+            await resource.commands.wait("run-job", offset=3, timeout=4.0)
+        assert exc_info.value.code == "request_error"
+
+    asyncio.run(scenario())
+
+
+def test_delete_maps_http_timeout_to_shell_provider_error() -> None:
+    request = httpx.Request("DELETE", "http://shellctl.example/v1/jobs/run-job")
+
+    @dataclass(slots=True)
+    class DeleteTimeoutClient(FakeShellctlClient):
+        async def delete(self, job_id, *, force=False, grace_seconds=None):
+            self.delete_calls.append((job_id, force, grace_seconds))
+            raise httpx.ReadTimeout("delete timed out", request=request)
+
+    client = DeleteTimeoutClient()
+
+    async def scenario() -> None:
+        resource = await _provider(client).create()
+        with pytest.raises(ShellProviderError, match="delete timed out") as exc_info:
+            await resource.commands.delete("run-job", force=True, grace_seconds=2.0)
+        assert exc_info.value.code == "timeout"
+
+    asyncio.run(scenario())
+    assert client.delete_calls == [("run-job", True, 2.0)]
+
+
+def test_delete_maps_http_request_error_to_shell_provider_error() -> None:
+    request = httpx.Request("DELETE", "http://shellctl.example/v1/jobs/run-job")
+
+    @dataclass(slots=True)
+    class DeleteRequestErrorClient(FakeShellctlClient):
+        async def delete(self, job_id, *, force=False, grace_seconds=None):
+            self.delete_calls.append((job_id, force, grace_seconds))
+            raise httpx.ConnectError("delete connection failed", request=request)
+
+    client = DeleteRequestErrorClient()
+
+    async def scenario() -> None:
+        resource = await _provider(client).create()
+        with pytest.raises(ShellProviderError, match="delete connection failed") as exc_info:
+            await resource.commands.delete("run-job", force=True, grace_seconds=2.0)
+        assert exc_info.value.code == "request_error"
+
+    asyncio.run(scenario())
+    assert client.delete_calls == [("run-job", True, 2.0)]
+
+
 def test_files_upload_and_download_still_work() -> None:
     content = b"hello \x00 world"
     encoded = base64.b64encode(content).decode("ascii")
diff --git a/dify-agent/tests/local/dify_agent/agent_stub/cli/test_config.py b/dify-agent/tests/local/dify_agent/agent_stub/cli/test_config.py
index b9fbe015969..0032d9c6142 100644
--- a/dify-agent/tests/local/dify_agent/agent_stub/cli/test_config.py
+++ b/dify-agent/tests/local/dify_agent/agent_stub/cli/test_config.py
@@ -77,34 +77,16 @@ def test_pull_config_files_from_environment_downloads_visible_files(
     assert (tmp_path / "guide.txt").read_bytes() == b"guide-bytes"
 
 
-def test_pull_config_env_from_environment_writes_only_declared_keys(
-    monkeypatch: pytest.MonkeyPatch,
-    tmp_path: Path,
-) -> None:
-    monkeypatch.setattr(config_cli, "manifest_from_environment", _manifest_payload)
-    monkeypatch.setenv("API_KEY", "plain")
-    monkeypatch.setenv("JSON_VALUE", "two words")
-
-    result = config_cli.pull_config_env_from_environment(local_path=str(tmp_path / ".env"))
-
-    assert result.read_text(encoding="utf-8") == 'API_KEY=plain\nJSON_VALUE="two words"\n'
-
-
-def test_pull_config_env_and_note_use_hidden_default_dir(
+def test_pull_config_note_uses_hidden_default_dir(
     monkeypatch: pytest.MonkeyPatch,
     tmp_path: Path,
 ) -> None:
     monkeypatch.chdir(tmp_path)
     monkeypatch.setattr(config_cli, "manifest_from_environment", _manifest_payload)
-    monkeypatch.setenv("API_KEY", "plain")
-    monkeypatch.setenv("JSON_VALUE", "two words")
 
-    env_path = config_cli.pull_config_env_from_environment()
     note_path = config_cli.pull_config_note_from_environment()
 
-    assert env_path == (tmp_path / ".dify_conf" / ".env").resolve()
     assert note_path == (tmp_path / ".dify_conf" / "note.md").resolve()
-    assert env_path.read_text(encoding="utf-8") == 'API_KEY=plain\nJSON_VALUE="two words"\n'
     assert note_path.read_text(encoding="utf-8") == "Use carefully."
 
 
@@ -177,31 +159,6 @@ def test_push_config_note_from_environment_reads_stdin(monkeypatch: pytest.Monke
     assert request.note == "from-stdin"
 
 
-def test_push_config_env_from_environment_reads_default_file(
-    monkeypatch: pytest.MonkeyPatch,
-    tmp_path: Path,
-) -> None:
-    monkeypatch.chdir(tmp_path)
-    env_path = tmp_path / ".dify_conf" / ".env"
-    env_path.parent.mkdir()
-    env_path.write_text("API_KEY=value\n", encoding="utf-8")
-    monkeypatch.setattr(config_cli, "read_agent_stub_environment", lambda: _environment())
-
-    captured: dict[str, object] = {}
-
-    def fake_push_sync(**kwargs):
-        captured.update(kwargs)
-        return _manifest_payload()
-
-    monkeypatch.setattr(config_cli, "request_agent_stub_config_push_sync", fake_push_sync)
-
-    config_cli.push_config_env_from_environment(None)
-
-    request = cast(AgentStubConfigPushRequest, captured["request"])
-    assert request.env_text == "API_KEY=value\n"
-    assert request.note is None
-
-
 def test_push_config_env_from_environment_reads_explicit_file(
     monkeypatch: pytest.MonkeyPatch,
     tmp_path: Path,
@@ -264,7 +221,7 @@ def test_push_config_files_from_environment_builds_upload_items(
 
     monkeypatch.setattr(config_cli, "request_agent_stub_config_push_sync", fake_push_sync)
 
-    config_cli.push_config_files_from_environment([str(file_path)], None)
+    config_cli.push_config_files_from_environment([str(file_path)])
 
     request = cast(AgentStubConfigPushRequest, captured["request"])
     assert request.files[0].name == "guide.txt"
@@ -273,19 +230,9 @@ def test_push_config_files_from_environment_builds_upload_items(
     assert request.skills == []
 
 
-def test_push_config_files_from_environment_validates_name_usage(tmp_path: Path) -> None:
-    first = tmp_path / "a.txt"
-    second = tmp_path / "b.txt"
-    first.write_text("a", encoding="utf-8")
-    second.write_text("b", encoding="utf-8")
-
-    with pytest.raises(AgentStubValidationError, match="--name requires exactly one PATH"):
-        config_cli.push_config_files_from_environment([str(first), str(second)], "renamed.txt")
-
-
 def test_push_config_files_from_environment_rejects_empty_paths() -> None:
     with pytest.raises(AgentStubValidationError, match="at least one file path is required"):
-        config_cli.push_config_files_from_environment([], None)
+        config_cli.push_config_files_from_environment([])
 
 
 def test_delete_config_files_from_environment_builds_delete_items(monkeypatch: pytest.MonkeyPatch) -> None:
diff --git a/dify-agent/tests/local/dify_agent/agent_stub/cli/test_main.py b/dify-agent/tests/local/dify_agent/agent_stub/cli/test_main.py
index 0aedc6f435f..b87e79daa35 100644
--- a/dify-agent/tests/local/dify_agent/agent_stub/cli/test_main.py
+++ b/dify-agent/tests/local/dify_agent/agent_stub/cli/test_main.py
@@ -299,12 +299,11 @@ def test_cli_config_manifest_omits_hash_fields(
             "push_config_note_from_environment",
             {"local_path": "/tmp/note.md"},
         ),
-        (["config", "env", "push"], "push_config_env_from_environment", {"local_path": None}),
         (["config", "env", "push", "/tmp/.env"], "push_config_env_from_environment", {"local_path": "/tmp/.env"}),
         (
-            ["config", "files", "push", "/tmp/guide.txt", "--name", "runtime.txt"],
+            ["config", "files", "push", "/tmp/guide.txt"],
             "push_config_files_from_environment",
-            {"paths": ["/tmp/guide.txt"], "name": "runtime.txt"},
+            {"paths": ["/tmp/guide.txt"]},
         ),
         (
             ["config", "files", "delete", "old.txt", "legacy.txt"],
diff --git a/dify-agent/tests/local/dify_agent/agent_stub/server/test_agent_stub_files.py b/dify-agent/tests/local/dify_agent/agent_stub/server/test_agent_stub_files.py
index 3c713aedc34..164fd7487d7 100644
--- a/dify-agent/tests/local/dify_agent/agent_stub/server/test_agent_stub_files.py
+++ b/dify-agent/tests/local/dify_agent/agent_stub/server/test_agent_stub_files.py
@@ -23,6 +23,7 @@ def _principal() -> AgentStubPrincipal:
             user_id="user-1",
             user_from="account",
             workflow_id="workflow-1",
+            conversation_id="conversation-1",
             agent_mode="workflow_run",
             invoke_from="service-api",
         ),
@@ -54,6 +55,7 @@ def test_dify_api_agent_stub_file_handler_injects_execution_context_for_upload(m
             "user_id": "user-1",
             "filename": "report.pdf",
             "mimetype": "application/pdf",
+            "conversation_id": "conversation-1",
         }
         return httpx.Response(200, json={"data": {"url": "https://files.example.com/upload"}})
 
diff --git a/dify-agent/tests/local/dify_agent/client/test_client.py b/dify-agent/tests/local/dify_agent/client/test_client.py
index d1b06ad5fc9..3a1525f015c 100644
--- a/dify-agent/tests/local/dify_agent/client/test_client.py
+++ b/dify-agent/tests/local/dify_agent/client/test_client.py
@@ -137,9 +137,9 @@ class DisconnectingSyncStream(httpx.SyncByteStream):
 
 
 def test_sse_decoder_accepts_function_tool_result_part_alias(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setattr(client_module, "_FUNCTION_TOOL_RESULT_PAYLOAD_KEY", "result")
+    monkeypatch.setattr(client_module, "_function_tool_result_payload_key_cache", "part")
     decoder = client_module._SSEDecoder()
-    payload = _function_tool_result_payload("part")
+    payload = _function_tool_result_payload("result")
 
     assert decoder.feed_line(f"data: {json.dumps(payload)}") is None
     event = decoder.feed_line("")
@@ -153,7 +153,7 @@ def test_sse_decoder_accepts_function_tool_result_part_alias(monkeypatch: pytest
 def test_function_tool_result_payload_normalization_supports_old_part_schema(
     monkeypatch: pytest.MonkeyPatch,
 ) -> None:
-    monkeypatch.setattr(client_module, "_FUNCTION_TOOL_RESULT_PAYLOAD_KEY", "part")
+    monkeypatch.setattr(client_module, "_function_tool_result_payload_key_cache", "part")
     payload = _function_tool_result_payload("result")
 
     normalized = client_module._normalize_run_event_payload_for_local_pydantic_ai(payload)
diff --git a/dify-agent/tests/local/dify_agent/layers/conftest.py b/dify-agent/tests/local/dify_agent/layers/conftest.py
deleted file mode 100644
index 9d393301962..00000000000
--- a/dify-agent/tests/local/dify_agent/layers/conftest.py
+++ /dev/null
@@ -1,32 +0,0 @@
-from __future__ import annotations
-
-import sys
-import types
-
-from agenton.layers import EmptyLayerConfig, EmptyRuntimeState, NoLayerDeps, PlainLayer
-from pydantic import BaseModel
-
-
-class _BaseSettings(BaseModel):
-    """Minimal test stub for environments without pydantic-settings installed."""
-
-
-class _SettingsConfigDict(dict[str, object]):
-    """Minimal callable mapping used by ShellAdapterSettings.model_config."""
-
-
-if "pydantic_settings" not in sys.modules:
-    stub = types.ModuleType("pydantic_settings")
-    setattr(stub, "BaseSettings", _BaseSettings)
-    setattr(stub, "SettingsConfigDict", _SettingsConfigDict)
-    sys.modules["pydantic_settings"] = stub
-
-
-if "dify_agent.layers.execution_context.layer" not in sys.modules:
-    execution_context_stub = types.ModuleType("dify_agent.layers.execution_context.layer")
-
-    class DifyExecutionContextLayer(PlainLayer[NoLayerDeps, EmptyLayerConfig, EmptyRuntimeState]):
-        """Minimal test stub for shell-layer type-only imports."""
-
-    setattr(execution_context_stub, "DifyExecutionContextLayer", DifyExecutionContextLayer)
-    sys.modules["dify_agent.layers.execution_context.layer"] = execution_context_stub
diff --git a/dify-agent/tests/local/dify_agent/layers/dify_plugin/test_layers.py b/dify-agent/tests/local/dify_agent/layers/dify_plugin/test_layers.py
index a1cb137c10c..995fea0120c 100644
--- a/dify-agent/tests/local/dify_agent/layers/dify_plugin/test_layers.py
+++ b/dify-agent/tests/local/dify_agent/layers/dify_plugin/test_layers.py
@@ -1,5 +1,6 @@
 import asyncio
 import json
+from types import SimpleNamespace
 
 import httpx
 import pytest
@@ -19,7 +20,7 @@ from dify_agent.layers.dify_plugin.configs import (
     DifyPluginToolsLayerConfig,
 )
 from dify_agent.layers.dify_plugin.llm_layer import DifyPluginLLMLayer
-from dify_agent.layers.dify_plugin.tools_layer import DifyPluginToolsLayer
+from dify_agent.layers.dify_plugin.tools_layer import DifyPluginToolsLayer, _PluginToolFileContext
 from dify_agent.layers.execution_context import DifyExecutionContextLayerConfig
 from dify_agent.layers.execution_context.layer import DifyExecutionContextLayer
 
@@ -91,6 +92,17 @@ def _execution_context_provider() -> LayerProvider[DifyExecutionContextLayer]:
     )
 
 
+def _tools_provider() -> LayerProvider[DifyPluginToolsLayer]:
+    return LayerProvider.from_factory(
+        layer_type=DifyPluginToolsLayer,
+        create=lambda config: DifyPluginToolsLayer.from_config_with_settings(
+            DifyPluginToolsLayerConfig.model_validate(config),
+            inner_api_url="http://dify-api",
+            inner_api_key="inner-secret",
+        ),
+    )
+
+
 def _prepared_tool_parameters() -> list[DifyPluginToolParameter]:
     return [
         DifyPluginToolParameter(
@@ -151,6 +163,42 @@ def _llm_only_parameter(*, name: str, description: str, default: JsonValue = Non
     )
 
 
+def _file_parameter(
+    *,
+    name: str = "source",
+    parameter_type: DifyPluginToolParameterType = DifyPluginToolParameterType.FILE,
+) -> DifyPluginToolParameter:
+    return DifyPluginToolParameter(
+        name=name,
+        type=parameter_type,
+        form=DifyPluginToolParameterForm.LLM,
+        required=True,
+        llm_description="Source file",
+    )
+
+
+def _file_tools_config(
+    *,
+    parameter_type: DifyPluginToolParameterType = DifyPluginToolParameterType.FILE,
+) -> DifyPluginToolsLayerConfig:
+    return DifyPluginToolsLayerConfig(
+        tools=[
+            DifyPluginToolConfig(
+                plugin_id="langgenius/tools",
+                provider="search",
+                tool_name="read_file",
+                credential_type="api-key",
+                parameters=[_file_parameter(parameter_type=parameter_type)],
+                parameters_json_schema={
+                    "type": "object",
+                    "properties": {"source": {"type": "string"}},
+                    "required": ["source"],
+                },
+            )
+        ]
+    )
+
+
 def _invoke_stream_response(
     *,
     error_payload: dict[str, object] | None = None,
@@ -204,6 +252,28 @@ def _tool_transport(
     return httpx.MockTransport(handler)
 
 
+def _file_tool_transport(
+    *,
+    expected_source: object,
+    download_response: dict[str, object] | None = None,
+) -> httpx.MockTransport:
+    def handler(request: httpx.Request) -> httpx.Response:
+        if request.url.path.endswith("/inner/api/download/file/request"):
+            assert download_response is not None
+            payload = json.loads(request.content.decode("utf-8"))
+            assert payload["file"] == {"transfer_method": "tool_file", "reference": "dify-file-ref:file-1"}
+            return httpx.Response(200, json={"data": download_response})
+
+        if request.url.path.endswith("/dispatch/tool/invoke"):
+            payload = json.loads(request.content.decode("utf-8"))
+            assert payload["data"]["tool_parameters"]["source"] == expected_source
+            return _invoke_stream_response()
+
+        raise AssertionError(f"Unexpected request path: {request.url.path}")
+
+    return httpx.MockTransport(handler)
+
+
 def test_dify_plugin_type_id_constants_match_implementation_classes() -> None:
     assert DIFY_PLUGIN_LLM_LAYER_TYPE_ID == DifyPluginLLMLayer.type_id
     assert DIFY_PLUGIN_TOOLS_LAYER_TYPE_ID == DifyPluginToolsLayer.type_id
@@ -245,7 +315,7 @@ def test_dify_plugin_tools_layer_uses_prepared_tool_definition_and_invokes_daemo
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=_tool_transport()) as client:
@@ -253,7 +323,7 @@ def test_dify_plugin_tools_layer_uses_prepared_tool_definition_and_invokes_daemo
                 configs={"execution_context": _execution_context_config(), "tools": _tools_config()}
             ) as run:
                 tools_layer = run.get_layer("tools", DifyPluginToolsLayer)
-                tool = (await tools_layer.get_tools(http_client=client))[0]
+                tool = (await tools_layer.get_tools(http_client=client, dify_api_http_client=client))[0]
 
                 tool_def = await tool.prepare_tool_def(None)  # pyright: ignore[reportArgumentType]
                 result = await tool.function_schema.call(
@@ -322,14 +392,16 @@ def test_dify_plugin_tools_layer_uses_each_tool_plugin_id_for_transport() -> Non
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": tools_config}
             ) as run:
-                tools = await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client)
+                tools = await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                    http_client=client, dify_api_http_client=client
+                )
 
                 await tools[0].function_schema.call({"query": "first"}, None)  # pyright: ignore[reportArgumentType]
                 await tools[1].function_schema.call({"query": "second"}, None)  # pyright: ignore[reportArgumentType]
@@ -432,14 +504,18 @@ def test_dify_plugin_tools_layer_casts_prepared_parameter_values_before_invocati
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": tools_config}
             ) as run:
-                tool = (await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client))[0]
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
+                )[0]
 
                 result = await tool.function_schema.call(
                     {
@@ -496,14 +572,18 @@ def test_dify_plugin_tools_layer_sends_prepared_parameter_defaults_to_daemon() -
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=httpx.MockTransport(handler)) as client:
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": tools_config}
             ) as run:
-                tool = (await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client))[0]
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
+                )[0]
 
                 result = await tool.function_schema.call(
                     {"query": "dify"},
@@ -515,12 +595,289 @@ def test_dify_plugin_tools_layer_sends_prepared_parameter_defaults_to_daemon() -
     asyncio.run(scenario())
 
 
+def test_dify_plugin_tools_layer_converts_remote_url_file_string() -> None:
+    async def scenario() -> None:
+        expected_source = {
+            "dify_model_identity": "__dify__file__",
+            "type": "custom",
+            "url": "https://example.com/report.pdf",
+            "filename": "report.pdf",
+            "mime_type": "application/pdf",
+            "extension": ".pdf",
+            "size": -1,
+        }
+        compositor = Compositor(
+            [
+                LayerNode("execution_context", _execution_context_provider()),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
+            ]
+        )
+        async with httpx.AsyncClient(transport=_file_tool_transport(expected_source=expected_source)) as client:
+            async with compositor.enter(
+                configs={"execution_context": _execution_context_config(), "tools": _file_tools_config()}
+            ) as run:
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client,
+                        dify_api_http_client=client,
+                    )
+                )[0]
+                result = await tool.function_schema.call(
+                    {"source": "https://example.com/report.pdf"},
+                    None,  # pyright: ignore[reportArgumentType]
+                )
+
+        assert result == 'found {"count": 1}'
+
+    asyncio.run(scenario())
+
+
+def test_dify_plugin_tools_layer_converts_remote_url_file_mapping() -> None:
+    async def scenario() -> None:
+        expected_source = {
+            "dify_model_identity": "__dify__file__",
+            "type": "custom",
+            "url": "https://example.com/report.pdf",
+            "filename": "report.pdf",
+            "mime_type": "application/pdf",
+            "extension": ".pdf",
+            "size": -1,
+        }
+        compositor = Compositor(
+            [
+                LayerNode("execution_context", _execution_context_provider()),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
+            ]
+        )
+        async with httpx.AsyncClient(transport=_file_tool_transport(expected_source=expected_source)) as client:
+            async with compositor.enter(
+                configs={"execution_context": _execution_context_config(), "tools": _file_tools_config()}
+            ) as run:
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client,
+                        dify_api_http_client=client,
+                    )
+                )[0]
+                result = await tool.function_schema.call(
+                    {"source": {"transfer_method": "remote_url", "url": "https://example.com/report.pdf"}},
+                    None,  # pyright: ignore[reportArgumentType]
+                )
+
+        assert result == 'found {"count": 1}'
+
+    asyncio.run(scenario())
+
+
+def test_dify_plugin_tools_layer_resolves_tool_file_mapping_before_invocation() -> None:
+    async def scenario() -> None:
+        expected_source = {
+            "dify_model_identity": "__dify__file__",
+            "type": "custom",
+            "url": "https://signed.example/report.pdf",
+            "filename": "report.pdf",
+            "mime_type": "application/pdf",
+            "extension": ".pdf",
+            "size": 42,
+        }
+        compositor = Compositor(
+            [
+                LayerNode("execution_context", _execution_context_provider()),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
+            ]
+        )
+        async with httpx.AsyncClient(
+            transport=_file_tool_transport(
+                expected_source=expected_source,
+                download_response={
+                    "filename": "report.pdf",
+                    "mime_type": "application/pdf",
+                    "size": 42,
+                    "download_url": "https://signed.example/report.pdf",
+                },
+            )
+        ) as client:
+            async with compositor.enter(
+                configs={"execution_context": _execution_context_config(), "tools": _file_tools_config()}
+            ) as run:
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client,
+                        dify_api_http_client=client,
+                    )
+                )[0]
+                result = await tool.function_schema.call(
+                    {"source": {"transfer_method": "tool_file", "reference": "dify-file-ref:file-1"}},
+                    None,  # pyright: ignore[reportArgumentType]
+                )
+
+        assert result == 'found {"count": 1}'
+
+    asyncio.run(scenario())
+
+
+def test_dify_plugin_tools_layer_converts_files_parameter_values() -> None:
+    async def scenario() -> None:
+        expected_source = [
+            {
+                "dify_model_identity": "__dify__file__",
+                "type": "custom",
+                "url": "https://example.com/a.txt",
+                "filename": "a.txt",
+                "mime_type": "text/plain",
+                "extension": ".txt",
+                "size": -1,
+            },
+            {
+                "dify_model_identity": "__dify__file__",
+                "type": "custom",
+                "url": "https://example.com/b.txt",
+                "filename": "b.txt",
+                "mime_type": "text/plain",
+                "extension": ".txt",
+                "size": -1,
+            },
+        ]
+        compositor = Compositor(
+            [
+                LayerNode("execution_context", _execution_context_provider()),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
+            ]
+        )
+        async with httpx.AsyncClient(transport=_file_tool_transport(expected_source=expected_source)) as client:
+            async with compositor.enter(
+                configs={
+                    "execution_context": _execution_context_config(),
+                    "tools": _file_tools_config(parameter_type=DifyPluginToolParameterType.FILES),
+                }
+            ) as run:
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client,
+                        dify_api_http_client=client,
+                    )
+                )[0]
+                result = await tool.function_schema.call(
+                    {"source": ["https://example.com/a.txt", "https://example.com/b.txt"]},
+                    None,  # pyright: ignore[reportArgumentType]
+                )
+
+        assert result == 'found {"count": 1}'
+
+    asyncio.run(scenario())
+
+
+def test_dify_plugin_tools_layer_rejects_multiple_values_for_single_file() -> None:
+    async def scenario() -> None:
+        compositor = Compositor(
+            [
+                LayerNode("execution_context", _execution_context_provider()),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
+            ]
+        )
+        async with httpx.AsyncClient(transport=httpx.MockTransport(lambda _request: httpx.Response(500))) as client:
+            async with compositor.enter(
+                configs={"execution_context": _execution_context_config(), "tools": _file_tools_config()}
+            ) as run:
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client,
+                        dify_api_http_client=client,
+                    )
+                )[0]
+                result = await tool.function_schema.call(
+                    {"source": ["https://example.com/a.txt", "https://example.com/b.txt"]},
+                    None,  # pyright: ignore[reportArgumentType]
+                )
+        assert "only accepts one file" in result
+
+    asyncio.run(scenario())
+
+
+def test_dify_plugin_tools_layer_rejects_path_without_shell() -> None:
+    async def scenario() -> None:
+        compositor = Compositor(
+            [
+                LayerNode("execution_context", _execution_context_provider()),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
+            ]
+        )
+        async with httpx.AsyncClient(transport=httpx.MockTransport(lambda _request: httpx.Response(500))) as client:
+            async with compositor.enter(
+                configs={"execution_context": _execution_context_config(), "tools": _file_tools_config()}
+            ) as run:
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client,
+                        dify_api_http_client=client,
+                    )
+                )[0]
+                result = await tool.function_schema.call(
+                    {"source": "outputs/report.pdf"},
+                    None,  # pyright: ignore[reportArgumentType]
+                )
+        assert "require an active shell layer" in result
+
+    asyncio.run(scenario())
+
+
+def test_plugin_tool_file_context_uploads_sandbox_path_and_resolves_signed_url() -> None:
+    class FakeShell:
+        script: str | None = None
+
+        async def run_remote_script_complete(self, script: str, **_kwargs: object) -> object:
+            self.script = script
+            return SimpleNamespace(
+                exit_code=0,
+                status="done",
+                output_complete=True,
+                output=(
+                    "noise\n"
+                    "<<<DIFY_PLUGIN_TOOL_FILE_UPLOAD_BEGIN>>>"
+                    '{"transfer_method":"tool_file","reference":"dify-file-ref:file-1"}'
+                    "<<<DIFY_PLUGIN_TOOL_FILE_UPLOAD_END>>>\n"
+                ),
+            )
+
+    class FakeFileClient:
+        async def request_download(self, **_kwargs: object) -> object:
+            return SimpleNamespace(
+                filename="report.pdf",
+                mime_type="application/pdf",
+                size=42,
+                download_url="https://signed.example/report.pdf",
+            )
+
+    async def scenario() -> None:
+        shell = FakeShell()
+        context = _PluginToolFileContext(
+            file_client=FakeFileClient(),  # type: ignore[arg-type]
+            execution_context=_execution_context_config(),
+            shell=shell,  # type: ignore[arg-type]
+        )
+        result = await context.to_plugin_file_parameter("outputs/report.pdf")
+
+        assert result == {
+            "dify_model_identity": "__dify__file__",
+            "type": "custom",
+            "url": "https://signed.example/report.pdf",
+            "filename": "report.pdf",
+            "mime_type": "application/pdf",
+            "extension": ".pdf",
+            "size": 42,
+        }
+        assert shell.script is not None
+        assert "outputs/report.pdf" in shell.script
+
+    asyncio.run(scenario())
+
+
 def test_dify_plugin_tools_layer_requires_hidden_runtime_parameters_in_prepared_config() -> None:
     async def scenario() -> None:
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=_tool_transport()) as client:
@@ -531,7 +888,9 @@ def test_dify_plugin_tools_layer_requires_hidden_runtime_parameters_in_prepared_
                 }
             ) as run:
                 with pytest.raises(ValueError, match="requires non-LLM runtime_parameters for: auth_scope"):
-                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client)
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
 
     asyncio.run(scenario())
 
@@ -541,7 +900,7 @@ def test_dify_plugin_tools_layer_returns_agent_friendly_error_text() -> None:
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(
@@ -555,7 +914,11 @@ def test_dify_plugin_tools_layer_returns_agent_friendly_error_text() -> None:
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": _tools_config()}
             ) as run:
-                tool = (await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client))[0]
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
+                )[0]
                 result = await tool.function_schema.call(
                     {"query": "dify", "region": "global"},
                     None,  # pyright: ignore[reportArgumentType]
@@ -571,7 +934,7 @@ def test_dify_plugin_tools_layer_propagates_unexpected_transport_errors() -> Non
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
 
@@ -585,7 +948,11 @@ def test_dify_plugin_tools_layer_propagates_unexpected_transport_errors() -> Non
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": _tools_config()}
             ) as run:
-                tool = (await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client))[0]
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
+                )[0]
 
                 with pytest.raises(RuntimeError, match="unexpected transport failure"):
                     await tool.function_schema.call(
@@ -633,14 +1000,18 @@ def test_dify_plugin_tools_layer_maps_nested_plugin_invoke_errors_to_agent_text(
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=_tool_transport(invoke_error_payload=invoke_error_payload)) as client:
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": _tools_config()}
             ) as run:
-                tool = (await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client))[0]
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
+                )[0]
                 result = await tool.function_schema.call(
                     {"query": "dify", "region": "global"},
                     None,  # pyright: ignore[reportArgumentType]
@@ -656,14 +1027,18 @@ def test_dify_plugin_tools_layer_merges_blob_chunks_before_observation_conversio
         compositor = Compositor(
             [
                 LayerNode("execution_context", _execution_context_provider()),
-                LayerNode("tools", DifyPluginToolsLayer, deps={"execution_context": "execution_context"}),
+                LayerNode("tools", _tools_provider(), deps={"execution_context": "execution_context"}),
             ]
         )
         async with httpx.AsyncClient(transport=_tool_transport(chunked_blob=True)) as client:
             async with compositor.enter(
                 configs={"execution_context": _execution_context_config(), "tools": _tools_config()}
             ) as run:
-                tool = (await run.get_layer("tools", DifyPluginToolsLayer).get_tools(http_client=client))[0]
+                tool = (
+                    await run.get_layer("tools", DifyPluginToolsLayer).get_tools(
+                        http_client=client, dify_api_http_client=client
+                    )
+                )[0]
                 result = await tool.function_schema.call(
                     {"query": "dify", "region": "global"},
                     None,  # pyright: ignore[reportArgumentType]
diff --git a/dify-agent/tests/local/dify_agent/layers/execution_context/test_configs.py b/dify-agent/tests/local/dify_agent/layers/execution_context/test_configs.py
index cf41a884f4c..2c26f0dfb14 100644
--- a/dify-agent/tests/local/dify_agent/layers/execution_context/test_configs.py
+++ b/dify-agent/tests/local/dify_agent/layers/execution_context/test_configs.py
@@ -8,6 +8,7 @@ from dify_agent.layers.execution_context import DifyExecutionContextLayerConfig
 def test_execution_context_package_exports_client_safe_config_symbols_only() -> None:
     assert execution_context_exports.__all__ == [
         "DIFY_EXECUTION_CONTEXT_LAYER_TYPE_ID",
+        "DifyExecutionContextAgentConfigVersionKind",
         "DifyExecutionContextAgentMode",
         "DifyExecutionContextInvokeFrom",
         "DifyExecutionContextLayerConfig",
diff --git a/dify-agent/tests/local/dify_agent/layers/shell/test_layer.py b/dify-agent/tests/local/dify_agent/layers/shell/test_layer.py
index f4bf3fa8b99..f2a702139bb 100644
--- a/dify-agent/tests/local/dify_agent/layers/shell/test_layer.py
+++ b/dify-agent/tests/local/dify_agent/layers/shell/test_layer.py
@@ -9,7 +9,7 @@ from typing import cast
 import pytest
 
 import dify_agent.layers.shell.layer as shell_layer_module
-from dify_agent.layers.shell import DIFY_SHELL_LAYER_TYPE_ID, DifyShellLayerConfig
+from dify_agent.layers.shell import DIFY_SHELL_LAYER_TYPE_ID, DifyShellEnvVarConfig, DifyShellLayerConfig
 from dify_agent.layers.shell.layer import (
     CompleteRemoteCommandResult,
     DEFAULT_TERMINATE_GRACE_SECONDS,
@@ -20,9 +20,11 @@ from dify_agent.adapters.shell.protocols import (
     ShellCommandResult,
     ShellCommandStatus,
     ShellFileTransferProtocol,
+    ShellProviderError,
     ShellProviderProtocol,
     ShellResourceProtocol,
 )
+from dify_agent.layers.execution_context import DifyExecutionContextLayerConfig
 
 
 def _command_result(
@@ -82,6 +84,17 @@ def _assert_error_observation(result: object, *, job_id: str | None = None, incl
         assert includes in result["error"]
 
 
+def _capture_logged_exceptions(monkeypatch: pytest.MonkeyPatch) -> list[tuple[str, tuple[object, ...]]]:
+    logged: list[tuple[str, tuple[object, ...]]] = []
+
+    def fake_exception(message: str, *args: object, **kwargs: object) -> None:
+        del kwargs
+        logged.append((message, args))
+
+    monkeypatch.setattr(shell_layer_module.logger, "exception", fake_exception)
+    return logged
+
+
 @dataclass(slots=True)
 class RunCall:
     script: str
@@ -123,6 +136,10 @@ class DeleteCall:
     grace_seconds: float | None
 
 
+class _UnexpectedToolError(Exception):
+    pass
+
+
 class FakeFiles(ShellFileTransferProtocol):
     async def upload(self, *, content: bytes, remote_path: str, cwd: str | None = None) -> None:
         raise AssertionError("resource.files should not be used by production shell layer logic")
@@ -216,6 +233,43 @@ def _layer(
     return layer, provider
 
 
+@dataclass(slots=True)
+class _ExecutionContextStub:
+    config: DifyExecutionContextLayerConfig
+
+
+def _bind_execution_context(layer: DifyShellLayer, *, agent_id: str | None = "agent-1") -> None:
+    layer.deps.execution_context = cast(
+        object, _ExecutionContextStub(config=_execution_context_config(agent_id=agent_id))
+    )
+
+
+def _execution_context_config(*, agent_id: str | None = None) -> DifyExecutionContextLayerConfig:
+    return DifyExecutionContextLayerConfig(
+        tenant_id="tenant-1",
+        user_id="user-1",
+        user_from="account",
+        agent_id=agent_id,
+        agent_mode="agent_app",
+        invoke_from="service-api",
+    )
+
+
+def _runtime_state(
+    *,
+    session_id: str = "abc12ff",
+    workspace_cwd: str = "~/workspace/abc12ff",
+    job_ids: list[str] | None = None,
+    job_offsets: dict[str, int] | None = None,
+) -> DifyShellRuntimeState:
+    return DifyShellRuntimeState(
+        session_id=session_id,
+        workspace_cwd=workspace_cwd,
+        job_ids=[] if job_ids is None else job_ids,
+        job_offsets={} if job_offsets is None else job_offsets,
+    )
+
+
 def test_shell_type_id_constant_matches_implementation_class() -> None:
     assert DIFY_SHELL_LAYER_TYPE_ID == DifyShellLayer.type_id
 
@@ -235,13 +289,15 @@ def test_resource_context_calls_provider_create_and_resource_close() -> None:
 def test_shell_layer_create_allocates_workspace_and_bootstraps(monkeypatch: pytest.MonkeyPatch) -> None:
     monkeypatch.setattr(shell_layer_module.time, "time", lambda: int("abc12", 16))
     monkeypatch.setattr(shell_layer_module.secrets, "token_hex", lambda _nbytes: "ff")
+    expected_home = "/home/agent-1"
+    expected_workspace_cwd = "/home/agent-1/workspace/abc12ff"
 
     def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
-        assert env is None
+        assert env == {"HOME": expected_home}
         if cwd is None:
             assert 'mkdir -p "$HOME/workspace"' in script
             return _command_result("mkdir-job", status="exited", done=True, exit_code=0)
-        assert cwd == "~/workspace/abc12ff"
+        assert cwd == expected_workspace_cwd
         assert "apt-get install -y ripgrep" in script
         return _command_result("bootstrap-job", status="exited", done=True, exit_code=0)
 
@@ -251,6 +307,7 @@ def test_shell_layer_create_allocates_workspace_and_bootstraps(monkeypatch: pyte
             cli_tools=[{"name": "ripgrep", "install_commands": ["apt-get install -y ripgrep"]}],
         ),
     )
+    _bind_execution_context(layer)
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -259,13 +316,56 @@ def test_shell_layer_create_allocates_workspace_and_bootstraps(monkeypatch: pyte
 
     asyncio.run(scenario())
 
-    assert layer.runtime_state == DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    assert layer.runtime_state.session_id == "abc12ff"
+    assert layer.runtime_state.workspace_cwd == "~/workspace/abc12ff"
     assert [call.job_id for call in provider.resource.commands.delete_calls] == ["mkdir-job", "bootstrap-job"]
 
 
+def test_shell_layer_uses_agent_specific_home_and_workspace_cwd(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    monkeypatch.setattr(shell_layer_module.time, "time", lambda: int("abc12", 16))
+    monkeypatch.setattr(shell_layer_module.secrets, "token_hex", lambda _nbytes: "ff")
+
+    expected_home = "/home/agent-1"
+    expected_workspace_cwd = "/home/agent-1/workspace/abc12ff"
+
+    def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
+        del timeout
+        if script.startswith('mkdir -p "$HOME/workspace";'):
+            assert cwd is None
+            assert env == {"HOME": expected_home}
+            return _command_result("mkdir-job", status="exited", done=True, exit_code=0)
+        if script == "pwd":
+            assert cwd == expected_workspace_cwd
+            assert env == {"HOME": expected_home}
+            return _command_result("user-job", status="exited", done=True, exit_code=0, output=expected_home, offset=13)
+        raise AssertionError(f"Unexpected script: {script!r}")
+
+    commands = FakeCommands(run_handler=run_handler)
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    tools = {tool.name: tool for tool in layer.tools}
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            await layer.on_context_create()
+            run_result = await tools["shell_run"].function_schema.call({"script": "pwd"}, None)  # pyright: ignore[reportArgumentType]
+            metadata, output = _parse_tagged_observation(run_result)
+            assert metadata["job_id"] == "user-job"
+            assert output == expected_home
+
+    asyncio.run(scenario())
+
+    assert layer.runtime_state.session_id == "abc12ff"
+    assert layer.runtime_state.workspace_cwd == "~/workspace/abc12ff"
+    assert layer.runtime_state.job_ids == ["user-job"]
+    assert [call.job_id for call in commands.delete_calls] == ["mkdir-job"]
+
+
 def test_shell_layer_suspend_does_not_close_before_resource_context_exits() -> None:
     layer, provider = _layer(commands=FakeCommands())
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -276,6 +376,46 @@ def test_shell_layer_suspend_does_not_close_before_resource_context_exits() -> N
     asyncio.run(scenario())
 
 
+def test_shell_layer_resume_recreates_live_home_and_workspace() -> None:
+    expected_home = "/home/agent-1"
+
+    def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
+        del timeout
+        assert script == 'mkdir -p "$HOME/workspace/abc12ff"'
+        assert cwd is None
+        assert env == {"HOME": expected_home}
+        return _command_result("resume-job", status="exited", done=True, exit_code=0)
+
+    commands = FakeCommands(run_handler=run_handler)
+    layer, provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            await layer.on_context_resume()
+
+    asyncio.run(scenario())
+
+    assert [call.job_id for call in commands.delete_calls] == ["resume-job"]
+    assert provider.resource.closed is True
+
+
+def test_shell_layer_resume_requires_agent_id_for_live_workspace_reentry() -> None:
+    commands = FakeCommands()
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer, agent_id=None)
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            with pytest.raises(ValueError, match="requires execution_context\\.agent_id"):
+                await layer.on_context_resume()
+
+    asyncio.run(scenario())
+    assert commands.run_calls == []
+
+
 def test_shell_layer_delete_cleans_workspace_and_tracked_jobs() -> None:
     def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
         del env, timeout
@@ -285,12 +425,8 @@ def test_shell_layer_delete_cleans_workspace_and_tracked_jobs() -> None:
 
     commands = FakeCommands(run_handler=run_handler)
     layer, provider = _layer(commands=commands)
-    layer.runtime_state = DifyShellRuntimeState(
-        session_id="abc12ff",
-        workspace_cwd="~/workspace/abc12ff",
-        job_ids=["user-job"],
-        job_offsets={"user-job": 9},
-    )
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 9})
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -307,8 +443,8 @@ def test_shell_layer_delete_cleans_workspace_and_tracked_jobs() -> None:
 def test_shell_layer_tools_map_inputs_and_maintain_offsets_with_tail_end() -> None:
     def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
         assert script == "pwd"
-        assert cwd == "~/workspace/abc12ff"
-        assert env is None
+        assert cwd == "/home/agent-1/workspace/abc12ff"
+        assert env == {"HOME": "/home/agent-1"}
         return _command_result(
             "user-job",
             status="running",
@@ -354,8 +490,9 @@ def test_shell_layer_tools_map_inputs_and_maintain_offsets_with_tail_end() -> No
         interrupt_handler=interrupt_handler,
     )
     layer, provider = _layer(commands=commands)
+    _bind_execution_context(layer)
     tools = {tool.name: tool for tool in layer.tools}
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -410,8 +547,8 @@ def test_shell_layer_tools_map_inputs_and_maintain_offsets_with_tail_end() -> No
 def test_shell_run_keeps_original_offset_when_tail_lookup_fails_for_truncated_output() -> None:
     def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
         assert script == "pwd"
-        assert cwd == "~/workspace/abc12ff"
-        assert env is None
+        assert cwd == "/home/agent-1/workspace/abc12ff"
+        assert env == {"HOME": "/home/agent-1"}
         return _command_result(
             "user-job",
             status="running",
@@ -427,8 +564,9 @@ def test_shell_run_keeps_original_offset_when_tail_lookup_fails_for_truncated_ou
 
     commands = FakeCommands(run_handler=run_handler, tail_handler=tail_handler)
     layer, provider = _layer(commands=commands)
+    _bind_execution_context(layer)
     tools = {tool.name: tool for tool in layer.tools}
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -465,8 +603,9 @@ def test_shell_run_formats_large_non_truncated_output_without_tail_lookup() -> N
         )
     )
     layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
     tools = {tool.name: tool for tool in layer.tools}
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -488,12 +627,7 @@ def test_shell_interrupt_succeeds_when_tail_lookup_fails() -> None:
     )
     layer, _provider = _layer(commands=commands)
     tools = {tool.name: tool for tool in layer.tools}
-    layer.runtime_state = DifyShellRuntimeState(
-        session_id="abc12ff",
-        workspace_cwd="~/workspace/abc12ff",
-        job_ids=["user-job"],
-        job_offsets={"user-job": 22},
-    )
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 22})
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -510,14 +644,270 @@ def test_shell_interrupt_succeeds_when_tail_lookup_fails() -> None:
     asyncio.run(scenario())
 
 
+def test_shell_run_returns_provider_timeout_error_observation_without_unexpected_logging(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        run_handler=lambda script, cwd, env, timeout: (_ for _ in ()).throw(
+            ShellProviderError("provider timed out", code="timeout")
+        )
+    )
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_run"].function_schema.call({"script": "pwd"}, None)  # pyright: ignore[reportArgumentType]
+            _assert_error_observation(result, includes="timeout")
+
+    asyncio.run(scenario())
+    assert logged == []
+
+
+def test_shell_wait_returns_provider_request_error_observation_with_job_id_and_no_unexpected_logging(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        wait_handler=lambda job_id, offset, timeout: (_ for _ in ()).throw(
+            ShellProviderError("provider unavailable", code="request_error")
+        )
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 3})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_wait"].function_schema.call(
+                {"job_id": "user-job", "timeout": 4.0},
+                None,  # pyright: ignore[reportArgumentType]
+            )
+            _assert_error_observation(result, job_id="user-job", includes="request_error")
+
+    asyncio.run(scenario())
+    assert logged == []
+
+
+def test_shell_input_returns_provider_timeout_error_observation_with_job_id_and_no_unexpected_logging(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        input_handler=lambda job_id, text, offset, timeout: (_ for _ in ()).throw(
+            ShellProviderError("provider timed out", code="timeout")
+        )
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 6})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_input"].function_schema.call(
+                {"job_id": "user-job", "text": "ls\n", "timeout": 5.0},
+                None,  # pyright: ignore[reportArgumentType]
+            )
+            _assert_error_observation(result, job_id="user-job", includes="timeout")
+
+    asyncio.run(scenario())
+    assert logged == []
+
+
+def test_shell_interrupt_returns_provider_request_error_observation_with_job_id_and_no_unexpected_logging(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        interrupt_handler=lambda job_id, grace_seconds: (_ for _ in ()).throw(
+            ShellProviderError("provider unavailable", code="request_error")
+        )
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 22})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_interrupt"].function_schema.call(
+                {"job_id": "user-job", "grace_seconds": 1.5},
+                None,  # pyright: ignore[reportArgumentType]
+            )
+            _assert_error_observation(result, job_id="user-job", includes="request_error")
+
+    asyncio.run(scenario())
+    assert logged == []
+
+
+def test_shell_run_returns_error_observation_and_logs_unexpected_exception(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        run_handler=lambda script, cwd, env, timeout: (_ for _ in ()).throw(_UnexpectedToolError("boom"))
+    )
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_run"].function_schema.call({"script": "pwd"}, None)  # pyright: ignore[reportArgumentType]
+            _assert_error_observation(result, includes="boom")
+
+    asyncio.run(scenario())
+    assert logged == [
+        ("Unexpected shell tool failure: tool=%s session_id=%s job_id=%s", ("shell_run", "abc12ff", None))
+    ]
+
+
+def test_shell_wait_returns_error_observation_and_logs_unexpected_exception(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        wait_handler=lambda job_id, offset, timeout: (_ for _ in ()).throw(_UnexpectedToolError("wait exploded"))
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 3})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_wait"].function_schema.call(
+                {"job_id": "user-job", "timeout": 4.0},
+                None,  # pyright: ignore[reportArgumentType]
+            )
+            _assert_error_observation(result, job_id="user-job", includes="wait exploded")
+
+    asyncio.run(scenario())
+    assert logged == [
+        ("Unexpected shell tool failure: tool=%s session_id=%s job_id=%s", ("shell_wait", "abc12ff", "user-job"))
+    ]
+
+
+def test_shell_input_returns_error_observation_and_logs_unexpected_exception(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        input_handler=lambda job_id, text, offset, timeout: (_ for _ in ()).throw(
+            _UnexpectedToolError("stdin exploded")
+        )
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 6})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_input"].function_schema.call(
+                {"job_id": "user-job", "text": "ls\n", "timeout": 5.0},
+                None,  # pyright: ignore[reportArgumentType]
+            )
+            _assert_error_observation(result, job_id="user-job", includes="stdin exploded")
+
+    asyncio.run(scenario())
+    assert logged == [
+        (
+            "Unexpected shell tool failure: tool=%s session_id=%s job_id=%s",
+            ("shell_input", "abc12ff", "user-job"),
+        )
+    ]
+
+
+def test_shell_interrupt_returns_error_observation_and_logs_unexpected_exception(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        interrupt_handler=lambda job_id, grace_seconds: (_ for _ in ()).throw(
+            _UnexpectedToolError("interrupt exploded")
+        )
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 22})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_interrupt"].function_schema.call(
+                {"job_id": "user-job", "grace_seconds": 1.5},
+                None,  # pyright: ignore[reportArgumentType]
+            )
+            _assert_error_observation(result, job_id="user-job", includes="interrupt exploded")
+
+    asyncio.run(scenario())
+    assert logged == [
+        (
+            "Unexpected shell tool failure: tool=%s session_id=%s job_id=%s",
+            ("shell_interrupt", "abc12ff", "user-job"),
+        )
+    ]
+
+
+def test_shell_interrupt_logs_unexpected_tail_failure_but_still_succeeds(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    logged = _capture_logged_exceptions(monkeypatch)
+    commands = FakeCommands(
+        interrupt_handler=lambda job_id, grace_seconds: _command_status(job_id, offset=22),
+        tail_handler=lambda job_id: (_ for _ in ()).throw(_UnexpectedToolError("tail exploded")),
+    )
+    layer, _provider = _layer(commands=commands)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state(job_ids=["user-job"], job_offsets={"user-job": 22})
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await tools["shell_interrupt"].function_schema.call({"job_id": "user-job"}, None)  # pyright: ignore[reportArgumentType]
+            metadata, output = _parse_tagged_observation(result)
+            assert metadata == {
+                "job_id": "user-job",
+                "status": "terminated",
+                "done": True,
+                "exit_code": 130,
+            }
+            assert output == "Job was interrupted."
+
+    asyncio.run(scenario())
+    assert logged == [
+        (
+            "Failed to fetch output path for interrupted shell job %s in session %s",
+            ("user-job", "abc12ff"),
+        )
+    ]
+
+
+def test_shell_run_propagates_cancelled_error() -> None:
+    commands = FakeCommands(
+        run_handler=lambda script, cwd, env, timeout: (_ for _ in ()).throw(asyncio.CancelledError())
+    )
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    tools = {tool.name: tool for tool in layer.tools}
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            with pytest.raises(asyncio.CancelledError):
+                await tools["shell_run"].function_schema.call({"script": "pwd"}, None)  # pyright: ignore[reportArgumentType]
+
+    asyncio.run(scenario())
+
+
 def test_run_remote_script_complete_uses_read_output_before_wait_and_deletes_job() -> None:
     events: list[str] = []
 
     def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
         events.append("run")
         assert script == "printf 'abcdefghi'"
-        assert cwd == "~/workspace/abc12ff"
-        assert env is None
+        assert cwd == "/home/agent-1/workspace/abc12ff"
+        assert env == {"HOME": "/home/agent-1"}
         return _command_result("remote-job", status="running", done=False, output="abc", offset=3, truncated=True)
 
     def wait_handler(job_id: str, offset: int, timeout: float) -> ShellCommandResult:
@@ -531,7 +921,8 @@ def test_run_remote_script_complete_uses_read_output_before_wait_and_deletes_job
 
     commands = FakeCommands(run_handler=run_handler, wait_handler=wait_handler)
     layer, _provider = _layer(commands=commands)
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -546,20 +937,126 @@ def test_run_remote_script_complete_uses_read_output_before_wait_and_deletes_job
     assert [call.job_id for call in commands.delete_calls] == ["remote-job"]
 
 
+def test_run_remote_script_complete_uses_agent_specific_home_and_workspace_cwd() -> None:
+    expected_home = "/home/agent-1"
+    expected_workspace_cwd = "/home/agent-1/workspace/abc12ff"
+
+    def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
+        del timeout
+        assert script == "pwd"
+        assert cwd == expected_workspace_cwd
+        assert env == {"HOME": expected_home}
+        return _command_result(
+            "remote-job",
+            status="exited",
+            done=True,
+            exit_code=0,
+            output=expected_home,
+            offset=len(expected_home),
+        )
+
+    commands = FakeCommands(run_handler=run_handler)
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await layer.run_remote_script_complete("pwd")
+            assert isinstance(result, CompleteRemoteCommandResult)
+            assert result.output == expected_home
+
+    asyncio.run(scenario())
+    assert [call.job_id for call in commands.delete_calls] == ["remote-job"]
+
+
+def test_run_remote_script_complete_passes_config_env_values_to_shellctl_env() -> None:
+    def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
+        del timeout
+        assert script == "printenv API_TOKEN"
+        assert cwd == "/home/agent-1/workspace/abc12ff"
+        assert env == {"HOME": "/home/agent-1", "API_TOKEN": "inline-secret-value"}
+        return _command_result(
+            "remote-job",
+            status="exited",
+            done=True,
+            exit_code=0,
+            output="inline-secret-value\n",
+            offset=len("inline-secret-value\n"),
+        )
+
+    commands = FakeCommands(run_handler=run_handler)
+    layer, _provider = _layer(
+        commands=commands,
+        config=DifyShellLayerConfig(env=[DifyShellEnvVarConfig(name="API_TOKEN", value="inline-secret-value")]),
+    )
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await layer.run_remote_script_complete("printenv API_TOKEN")
+            assert isinstance(result, CompleteRemoteCommandResult)
+            assert result.output == "inline-secret-value\n"
+
+    asyncio.run(scenario())
+    assert [call.job_id for call in commands.delete_calls] == ["remote-job"]
+
+
+def test_run_remote_script_uses_agent_specific_home_and_workspace_cwd() -> None:
+    expected_home = "/home/agent-1"
+    expected_workspace_cwd = "/home/agent-1/workspace/abc12ff"
+
+    def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
+        del timeout
+        assert script == "pwd"
+        assert cwd == expected_workspace_cwd
+        assert env == {"HOME": expected_home}
+        return _command_result(
+            "remote-job",
+            status="exited",
+            done=True,
+            exit_code=0,
+            output=expected_home,
+            offset=len(expected_home),
+        )
+
+    commands = FakeCommands(run_handler=run_handler)
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            result = await layer.run_remote_script("pwd")
+            assert isinstance(result, CompleteRemoteCommandResult)
+            assert result.output == expected_home
+
+    asyncio.run(scenario())
+    assert [call.job_id for call in commands.delete_calls] == ["remote-job"]
+
+
 def test_run_remote_script_complete_returns_incomplete_reason_for_output_limit() -> None:
-    commands = FakeCommands(
-        run_handler=lambda script, cwd, env, timeout: _command_result(
+    def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
+        del script, timeout
+        assert cwd == "/home/agent-1/workspace/abc12ff"
+        assert env == {"HOME": "/home/agent-1"}
+        return _command_result(
             "remote-job",
             status="running",
             done=False,
             output="hello world",
             offset=11,
             truncated=True,
-        ),
+        )
+
+    commands = FakeCommands(
+        run_handler=run_handler,
         interrupt_handler=lambda job_id, grace_seconds: _command_status(job_id, status="terminated", offset=11),
     )
     layer, _provider = _layer(commands=commands)
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -589,8 +1086,8 @@ def test_run_remote_script_complete_returns_incomplete_reason_for_timeout(
     def run_handler(script: str, cwd: str | None, env: Mapping[str, str] | None, timeout: float) -> ShellCommandResult:
         nonlocal now
         assert script == "sleep 10"
-        assert cwd == "~/workspace/abc12ff"
-        assert env is None
+        assert cwd == "/home/agent-1/workspace/abc12ff"
+        assert env == {"HOME": "/home/agent-1"}
         assert timeout == pytest.approx(60.0, rel=0, abs=0.01)
         now = 161.0
         return _command_result("remote-job", status="running", done=False, output="hello", offset=5)
@@ -606,7 +1103,8 @@ def test_run_remote_script_complete_returns_incomplete_reason_for_timeout(
         ),
     )
     layer, _provider = _layer(commands=commands)
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    _bind_execution_context(layer)
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -629,7 +1127,7 @@ def test_shell_layer_rejects_untracked_job_ids_without_provider_calls() -> None:
     commands = FakeCommands()
     layer, _provider = _layer(commands=commands)
     tools = {tool.name: tool for tool in layer.tools}
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         async with layer.resource_context():
@@ -649,10 +1147,26 @@ def test_shell_layer_rejects_untracked_job_ids_without_provider_calls() -> None:
     assert commands.interrupt_calls == []
 
 
+def test_shell_layer_requires_agent_id_for_live_command_execution(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setattr(shell_layer_module.time, "time", lambda: int("abc12", 16))
+    monkeypatch.setattr(shell_layer_module.secrets, "token_hex", lambda _nbytes: "ff")
+    commands = FakeCommands()
+    layer, _provider = _layer(commands=commands)
+    _bind_execution_context(layer, agent_id=None)
+
+    async def scenario() -> None:
+        async with layer.resource_context():
+            with pytest.raises(ValueError, match="requires execution_context\\.agent_id"):
+                await layer.on_context_create()
+
+    asyncio.run(scenario())
+    assert commands.run_calls == []
+
+
 def test_shell_layer_hooks_and_tools_fail_clearly_outside_active_resource_context() -> None:
     layer, _provider = _layer(commands=FakeCommands())
     tools = {tool.name: tool for tool in layer.tools}
-    layer.runtime_state = DifyShellRuntimeState(session_id="abc12ff", workspace_cwd="~/workspace/abc12ff")
+    layer.runtime_state = _runtime_state()
 
     async def scenario() -> None:
         result = await tools["shell_run"].function_schema.call({"script": "pwd"}, None)  # pyright: ignore[reportArgumentType]
diff --git a/dify-agent/tests/local/dify_agent/protocol/test_protocol_schemas.py b/dify-agent/tests/local/dify_agent/protocol/test_protocol_schemas.py
index b58e0818229..75fa8dda204 100644
--- a/dify-agent/tests/local/dify_agent/protocol/test_protocol_schemas.py
+++ b/dify-agent/tests/local/dify_agent/protocol/test_protocol_schemas.py
@@ -12,6 +12,7 @@ from dify_agent.layers.dify_plugin import DIFY_PLUGIN_LLM_LAYER_TYPE_ID, DIFY_PL
 from dify_agent.layers.output import DIFY_OUTPUT_LAYER_TYPE_ID, DifyOutputLayerConfig
 from dify_agent.protocol import DIFY_AGENT_HISTORY_LAYER_ID, DIFY_AGENT_MODEL_LAYER_ID, DIFY_AGENT_OUTPUT_LAYER_ID
 from dify_agent.protocol.schemas import (
+    AgentRunUsage,
     RUN_EVENT_ADAPTER,
     CreateRunRequest,
     DeferredToolCallPayload,
@@ -171,6 +172,7 @@ def test_create_run_request_accepts_dto_first_public_composition_and_normalizes_
         "conversation_id": None,
         "agent_id": None,
         "agent_config_version_id": None,
+        "agent_config_version_kind": None,
         "agent_mode": "workflow_run",
         "invoke_from": "service-api",
         "trace_id": "trace-1",
@@ -403,6 +405,27 @@ def test_run_succeeded_event_round_trips_explicit_json_null_output() -> None:
     assert b'"deferred_tool_call"' not in payload
 
 
+def test_run_succeeded_event_round_trips_usage() -> None:
+    event = RunSucceededEvent(
+        run_id="run-usage",
+        data=RunSucceededEventData(
+            output="done",
+            session_snapshot=CompositorSessionSnapshot(layers=[]),
+            usage=AgentRunUsage(prompt_tokens=3, completion_tokens=5),
+        ),
+    )
+
+    payload = RUN_EVENT_ADAPTER.dump_json(event)
+    decoded = RUN_EVENT_ADAPTER.validate_json(payload)
+
+    assert isinstance(decoded, RunSucceededEvent)
+    assert decoded.data.usage is not None
+    assert decoded.data.usage.prompt_tokens == 3
+    assert decoded.data.usage.completion_tokens == 5
+    assert decoded.data.usage.total_tokens == 8
+    assert b'"usage"' in payload
+
+
 def test_on_exit_accept_layer_overrides() -> None:
     request = CreateRunRequest.model_validate(
         {
diff --git a/dify-agent/tests/local/dify_agent/runtime/test_runner.py b/dify-agent/tests/local/dify_agent/runtime/test_runner.py
index b9f5b5bf467..6c05cb65703 100644
--- a/dify-agent/tests/local/dify_agent/runtime/test_runner.py
+++ b/dify-agent/tests/local/dify_agent/runtime/test_runner.py
@@ -384,6 +384,8 @@ def test_runner_emits_terminal_success_and_snapshot(monkeypatch: pytest.MonkeyPa
     terminal = sink.events["run-1"][-1]
     assert isinstance(terminal, RunSucceededEvent)
     assert terminal.data.output == "done"
+    assert terminal.data.usage is not None
+    assert terminal.data.usage.total_tokens > 0
     assert [layer.name for layer in terminal.data.session_snapshot.layers] == [
         "prompt",
         "renamed-execution-context",
diff --git a/dify-agent/tests/local/dify_agent/server/test_sandbox_files.py b/dify-agent/tests/local/dify_agent/server/test_sandbox_files.py
index eae7f945bfb..880aa94cfa2 100644
--- a/dify-agent/tests/local/dify_agent/server/test_sandbox_files.py
+++ b/dify-agent/tests/local/dify_agent/server/test_sandbox_files.py
@@ -3,15 +3,19 @@ from __future__ import annotations
 import asyncio
 import base64
 import json
+import os
+from pathlib import Path
+import subprocess
+import sys
 from collections.abc import Callable, Mapping
 from dataclasses import dataclass
-from typing import Literal
+from typing import Literal, cast
 
 import pytest
 from agenton.compositor import CompositorSessionSnapshot, LayerProvider
 from agenton.compositor.schemas import LayerSessionSnapshot
 from agenton.layers.base import LifecycleState
-from dify_agent.adapters.shell.shellctl import ShellctlProvider
+from dify_agent.adapters.shell.shellctl import ShellctlClientProtocol, ShellctlProvider
 from dify_agent.agent_stub.server.shell_agent_stub_env import (
     AGENT_STUB_API_BASE_URL_ENV_VAR,
     AGENT_STUB_AUTH_JWE_ENV_VAR,
@@ -32,10 +36,13 @@ from dify_agent.protocol import (
     build_sandbox_locator_from_run_request,
 )
 from dify_agent.server.sandbox_files import (
+    _LIST_SCRIPT,
     SandboxFileError,
     SandboxFileService,
     _OUTPUT_BEGIN,
     _OUTPUT_END,
+    _READ_SCRIPT,
+    _UPLOAD_SCRIPT,
     _decode_sandbox_payload,
     _shell_result_details,
 )
@@ -57,35 +64,35 @@ class _Job:
 class RunCall:
     script: str
     cwd: str | None
-    env: Mapping[str, str] | None
+    env: dict[str, str] | None
     timeout: float
 
 
 class FakeShellctlClient:
-    def __init__(self, *, run_handler: Callable[[str, str | None, Mapping[str, str] | None, float], _Job]) -> None:
+    def __init__(self, *, run_handler: Callable[[str, str | None, dict[str, str] | None, float], _Job]) -> None:
         self.run_handler = run_handler
         self.run_calls: list[RunCall] = []
         self.delete_calls: list[str] = []
 
     async def run(
-        self, script: str, *, cwd: str | None = None, env: Mapping[str, str] | None = None, timeout: float = 10.0
-    ):
+        self, script: str, *, cwd: str | None = None, env: dict[str, str] | None = None, timeout: float = 10.0
+    ) -> _Job:
         self.run_calls.append(RunCall(script=script, cwd=cwd, env=env, timeout=timeout))
         return self.run_handler(script, cwd, env, timeout)
 
-    async def wait(self, job_id: str, *, offset: int, timeout: float = 10.0):
+    async def wait(self, job_id: str, *, offset: int, timeout: float = 10.0) -> _Job:
         raise AssertionError(f"Unexpected wait() call for {job_id} offset={offset} timeout={timeout}")
 
-    async def input(self, job_id: str, text: str, *, offset: int, timeout: float = 10.0):
+    async def input(self, job_id: str, text: str, *, offset: int, timeout: float = 10.0) -> _Job:
         raise AssertionError(f"Unexpected input() call for {job_id} text={text!r}")
 
-    async def tail(self, job_id: str):
+    async def tail(self, job_id: str) -> _Job:
         raise AssertionError(f"Unexpected tail() call for {job_id}")
 
-    async def terminate(self, job_id: str, grace_seconds: float = 10.0):
+    async def terminate(self, job_id: str, grace_seconds: float = 10.0) -> _Job:
         raise AssertionError(f"Unexpected terminate() call for {job_id} grace={grace_seconds}")
 
-    async def delete(self, job_id: str, *, force: bool = False, grace_seconds: float | None = None):
+    async def delete(self, job_id: str, *, force: bool = False, grace_seconds: float | None = None) -> None:
         del force, grace_seconds
         self.delete_calls.append(job_id)
         return None
@@ -125,6 +132,28 @@ def _complete_result(
     )
 
 
+def _run_embedded_script(
+    script_source: str,
+    *,
+    args: list[str],
+    cwd: Path,
+    env: Mapping[str, str] | None = None,
+) -> dict[str, object]:
+    merged_env = dict(os.environ)
+    if env is not None:
+        merged_env.update(env)
+    completed = subprocess.run(
+        [sys.executable, "-", *args],
+        input=script_source,
+        text=True,
+        capture_output=True,
+        cwd=cwd,
+        env=merged_env,
+        check=False,
+    )
+    return _decode_sandbox_payload(_complete_result(output=completed.stdout, exit_code=completed.returncode))
+
+
 def _execution_context() -> DifyExecutionContextLayerConfig:
     return DifyExecutionContextLayerConfig(
         tenant_id="tenant-1",
@@ -169,7 +198,7 @@ def _locator() -> SandboxLocator:
 
 
 def _service(
-    run_handler: Callable[[str, str | None, Mapping[str, str] | None, float], _Job],
+    run_handler: Callable[[str, str | None, dict[str, str] | None, float], _Job],
 ) -> tuple[SandboxFileService, FakeShellctlClient]:
     client = FakeShellctlClient(run_handler=run_handler)
     execution_context_provider = LayerProvider.from_factory(
@@ -187,7 +216,7 @@ def _service(
             shell_provider=ShellctlProvider(
                 entrypoint="http://shellctl",
                 token="",
-                client_factory=lambda: client,
+                client_factory=lambda: cast(ShellctlClientProtocol, cast(object, client)),
             ),
             agent_stub_api_base_url="https://agent.example.com/agent-stub",
             agent_stub_token_factory=lambda execution_context, *, session_id: (
@@ -198,6 +227,19 @@ def _service(
     return SandboxFileService(layer_providers=(execution_context_provider, shell_provider)), client
 
 
+def _sandbox_python_run_call(client: FakeShellctlClient) -> RunCall:
+    for run_call in reversed(client.run_calls):
+        if run_call.script.startswith("python3 - "):
+            return run_call
+    raise AssertionError("sandbox python script was not executed")
+
+
+def _sandbox_list_entries(payload: dict[str, object]) -> list[object]:
+    entries = payload.get("entries")
+    assert isinstance(entries, list)
+    return entries
+
+
 def test_list_files_runs_fixed_script_and_parses_response() -> None:
     service, client = _service(
         lambda script, cwd, env, timeout: _Job(
@@ -215,13 +257,153 @@ def test_list_files_runs_fixed_script_and_parses_response() -> None:
     result = asyncio.run(service.list_files(SandboxListRequest(locator=_locator(), path=".")))
 
     assert result.entries[0].name == "notes.txt"
-    assert client.run_calls[0].cwd == "~/workspace/abc12ff"
-    assert client.run_calls[0].env is None
-    assert "python3 - . 1000 <<'PY'" in client.run_calls[0].script
-    assert client.delete_calls == ["sandbox-job"]
+    script_call = _sandbox_python_run_call(client)
+    assert script_call.cwd == "/home/agent-1/workspace/abc12ff"
+    assert script_call.env == {"HOME": "/home/agent-1"}
+    assert "python3 - . 1000 <<'PY'" in script_call.script
+    assert client.delete_calls[-1] == "sandbox-job"
 
 
-@pytest.mark.parametrize("bad_path", ["/etc/passwd", "~/secret-dir", "bad\x00path"])
+def test_list_files_allows_parent_relative_paths() -> None:
+    service, client = _service(
+        lambda script, cwd, env, timeout: _Job(
+            job_id="sandbox-job",
+            output=_wrap({"path": "../shared", "entries": [], "truncated": False}),
+        )
+    )
+
+    result = asyncio.run(service.list_files(SandboxListRequest(locator=_locator(), path="../shared")))
+
+    assert result.path == "../shared"
+    assert "python3 - ../shared 1000 <<'PY'" in _sandbox_python_run_call(client).script
+
+
+@pytest.mark.parametrize(
+    ("path", "expected_command"),
+    [
+        ("~", "python3 - '~' 1000 <<'PY'"),
+        ("~/shared", "python3 - '~/shared' 1000 <<'PY'"),
+    ],
+)
+def test_list_files_allows_home_relative_paths(path: str, expected_command: str) -> None:
+    service, client = _service(
+        lambda script, cwd, env, timeout: _Job(
+            job_id="sandbox-job",
+            output=_wrap({"path": path, "entries": [], "truncated": False}),
+        )
+    )
+
+    result = asyncio.run(service.list_files(SandboxListRequest(locator=_locator(), path=path)))
+
+    assert result.path == path
+    assert expected_command in _sandbox_python_run_call(client).script
+
+
+def test_embedded_scripts_allow_parent_relative_paths(tmp_path: Path) -> None:
+    workspace_dir = tmp_path / "workspace"
+    cwd = workspace_dir / "run"
+    shared_dir = workspace_dir / "shared"
+    cwd.mkdir(parents=True)
+    shared_dir.mkdir()
+    notes_path = shared_dir / "notes.txt"
+    notes_path.write_text("hello", encoding="utf-8")
+
+    bin_dir = tmp_path / "bin"
+    bin_dir.mkdir()
+    fake_dify_agent = bin_dir / "dify-agent"
+    fake_dify_agent.write_text(
+        "\n".join(
+            [
+                "#!/usr/bin/env python3",
+                "import json",
+                "import sys",
+                'if sys.argv[1:] != ["file", "upload", "../shared/notes.txt"]:',
+                '    raise SystemExit(f"unexpected args: {sys.argv[1:]!r}")',
+                'print(json.dumps({"transfer_method": "tool_file", "reference": "file-ref"}))',
+            ]
+        )
+        + "\n",
+        encoding="utf-8",
+    )
+    fake_dify_agent.chmod(0o755)
+
+    list_payload = _run_embedded_script(_LIST_SCRIPT, args=["../shared", "1000"], cwd=cwd)
+    read_payload = _run_embedded_script(_READ_SCRIPT, args=["../shared/notes.txt", "8"], cwd=cwd)
+    upload_payload = _run_embedded_script(
+        _UPLOAD_SCRIPT,
+        args=["../shared/notes.txt"],
+        cwd=cwd,
+        env={"PATH": f"{bin_dir}{os.pathsep}{os.environ.get('PATH', os.defpath)}"},
+    )
+
+    assert list_payload["path"] == "../shared"
+    assert any(
+        isinstance(entry, dict) and entry.get("name") == "notes.txt" for entry in _sandbox_list_entries(list_payload)
+    )
+    assert read_payload == {
+        "path": "../shared/notes.txt",
+        "size": 5,
+        "truncated": False,
+        "binary": False,
+        "text": "hello",
+    }
+    assert upload_payload == {
+        "path": "../shared/notes.txt",
+        "file": {"transfer_method": "tool_file", "reference": "file-ref"},
+    }
+
+
+def test_embedded_scripts_expand_home_relative_paths(tmp_path: Path) -> None:
+    cwd = tmp_path / "workspace" / "run"
+    cwd.mkdir(parents=True)
+    home_dir = tmp_path / "home" / "agent-1"
+    shared_dir = home_dir / "shared"
+    shared_dir.mkdir(parents=True)
+    notes_path = shared_dir / "notes.txt"
+    notes_path.write_text("hello", encoding="utf-8")
+
+    bin_dir = tmp_path / "bin"
+    bin_dir.mkdir()
+    fake_dify_agent = bin_dir / "dify-agent"
+    fake_dify_agent.write_text(
+        "\n".join(
+            [
+                "#!/usr/bin/env python3",
+                "import json",
+                "import sys",
+                'if sys.argv[1:] != ["file", "upload", "~/shared/notes.txt"]:',
+                '    raise SystemExit(f"unexpected args: {sys.argv[1:]!r}")',
+                'print(json.dumps({"transfer_method": "tool_file", "reference": "file-ref"}))',
+            ]
+        )
+        + "\n",
+        encoding="utf-8",
+    )
+    fake_dify_agent.chmod(0o755)
+
+    script_env = {"HOME": str(home_dir), "PATH": f"{bin_dir}{os.pathsep}{os.environ.get('PATH', os.defpath)}"}
+    list_payload = _run_embedded_script(_LIST_SCRIPT, args=["~/shared", "1000"], cwd=cwd, env=script_env)
+    read_payload = _run_embedded_script(_READ_SCRIPT, args=["~/shared/notes.txt", "8"], cwd=cwd, env=script_env)
+    upload_payload = _run_embedded_script(_UPLOAD_SCRIPT, args=["~/shared/notes.txt"], cwd=cwd, env=script_env)
+
+    assert list_payload["path"] == "~/shared"
+    assert any(
+        isinstance(entry, dict) and entry.get("name") == "notes.txt" for entry in _sandbox_list_entries(list_payload)
+    )
+    assert read_payload == {
+        "path": "~/shared/notes.txt",
+        "size": 5,
+        "truncated": False,
+        "binary": False,
+        "text": "hello",
+    }
+    assert upload_payload == {
+        "path": "~/shared/notes.txt",
+        "file": {"transfer_method": "tool_file", "reference": "file-ref"},
+    }
+
+
+@pytest.mark.parametrize("bad_path", ["/etc/passwd", "~other/secret-dir", "bad\x00path"])
 def test_list_files_rejects_invalid_paths_before_shell_execution(bad_path: str) -> None:
     service, client = _service(lambda script, cwd, env, timeout: _Job(job_id="sandbox-job", output="unused"))
 
@@ -262,8 +444,10 @@ def test_upload_injects_agent_stub_env_and_returns_mapping() -> None:
 
     assert result.file.transfer_method == "tool_file"
     assert result.file.reference == "file-ref"
-    assert client.run_calls[0].cwd == "~/workspace/abc12ff"
-    assert client.run_calls[0].env == {
+    script_call = _sandbox_python_run_call(client)
+    assert script_call.cwd == "/home/agent-1/workspace/abc12ff"
+    assert script_call.env == {
+        "HOME": "/home/agent-1",
         AGENT_STUB_API_BASE_URL_ENV_VAR: "https://agent.example.com/agent-stub",
         AGENT_STUB_AUTH_JWE_ENV_VAR: "token-for:tenant-1:abc12ff",
         AGENT_STUB_DRIVE_BASE_ENV_VAR: "/mnt/drive/agent-1",
@@ -291,4 +475,38 @@ def test_read_file_uses_complete_mode_and_parses_response() -> None:
     result = asyncio.run(service.read_file(SandboxReadRequest(locator=_locator(), path="notes.txt", max_bytes=8)))
 
     assert result.text == "hello"
-    assert "python3 - notes.txt 8 <<'PY'" in client.run_calls[0].script
+    assert "python3 - notes.txt 8 <<'PY'" in _sandbox_python_run_call(client).script
+
+
+@pytest.mark.parametrize(
+    ("sandbox_request", "expected_command"),
+    [
+        (SandboxReadRequest(locator=_locator(), path="../notes.txt", max_bytes=8), "python3 - ../notes.txt 8 <<'PY'"),
+        (SandboxUploadRequest(locator=_locator(), path="../report.txt"), "python3 - ../report.txt <<'PY'"),
+        (SandboxReadRequest(locator=_locator(), path="~/notes.txt", max_bytes=8), "python3 - '~/notes.txt' 8 <<'PY'"),
+        (SandboxUploadRequest(locator=_locator(), path="~/report.txt"), "python3 - '~/report.txt' <<'PY'"),
+    ],
+)
+def test_read_and_upload_allow_relative_paths(
+    sandbox_request: SandboxReadRequest | SandboxUploadRequest, expected_command: str
+) -> None:
+    expected_path = sandbox_request.path
+    service, client = _service(
+        lambda script, cwd, env, timeout: _Job(
+            job_id="sandbox-job",
+            output=_wrap(
+                {"path": expected_path, "size": 5, "truncated": False, "binary": False, "text": "hello"}
+                if isinstance(sandbox_request, SandboxReadRequest)
+                else {"path": expected_path, "file": {"transfer_method": "tool_file", "reference": "file-ref"}}
+            ),
+        )
+    )
+
+    if isinstance(sandbox_request, SandboxReadRequest):
+        result = asyncio.run(service.read_file(sandbox_request))
+        assert result.path == expected_path
+    else:
+        result = asyncio.run(service.upload_file(sandbox_request))
+        assert result.path == expected_path
+
+    assert expected_command in _sandbox_python_run_call(client).script
diff --git a/dify-agent/tests/local/dify_agent/test_import_boundaries.py b/dify-agent/tests/local/dify_agent/test_import_boundaries.py
index c24941fae7f..9a5e3b34a73 100644
--- a/dify-agent/tests/local/dify_agent/test_import_boundaries.py
+++ b/dify-agent/tests/local/dify_agent/test_import_boundaries.py
@@ -112,7 +112,7 @@ def test_protocol_and_dify_plugin_exports_do_not_import_server_only_modules() ->
         assertions=[
             "assert hasattr(dify_agent_protocol, 'PydanticAIStreamRunEvent')",
             "assert dify_agent_layers_drive.__all__ == ['DIFY_DRIVE_LAYER_TYPE_ID', 'DifyDriveLayerConfig', 'DifyDriveSkillConfig']",
-            "assert dify_agent_layers_execution_context.__all__ == ['DIFY_EXECUTION_CONTEXT_LAYER_TYPE_ID', 'DifyExecutionContextAgentMode', 'DifyExecutionContextInvokeFrom', 'DifyExecutionContextLayerConfig', 'DifyExecutionContextUserFrom']",
+            "assert dify_agent_layers_execution_context.__all__ == ['DIFY_EXECUTION_CONTEXT_LAYER_TYPE_ID', 'DifyExecutionContextAgentConfigVersionKind', 'DifyExecutionContextAgentMode', 'DifyExecutionContextInvokeFrom', 'DifyExecutionContextLayerConfig', 'DifyExecutionContextUserFrom']",
             "assert dify_agent_layers_ask_human.__all__ == ['AskHumanAction', 'AskHumanActionStyle', 'AskHumanField', 'AskHumanFieldType', 'AskHumanFileField', 'AskHumanFileListField', 'AskHumanParagraphField', 'AskHumanResultStatus', 'AskHumanSelectField', 'AskHumanSelectOption', 'AskHumanSelectedAction', 'AskHumanToolArgs', 'AskHumanToolResult', 'AskHumanUrgency', 'DEFAULT_ASK_HUMAN_TOOL_DESCRIPTION', 'DIFY_ASK_HUMAN_LAYER_TYPE_ID', 'DifyAskHumanLayerConfig']",
             "assert dify_agent_layers_dify_plugin.__all__ == ['DIFY_PLUGIN_LLM_LAYER_TYPE_ID', 'DIFY_PLUGIN_TOOLS_LAYER_TYPE_ID', 'DifyPluginCredentialValue', 'DifyPluginLLMLayerConfig', 'DifyPluginToolCredentialType', 'DifyPluginToolConfig', 'DifyPluginToolOption', 'DifyPluginToolParameter', 'DifyPluginToolParameterForm', 'DifyPluginToolParameterType', 'DifyPluginToolsLayerConfig', 'DifyPluginToolValue']",
             "assert dify_agent_layers_knowledge.__all__ == ['DIFY_KNOWLEDGE_BASE_LAYER_TYPE_ID', 'DifyKnowledgeBaseLayerConfig', 'DifyKnowledgeDatasetConfig', 'DifyKnowledgeEagerResult', 'DifyKnowledgeMetadataCondition', 'DifyKnowledgeMetadataConditions', 'DifyKnowledgeMetadataFilteringConfig', 'DifyKnowledgeModelConfig', 'DifyKnowledgeQueryConfig', 'DifyKnowledgeRerankingModelConfig', 'DifyKnowledgeRetrievalConfig', 'DifyKnowledgeRuntimeState', 'DifyKnowledgeSetConfig']",
diff --git a/dify-agent/tests/local/test_packaging.py b/dify-agent/tests/local/test_packaging.py
index c91ce3a7af4..610c4850127 100644
--- a/dify-agent/tests/local/test_packaging.py
+++ b/dify-agent/tests/local/test_packaging.py
@@ -9,7 +9,7 @@ PROJECT_ROOT = Path(__file__).resolve().parents[2]
 CLIENT_SHARED_DTO_DEPENDENCIES = {
     "httpx==0.28.1",
     "pydantic>=2.12.5,<2.13",
-    "pydantic-ai-slim>=1.85.1,<2.0.0",
+    "pydantic-ai-slim>=1.102.0,<2.0.0",
     "typer>=0.16.1,<0.17",
     "typing-extensions>=4.12.2,<5.0.0",
 }
@@ -23,7 +23,7 @@ SERVER_RUNTIME_DEPENDENCIES = {
     "pydantic-ai-slim[anthropic,google,openai]>=1.85.1,<2.0.0",
     "pydantic-settings>=2.12.0,<3.0.0",
     "redis>=7.4.0,<8.0.0",
-    "shell-session-manager==2.3.0",
+    "shell-session-manager==2.3.1",
     "uvicorn[standard]==0.46.0",
 }
 
@@ -105,4 +105,5 @@ def test_local_sandbox_dockerfile_installs_stub_client_and_shellctl() -> None:
     assert "VIRTUAL_ENV" not in dockerfile
     assert "uv sync" not in dockerfile
     assert "mkdir -p /mnt/drive" in dockerfile
+    assert "chown dify:dify /home" in dockerfile
     assert '["shellctl", "serve", "--listen", "0.0.0.0:5004"]' in dockerfile
diff --git a/dify-agent/uv.lock b/dify-agent/uv.lock
index df40ea9fe1a..c0b7fa8897d 100644
--- a/dify-agent/uv.lock
+++ b/dify-agent/uv.lock
@@ -650,7 +650,7 @@ requires-dist = [
     { name = "pydantic-ai-slim", extras = ["anthropic", "google", "openai"], marker = "extra == 'server'", specifier = ">=1.85.1,<2.0.0" },
     { name = "pydantic-settings", marker = "extra == 'server'", specifier = ">=2.12.0,<3.0.0" },
     { name = "redis", marker = "extra == 'server'", specifier = ">=7.4.0,<8.0.0" },
-    { name = "shell-session-manager", marker = "extra == 'server'", specifier = "==2.3.0" },
+    { name = "shell-session-manager", marker = "extra == 'server'", specifier = "==2.4.0" },
     { name = "typer", specifier = ">=0.16.1,<0.17" },
     { name = "typing-extensions", specifier = ">=4.12.2,<5.0.0" },
     { name = "uvicorn", extras = ["standard"], marker = "extra == 'server'", specifier = "==0.46.0" },
@@ -3503,7 +3503,7 @@ wheels = [
 
 [[package]]
 name = "shell-session-manager"
-version = "2.3.0"
+version = "2.4.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "aiosqlite" },
@@ -3515,9 +3515,9 @@ dependencies = [
     { name = "typer" },
     { name = "uvicorn" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/ed/27/64b6086f54a700836c3fe6fc403e1a180a5c0285b528588663d0de951a6b/shell_session_manager-2.3.0.tar.gz", hash = "sha256:59598f4824623053405a6d0ba85c931e263597752aa6b91e67b0e523474dccf5", size = 51361, upload-time = "2026-06-28T18:32:57.249Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/46/c4/5ad9b6a3f8a4ca78acd4fa7a44f55b3f41ce7d539d648d9beee30789dd5b/shell_session_manager-2.4.0.tar.gz", hash = "sha256:cffeff439a149fa0626dd6fe940cd2d010005d97b429769e607a2b09ae64cb84", size = 64571, upload-time = "2026-07-03T19:11:18.931Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/fc/26/07e4815e630316f1840e36b686f91bd389940437e38b897a0231c487298f/shell_session_manager-2.3.0-py3-none-any.whl", hash = "sha256:7727cc328661d93dd40d6f42d2c6b46c95d49717698558e81d78b41a42501b4f", size = 48900, upload-time = "2026-06-28T18:32:55.937Z" },
+    { url = "https://files.pythonhosted.org/packages/be/8c/d2f6bb7d461810c22d10b775aeec3b4374eec7fb9fa503d00653d45cd8b3/shell_session_manager-2.4.0-py3-none-any.whl", hash = "sha256:5bb2bb43def96b354817ba6a45ab88fd236897a07c7f6c94b53f3f0cf30fabe6", size = 56513, upload-time = "2026-07-03T19:11:17.506Z" },
 ]
 
 [[package]]
diff --git a/e2e/.env.example b/e2e/.env.example
new file mode 100644
index 00000000000..530736a3546
--- /dev/null
+++ b/e2e/.env.example
@@ -0,0 +1,50 @@
+# Copy this file to e2e/.env.local for local runs.
+# Do not commit e2e/.env.local.
+
+E2E_BASE_URL=http://127.0.0.1:3000
+E2E_API_URL=http://127.0.0.1:5001
+
+E2E_STABLE_MODEL_PROVIDER=openai
+E2E_STABLE_MODEL_NAME=gpt-5-nano
+E2E_STABLE_MODEL_TYPE=llm
+
+E2E_AGENT_DECISION_MODEL_PROVIDER=openai
+E2E_AGENT_DECISION_MODEL_NAME=gpt-5.5
+E2E_AGENT_DECISION_MODEL_TYPE=llm
+
+E2E_MODEL_PROVIDER_CREDENTIALS_JSON='{"openai_api_key":"replace-with-real-key"}'
+
+# Optional: external runtime CI/local profile. Defaults prepare Agent V2 runtime
+# seed resources and then runs every @external-model/@external-tool scenario.
+# E2E_EXTERNAL_RUNTIME_SEED_SPECS=agent-v2:external-runtime
+# E2E_EXTERNAL_RUNTIME_TAGS=(@external-model or @external-tool) and not @feature-gated and not @skip and not @preview
+
+# Optional: start the standalone dify-agent backend for Agent V2 runtime scenarios.
+# This is required for scenarios tagged @agent-backend-runtime. It is separate
+# from @external-model because other external-model scenarios may not use dify-agent.
+# When enabled, the E2E runner also starts the shellctl local sandbox required by
+# dify-agent's dify.shell/config runtime layer.
+# E2E_START_AGENT_BACKEND=1
+# E2E_AGENT_BACKEND_PORT=5050
+# E2E_AGENT_BACKEND_URL=http://127.0.0.1:5050
+# E2E_SHELLCTL_PORT=5004
+# E2E_SHELLCTL_URL=http://127.0.0.1:5004
+# E2E_SHELLCTL_IMAGE=dify-agent-local-sandbox:e2e
+# E2E_SHELLCTL_AUTH_TOKEN=
+
+# Optional: Agent V2 seed installs the latest OpenAI, JSON Process, and Tavily
+# marketplace plugins by default. Override ids when testing a different provider
+# set, or pin exact marketplace package identifiers for reproducible CI runs.
+# E2E_MARKETPLACE_API_URL=https://marketplace.dify.ai
+# E2E_MARKETPLACE_PLUGIN_IDS=langgenius/openai,langgenius/json_process,langgenius/tavily
+# E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS=
+
+# Optional: required only for OAuth tool fixture scenarios.
+# E2E_OAUTH_TOOL_CREDENTIAL_ID=
+# E2E_OAUTH_TOOL_PROVIDER=
+# E2E_OAUTH_TOOL_NAME=
+
+# Optional: required only for broken model recovery scenarios.
+# E2E_BROKEN_MODEL_PROVIDER=
+# E2E_BROKEN_MODEL_NAME=broken-model
+# E2E_BROKEN_MODEL_TYPE=llm
diff --git a/e2e/.gitignore b/e2e/.gitignore
index 94517c8409d..a35ee3849b0 100644
--- a/e2e/.gitignore
+++ b/e2e/.gitignore
@@ -5,3 +5,6 @@ test-results/
 cucumber-report/
 .logs/
 .generated-test-materials/
+seed-report/
+.env.local
+.env.*.local
diff --git a/e2e/AGENTS.md b/e2e/AGENTS.md
index d80f6755bad..6f41c0e921e 100644
--- a/e2e/AGENTS.md
+++ b/e2e/AGENTS.md
@@ -53,6 +53,12 @@ pnpm -C e2e e2e:full
 # run a tagged subset
 pnpm -C e2e e2e -- --tags @smoke
 
+# prepare external runtime seed resources for opt-in external suites
+pnpm -C e2e e2e:external:prepare
+
+# run scenarios that call real external providers
+pnpm -C e2e e2e:external
+
 # headed browser
 pnpm -C e2e e2e:headed -- --tags @smoke
 
@@ -74,8 +80,8 @@ flowchart TD
   C --> D["Cucumber loads config, steps, and support modules"]
   D --> E["BeforeAll bootstraps shared auth state via /install"]
   E --> F{"Which command is running?"}
-  F -->|`pnpm -C e2e e2e`| G["Run config default tags: not @fresh and not @skip and not @preview"]
-  F -->|`pnpm -C e2e e2e:full*`| H["Override tags to not @skip and not @preview"]
+  F -->|`pnpm -C e2e e2e`| G["Run config default tags: not @fresh and not @skip and not @preview and not @external-model and not @external-tool"]
+  F -->|`pnpm -C e2e e2e:full*`| H["Override tags to not @skip and not @preview and not @external-model and not @external-tool"]
   G --> I["Per-scenario BrowserContext from shared browser"]
   H --> I
   I --> J["Failure artifacts written to cucumber-report/artifacts"]
@@ -105,7 +111,7 @@ Behavior depends on instance state:
 - uninitialized instance: completes install and stores authenticated state
 - initialized instance: signs in and reuses authenticated state
 
-Because of that, the `@fresh` install scenario only runs in the `pnpm -C e2e e2e:full*` flows. The default `pnpm -C e2e e2e*` flows exclude `@fresh` and `@preview` via Cucumber config tags so they can be re-run against an already initialized instance while keeping Builder Preview scenarios opt-in.
+Because of that, the `@fresh` install scenario only runs in the `pnpm -C e2e e2e:full*` flows. The default `pnpm -C e2e e2e*` flows exclude `@fresh`, `@preview`, `@external-model`, and `@external-tool` via Cucumber config tags so they can be re-run against an already initialized instance while keeping Builder Preview and real external runtime scenarios opt-in.
 
 Reset all persisted E2E state:
 
@@ -201,8 +207,14 @@ Feature: Create dataset
   - `@unauthenticated` — uses a clean BrowserContext with no cookies or storage
   - `@authenticated` — optional intent tag for readability or selective runs; it does not currently change hook behavior on its own
 - `@fresh` — only runs in `e2e:full` mode (requires uninitialized instance)
+- `@external-model` — scenario execution can call a real model provider. Use this only for runtime requests, not for scenarios that only require an active model fixture.
+- `@external-tool` — scenario execution can call a real third-party tool provider. Use this only for runtime tool execution, not for plugin installation, discovery, or local deterministic tools.
 - `@skip` — excluded from all runs
 
+External runtime commands are opt-in. `pnpm -C e2e e2e:external:prepare` reads `E2E_EXTERNAL_RUNTIME_SEED_SPECS`, defaulting to `agent-v2:external-runtime`, and runs the matching seed packs before the external suite. `pnpm -C e2e e2e:external` reads `E2E_EXTERNAL_RUNTIME_TAGS`, defaulting to `(@external-model or @external-tool) and not @feature-gated and not @skip and not @preview`.
+
+Some external runtime scenarios need feature-owned services in addition to a real model or tool provider. Do not overload `@external-model` or `@external-tool` to mean those services are available. For Agent v2, scenarios that require the standalone `dify-agent` run server use the feature tag `@agent-backend-runtime` plus the explicit step `the Agent v2 runtime backend is available`. Run them with `E2E_START_AGENT_BACKEND=1` to let E2E start `dify-agent` and the shellctl local sandbox required by its `dify.config`/`dify.shell` runtime layers, or set `E2E_AGENT_BACKEND_URL`/`AGENT_BACKEND_BASE_URL` when an existing server should be reused.
+
 Keep scenarios short and declarative. Each step should describe **what** the user does, not **how** the UI works.
 
 ### Step definition conventions
diff --git a/e2e/cucumber.config.ts b/e2e/cucumber.config.ts
index 3e6bd722556..49d47641bca 100644
--- a/e2e/cucumber.config.ts
+++ b/e2e/cucumber.config.ts
@@ -1,8 +1,10 @@
 import type { IConfiguration } from '@cucumber/cucumber'
+import './scripts/env-register'
 
 const hasCliTags = process.argv.some(arg => arg === '--tags' || arg.startsWith('--tags='))
+const defaultNonExternalTags = 'not @fresh and not @skip and not @preview and not @external-model and not @external-tool'
 const defaultTags = process.env.E2E_CUCUMBER_TAGS
-  || (hasCliTags ? undefined : 'not @fresh and not @skip and not @preview')
+  || (hasCliTags ? undefined : defaultNonExternalTags)
 
 const config = {
   format: [
diff --git a/e2e/features/agent-v2/AGENTS.md b/e2e/features/agent-v2/AGENTS.md
index 13d32647c11..6806fcac1d1 100644
--- a/e2e/features/agent-v2/AGENTS.md
+++ b/e2e/features/agent-v2/AGENTS.md
@@ -19,8 +19,8 @@ Use API setup for prerequisite state, then use Playwright only for user-observab
 Use tags in three layers:
 
 - Capability tags describe the product area: `@build`, `@files`, `@advanced-settings`, `@agent-edit`, `@publish`, `@access-point`, `@output-variables`, and similar tags.
-- Execution-scope tags describe how the scenario should be selected: `@core`, `@infra`, `@web-app-runtime`, `@service-api-runtime`, `@preview`, and `@feature-gated`.
-- Narrow fixture or sub-surface tags describe a specific dependency or slice: `@stable-model`, `@skill-fixture`, `@web-app-access`, `@workflow-reference`, `@files-limits`, and similar tags.
+- Execution-scope tags describe how the scenario should be selected: `@core`, `@infra`, `@web-app-runtime`, `@service-api-runtime`, `@agent-backend-runtime`, `@preview`, `@feature-gated`, `@external-model`, and `@external-tool`.
+- Narrow fixture or sub-surface tags describe a specific dependency or slice: `@stable-model`, `@agent-decision-model`, `@skill-fixture`, `@web-app-access`, `@workflow-reference`, `@files-limits`, and similar tags.
 
 - `@agent-v2` — required capability tag for all Agent v2 scenarios.
 - `@core` — stable non-runtime scenario expected to run in the regular Agent v2 suite when its explicit preconditions are met. Do not apply `@core` to Preview/Test Run, Web app chat runtime, or Backend service API chat runtime scenarios.
@@ -28,7 +28,7 @@ Use tags in three layers:
 - `@build` — Build mode and Build draft behavior.
 - `@build-unavailable-resources` — feature-gated Build chat recovery when the user requests unavailable Skills or Tools.
 - `@files` — Files section upload, display, and fixture behavior.
-- `@files-limits` — feature-gated file format, size, count, and in-progress upload limit behavior.
+- `@files-limits` — file limit behavior. Multiple-file drop is stable core coverage; format, size, count, and in-progress upload recovery remain feature-gated until their product contracts are stable.
 - `@knowledge` — Knowledge Retrieval configuration display, persistence, and reference cleanup.
 - `@advanced-settings` — Env Editor, Content Moderation, and related Advanced Settings behavior.
 - `@agent-create` — Agent Roster creation and initial Configure navigation.
@@ -36,17 +36,22 @@ Use tags in three layers:
 - `@publish` — publish and publish-bar state.
 - `@access-point` — Web app, Backend service API, and Workflow access surfaces.
 - `@stable-model` — active model fixture dependency. Apply this to every scenario that includes `the Agent Builder stable chat model is available` or otherwise requires an active model configured in the workspace.
+- `@agent-decision-model` — stronger active model fixture dependency for scenarios that specifically validate Agent autonomous planning or resource-selection behavior, such as generated-query Knowledge Retrieval. Do not use it for scenarios that only need a model to exist or answer a deterministic prompt.
 - `@tool-fixture` — preseeded Tool dependency such as `JSON Process / JSON Replace` or `Tavily / Tavily Search`.
 - `@skill-fixture` — checked-in or preseeded Skill dependency such as `e2e-summary-skill`.
 - `@knowledge-fixture` — preseeded dataset dependency such as `E2E Agent Knowledge Base`.
 - `@full-config-agent` — fixed `E2E New Agent Builder Full Config` Agent dependency.
 - `@tool-states-agent` — fixed `E2E New Agent Builder Tool States` Agent dependency.
+- `@oauth-tool-agent` — fixed `E2E Agent With OAuth Tool` Agent dependency for OAuth2 tool credential preservation.
 - `@file-tree-fixture` — fixed file-tree Agent drive/config-files dependency.
 - `@dual-retrieval-fixture` — fixed dual Knowledge Retrieval Agent dependency.
 - `@backend-api-access` — fixed or scenario-owned Backend service API access dependency.
 - `@published-web-app` — fixed or scenario-owned published Web app access dependency.
 - `@web-app-runtime` — published public Web app runtime behavior. Use it for scenarios that open the public Web app and assert real chat responses. Access Point URL, launch, customization, and settings surfaces remain `@access-point` behavior unless they send messages through the public Web app.
 - `@service-api-runtime` — Backend service API runtime behavior. Use it for scenarios that call the published service API and assert real chat responses. Endpoint display, copy, API key, and API reference surfaces remain `@access-point` behavior.
+- `@agent-backend-runtime` — standalone `dify-agent` run-server dependency. Apply it only when the scenario goes through Agent v2 runtime execution that requires `AGENT_BACKEND_BASE_URL`, such as Build chat/finalize, published Web app chat, or Backend service API chat. This is separate from `@external-model`; a scenario can call a real model without using `dify-agent`, and a future Agent v2 runtime scenario might exercise `dify-agent` without a paid model provider.
+- `@external-model` — package-wide external runtime tag. In Agent v2, apply it to Build chat generation/finalization, published Web app chat, Backend service API chat, and other Agent runtime requests that can invoke a real model provider. Do not use this for scenarios that only require a configured active model.
+- `@external-tool` — package-wide external runtime tag. In Agent v2, apply it only when the scenario executes a real external tool provider at runtime. Do not use this for plugin installation, tool discovery, tool configuration persistence, or local deterministic tools.
 - `@feature-gated` — product capability is optional. This tag alone does not skip execution; the scenario must include an explicit step that returns `skipped` with a blocked-precondition reason when the feature is unavailable.
 
 Use feature-level `@core` only when every scenario in the file is stable, non-runtime, and not feature-gated. If a feature file mixes stable scenarios with runtime, Preview, or feature-gated scenarios, put `@core` only on the stable scenarios. Keep runtime tags scenario-level so the regular core suite cannot inherit them accidentally.
@@ -107,6 +112,8 @@ Use `a basic configured Agent v2 test agent has been created via API` when a sce
 
 Use `a runnable Agent v2 test agent has been created via API` after `the Agent Builder stable chat model is available` when a scenario needs a real model-backed Agent. The step writes the preflight model into the Agent Soul model config through `features/agent-v2/support/agent-soul.ts` with deterministic E2E model settings; do not duplicate provider/model payload construction in individual steps.
 
+Use `a runnable Agent v2 test agent using the agent-decision model has been created via API` after `the Agent Builder agent-decision chat model is available` only when the scenario is asserting Agent autonomous decision quality rather than generic runtime availability. Keep this scoped to scenarios tagged `@agent-decision-model`.
+
 Use `the Agent v2 configuration should be saved automatically` after UI edits that rely on Configure autosave. It waits for the user-visible publish bar saved state; do not replace it with network-idle waits or internal store checks.
 
 API setup is acceptable for creating scenario-owned Agents, enabling Backend service API, writing composer drafts, seeding Build drafts, and preparing fixed state. The scenario must still assert user-visible behavior or a real persisted product contract through the public Console API. Do not assert only that a setup API call succeeded.
@@ -153,7 +160,11 @@ Agent Builder preflight is read-only. It checks long-lived seed resources and re
 
 Treat preseeded Agent Builder resources as environment contracts. Preflight can report that a stable model, dataset, Skill, Tool credential, fixed Agent, published Web app, or workflow reference is missing, inactive, not indexed, or drifted, but it must not repair that drift during a scenario. Seed scripts, CI bootstrap, or the documented environment maintenance flow own creating and keeping those resources valid.
 
-Use `the Agent Builder stable chat model is available` before scenarios that need a real Agent Soul model configuration. This includes true runtime scenarios, model-backed build-mode assertions, and Workflow Agent v2 node setup because the backend rejects Agent nodes without model config. Do not add the model preflight to pure navigation or identity checks unless the setup API itself requires model config. `E2E_STABLE_MODEL_PROVIDER`, `E2E_STABLE_MODEL_NAME`, and optional `E2E_STABLE_MODEL_TYPE` are selectors for a model already configured in the workspace; they are not provider credentials. The step defaults to `openai` / `gpt-5.4-mini` / `llm`, verifies the selected model is present and `active` through `/console/api/workspaces/current/models/model-types/{type}`, then stores it on `DifyWorld.agentBuilder.preflight.stableModel`.
+Use `the Agent v2 runtime backend is available` before scenarios tagged `@agent-backend-runtime`. The step checks the standalone `dify-agent` run server through `E2E_AGENT_BACKEND_URL`, `AGENT_BACKEND_BASE_URL`, or the default `http://127.0.0.1:5050` when `E2E_START_AGENT_BACKEND=1`. The E2E runner starts this service with `uv run --project dify-agent --extra server uvicorn dify_agent.server.app:app --host 127.0.0.1 --port 5050` only when `E2E_START_AGENT_BACKEND=1` is explicit. Because Agent App runtime always carries the `dify.config` layer and that layer depends on `dify.shell`, the same runner path also starts the `dify-agent/docker/local-sandbox` shellctl container on `E2E_SHELLCTL_PORT` (default `5004`) and injects `DIFY_AGENT_SHELLCTL_ENTRYPOINT` into the Agent backend. Missing or unreachable runtime backend or shellctl sandbox should be reported as a blocked precondition, not discovered later as a `/build-chat/finalize` 400 or Web app response timeout.
+
+Use `the Agent Builder stable chat model is available` before scenarios that need a real Agent Soul model configuration. This includes true runtime scenarios, model-backed build-mode assertions, and Workflow Agent v2 node setup because the backend rejects Agent nodes without model config. Do not add the model preflight to pure navigation or identity checks unless the setup API itself requires model config. `E2E_STABLE_MODEL_PROVIDER`, `E2E_STABLE_MODEL_NAME`, and optional `E2E_STABLE_MODEL_TYPE` are selectors for a model already configured in the workspace; they are not provider credentials. The step defaults to `openai` / `gpt-5-nano` / `llm`, verifies the selected model is present and `active` through `/console/api/workspaces/current/models/model-types/{type}`, then stores it on `DifyWorld.agentBuilder.preflight.stableModel`.
+
+Use `the Agent Builder agent-decision chat model is available` before scenarios that need a stronger model to exercise Agent autonomous planning, generated query selection, or tool/resource choice. `E2E_AGENT_DECISION_MODEL_PROVIDER`, `E2E_AGENT_DECISION_MODEL_NAME`, and optional `E2E_AGENT_DECISION_MODEL_TYPE` are selectors for a second active model fixture, defaulting to `openai` / `gpt-5.5` / `llm`. The step stores the model on `DifyWorld.agentBuilder.preflight.agentDecisionModel`. Do not use this fixture as a broad replacement for `@stable-model`; it is intentionally narrower and costlier.
 
 Keep `@stable-model` on Build draft apply scenarios that click `Apply`. The current product path calls `/build-chat/finalize` before applying the draft, and the backend returns `model is required` when the Agent Soul has no model config. Discard-only and pending-draft isolation scenarios can stay model-free when they do not finalize the Build draft.
 
@@ -163,8 +174,12 @@ Override the default selector only when a scenario or environment explicitly nee
 
 ```bash
 E2E_STABLE_MODEL_PROVIDER=openai
-E2E_STABLE_MODEL_NAME=gpt-5.4-mini
+E2E_STABLE_MODEL_NAME=gpt-5-nano
 E2E_STABLE_MODEL_TYPE=llm
+
+E2E_AGENT_DECISION_MODEL_PROVIDER=openai
+E2E_AGENT_DECISION_MODEL_NAME=gpt-5.5
+E2E_AGENT_DECISION_MODEL_TYPE=llm
 ```
 
 Dify may expose OpenAI as either `openai` or a plugin provider ID such as `langgenius/openai/openai`. The preflight accepts both forms for selection and stores the actual Console API provider ID for Agent Soul setup.
@@ -187,6 +202,8 @@ Use `the Agent Builder preseeded Agent "{agent}" includes the core fixture confi
 
 Use `the Agent Builder preseeded Agent "{agent}" includes the tool state fixture configuration` for the fixed Tool States Agent prerequisite. It composes the Summary Skill, JSON Replace tool, and Tavily Search tool preflights, then reads `/console/api/agent/{agent_id}/composer` to verify the Agent Soul includes JSON Replace, Tavily Search, and a Tavily credential reference. This proves the seed is configured to exercise tool status UI; keep actual invalid-credential errors in dependent user-visible configuration or runtime scenarios.
 
+Use `the Agent Builder preseeded Agent "{agent}" includes an OAuth2 tool credential` for the fixed OAuth Tool Agent prerequisite. It reads the Agent composer and verifies at least one Dify tool has `credential_type: oauth2` with a `credential_ref.id`. Scenarios that need to save or mutate this configuration should copy that tool config into a scenario-owned Agent instead of mutating the fixed preseeded Agent.
+
 Use `the Agent Builder preseeded Agent "{agent}" includes the dual retrieval fixture configuration` for the fixed Dual Retrieval Agent prerequisite. It composes the indexed knowledge-base preflight, then reads `/console/api/agent/{agent_id}/composer` to verify `agent_soul.knowledge.sets` includes both an Agent-decide generated query set and a custom user-query set using the fixed custom query.
 
 Use `the Agent Builder preseeded Agent "{agent}" includes the file tree fixture files` for file-tree display prerequisites. It verifies the Agent drive contains every file from `agentBuilderFileTreeFixtureFiles` through `/console/api/agent/{agent_id}/drive/files?prefix=files/`.
@@ -213,6 +230,6 @@ Order blocked steps by the real owner of the first unresolved condition. If a sc
 
 Use partial coverage only when current product behavior is intentionally narrower than the written requirement and the test still asserts a real user-visible behavior. Example: Files are currently flat in Agent config files, so the flat Files list can be asserted while tree display remains blocked until product support exists.
 
-File format, size, count, and in-progress upload limit cases are feature-gated until the product exposes stable Agent config file restrictions and user-visible recovery/error states. Do not convert `@files-limits` scenarios to passing tests by relying on default environment behavior; first align the product contract or seed configuration.
+Multiple-file drop is already covered as stable `@core @files-limits` behavior. File format, size, count, and in-progress upload limit cases remain feature-gated until the product exposes stable Agent config file restrictions and user-visible recovery/error states. Do not convert those gated `@files-limits` scenarios to passing tests by relying on default environment behavior; first align the product contract or seed configuration.
 
 Do not mark a scenario as complete if it only proves setup state and does not assert the user-visible behavior or persisted product contract required by the case.
diff --git a/e2e/features/agent-v2/access-point.feature b/e2e/features/agent-v2/access-point.feature
index 877776dc0cf..431133e50ac 100644
--- a/e2e/features/agent-v2/access-point.feature
+++ b/e2e/features/agent-v2/access-point.feature
@@ -162,10 +162,11 @@ Feature: Agent v2 Access Point
     When I refresh the current page
     Then Agent v2 Backend service API access should be in service
 
-  @service-api-runtime @backend-api-access @stable-model
+  @service-api-runtime @external-model @agent-backend-runtime @backend-api-access @stable-model
   Scenario: Published Agent v2 answers through Backend service API
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And Agent v2 Backend service API access has been enabled with a key via API
     When I open the Agent v2 configure page
@@ -174,10 +175,11 @@ Feature: Agent v2 Access Point
     When I send the Agent v2 Backend service API minimal request
     Then the Agent v2 Backend service API request should succeed with the normal E2E marker
 
-  @service-api-runtime @backend-api-access @stable-model
+  @service-api-runtime @external-model @agent-backend-runtime @backend-api-access @stable-model
   Scenario: Disabled Backend service API rejects requests and restored access succeeds
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And the Agent v2 draft has been published via API
     And Agent v2 Backend service API access has been enabled with a key via API
diff --git a/e2e/features/agent-v2/build-draft.feature b/e2e/features/agent-v2/build-draft.feature
index c36f9245280..9ef9cc3ab17 100644
--- a/e2e/features/agent-v2/build-draft.feature
+++ b/e2e/features/agent-v2/build-draft.feature
@@ -1,19 +1,17 @@
 @agent-v2 @authenticated @build
 Feature: Agent v2 build draft
-  @core @stable-model @tool-fixture @skill-fixture
+  @external-model @agent-backend-runtime @stable-model
   Scenario: Generating a Build draft leaves the normal Agent configuration unchanged
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
-    And the Agent Builder preseeded tool "JSON Process / JSON Replace" is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
-    And the e2e-summary-skill Skill is available to the Agent v2 test agent
     When I open the Agent v2 configure page
     And I generate an Agent v2 Build draft from the fixed instruction
     Then I should see the Agent v2 Build draft pending changes
     And I should see the Agent v2 Build mode confirmation state
     And the normal Agent v2 draft should still use the normal E2E prompt
     And the normal Agent v2 draft should not include the e2e-summary-skill Skill
-    And the normal Agent v2 draft should not include the Agent Builder JSON Replace tool
 
   @core
   Scenario: Discarding a Build draft keeps the original Agent configuration
@@ -58,10 +56,11 @@ Feature: Agent v2 build draft
     And the Agent v2 draft should not include the supported Build draft config
     And the Agent v2 Build draft should no longer be active
 
-  @core @stable-model
+  @external-model @agent-backend-runtime @stable-model
   Scenario: Applying a pending Build draft updates the normal Agent configuration
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And an Agent v2 Build draft uses the updated E2E prompt with the stable E2E model
     When I open the Agent v2 configure page
@@ -76,10 +75,11 @@ Feature: Agent v2 build draft
     Then I should see the updated E2E prompt in the Agent v2 prompt editor
     And the Agent v2 Build draft should no longer be active
 
-  @core @stable-model @skill-fixture
+  @external-model @agent-backend-runtime @stable-model @skill-fixture
   Scenario: Applying a Build draft updates supported configuration sections
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And an Agent v2 Build draft adds the supported E2E files, skills, and env
     When I open the Agent v2 configure page
@@ -103,10 +103,11 @@ Feature: Agent v2 build draft
     And I should see the supported E2E environment variable in Advanced Settings
     And the Agent v2 Build draft should no longer be active
 
-  @core @stable-model @skill-fixture
+  @external-model @agent-backend-runtime @stable-model @skill-fixture
   Scenario: Applying a Build draft with an existing Skill keeps a single Skill entry
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And an Agent v2 Build draft includes the existing e2e-summary-skill Skill
     When I open the Agent v2 configure page
diff --git a/e2e/features/agent-v2/files.feature b/e2e/features/agent-v2/files.feature
index 6b4fb44851a..700f90d2d30 100644
--- a/e2e/features/agent-v2/files.feature
+++ b/e2e/features/agent-v2/files.feature
@@ -52,13 +52,14 @@ Feature: Agent v2 files
     When I open the Agent v2 configure page
     Then Agent v2 oversized file rejection should be available
 
-  @files-limits @feature-gated
-  Scenario: Agent v2 single-batch file count limits are enforced
+  @core @files-limits
+  Scenario: Dropping multiple Agent v2 files at once is rejected
     Given I am signed in as the default E2E admin
-    And Agent v2 single-batch file count limits are available
     And a basic configured Agent v2 test agent has been created via API
     When I open the Agent v2 configure page
-    Then Agent v2 single-batch file count limits should be available
+    And I drop multiple Agent v2 files into the Files upload dialog
+    Then the Agent v2 Files upload dialog should reject the multiple-file drop
+    And I should not see the dropped Agent v2 files in the Files section
 
   @files-limits @feature-gated
   Scenario: Agent v2 total file count limits are enforced
diff --git a/e2e/features/agent-v2/knowledge.feature b/e2e/features/agent-v2/knowledge.feature
index 8e111418b9c..494f8f367db 100644
--- a/e2e/features/agent-v2/knowledge.feature
+++ b/e2e/features/agent-v2/knowledge.feature
@@ -24,12 +24,13 @@ Feature: Agent v2 Knowledge Retrieval
     When I refresh the current page
     Then I should see the Agent v2 Custom query Knowledge Retrieval settings
 
-  @service-api-runtime @stable-model @backend-api-access
+  @service-api-runtime @external-model @agent-backend-runtime @agent-decision-model @backend-api-access
   Scenario: Agent decide Knowledge Retrieval answers through Backend service API
     Given I am signed in as the default E2E admin
-    And the Agent Builder stable chat model is available
+    And the Agent Builder agent-decision chat model is available
+    And the Agent v2 runtime backend is available
     And the Agent Builder preseeded dataset "E2E Agent Knowledge Base" is indexed and ready
-    And a runnable Agent v2 test agent has been created via API
+    And a runnable Agent v2 test agent using the agent-decision model has been created via API
     And Agent v2 Backend service API access has been enabled with a key via API
     When I open the Agent v2 configure page
     And I add the Agent Builder knowledge base as an Agent decide Knowledge Retrieval
@@ -40,10 +41,11 @@ Feature: Agent v2 Knowledge Retrieval
     When I send the Agent v2 Backend service API knowledge request
     Then the Agent v2 Backend service API response should include the knowledge E2E marker
 
-  @service-api-runtime @stable-model @backend-api-access
+  @service-api-runtime @external-model @agent-backend-runtime @stable-model @backend-api-access
   Scenario: Custom query Knowledge Retrieval answers through Backend service API
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And the Agent Builder preseeded dataset "E2E Agent Knowledge Base" is indexed and ready
     And a runnable Agent v2 test agent has been created via API
     And Agent v2 Backend service API access has been enabled with a key via API
diff --git a/e2e/features/agent-v2/preflight.feature b/e2e/features/agent-v2/preflight.feature
index 54b19d813dc..1593fd758aa 100644
--- a/e2e/features/agent-v2/preflight.feature
+++ b/e2e/features/agent-v2/preflight.feature
@@ -14,6 +14,11 @@ Feature: Agent Builder preseeded environment
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
 
+  @agent-decision-model
+  Scenario: Agent-decision chat model is available
+    Given I am signed in as the default E2E admin
+    And the Agent Builder agent-decision chat model is available
+
   @broken-model
   Scenario: Broken chat model is available for recovery scenarios
     Given I am signed in as the default E2E admin
@@ -79,6 +84,11 @@ Feature: Agent Builder preseeded environment
     Given I am signed in as the default E2E admin
     And the Agent Builder preseeded Agent "E2E New Agent Builder Tool States" includes the tool state fixture configuration
 
+  @oauth-tool-agent
+  Scenario: OAuth2 tool Agent includes credential fixture configuration
+    Given I am signed in as the default E2E admin
+    And the Agent Builder preseeded Agent "E2E Agent With OAuth Tool" includes an OAuth2 tool credential
+
   @file-tree-fixture
   Scenario: File tree Agent includes fixture files
     Given I am signed in as the default E2E admin
diff --git a/e2e/features/agent-v2/publish.feature b/e2e/features/agent-v2/publish.feature
index 746b3c69965..d2a0bd1547a 100644
--- a/e2e/features/agent-v2/publish.feature
+++ b/e2e/features/agent-v2/publish.feature
@@ -58,10 +58,11 @@ Feature: Agent v2 publish
     And the normal Agent v2 draft should use the normal E2E prompt
     And the Agent v2 publish action should be available for unpublished changes
 
-  @web-app-runtime @published-web-app @stable-model
+  @web-app-runtime @external-model @agent-backend-runtime @published-web-app @stable-model
   Scenario: Published Agent v2 answers through Web app
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And Agent v2 Web app access has been enabled via API
     When I open the Agent v2 configure page
@@ -72,10 +73,11 @@ Feature: Agent v2 publish
     Then the Agent v2 Web app response should include the normal E2E marker
     When I close the Agent v2 Web app
 
-  @web-app-runtime @published-web-app @stable-model
+  @web-app-runtime @external-model @agent-backend-runtime @published-web-app @stable-model
   Scenario: Published Web app remains isolated from unpublished Agent v2 draft edits
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And Agent v2 Web app access has been enabled via API
     When I open the Agent v2 configure page
@@ -90,10 +92,11 @@ Feature: Agent v2 publish
     And the Agent v2 Web app response should not include the updated E2E marker
     When I close the Agent v2 Web app
 
-  @web-app-runtime @published-web-app @stable-model
+  @web-app-runtime @external-model @agent-backend-runtime @published-web-app @stable-model
   Scenario: Published Web app uses the latest Agent v2 published configuration
     Given I am signed in as the default E2E admin
     And the Agent Builder stable chat model is available
+    And the Agent v2 runtime backend is available
     And a runnable Agent v2 test agent has been created via API
     And Agent v2 Web app access has been enabled via API
     When I open the Agent v2 configure page
diff --git a/e2e/features/agent-v2/support/access-point.ts b/e2e/features/agent-v2/support/access-point.ts
index d8449a673c6..272364ba192 100644
--- a/e2e/features/agent-v2/support/access-point.ts
+++ b/e2e/features/agent-v2/support/access-point.ts
@@ -6,6 +6,7 @@ import type {
   ChatRequestPayloadWithUser,
   PostChatMessagesResponse,
 } from '@dify/contracts/api/service/types.gen'
+import type { APIResponse } from '@playwright/test'
 import { request } from '@playwright/test'
 import { createApiContext, expectApiResponseOK, setAppSiteEnabled } from '../../../support/api'
 import { getTestAgent } from './agent'
@@ -16,6 +17,84 @@ export type AgentServiceApiChatResult = {
   status: number
 }
 
+type ServiceApiSseEvent = {
+  data: unknown
+  event?: string
+}
+
+async function parseServiceApiChatResponse(response: APIResponse) {
+  const contentType = response.headers()['content-type'] ?? ''
+  const text = await response.text().catch(() => '')
+
+  if (contentType.includes('text/event-stream'))
+    return parseServiceApiSseText(text)
+
+  if (contentType.includes('application/json')) {
+    try {
+      return JSON.parse(text) as unknown
+    }
+    catch {
+      return { message: text }
+    }
+  }
+
+  try {
+    return JSON.parse(text) as unknown
+  }
+  catch {
+    return { message: text }
+  }
+}
+
+function parseServiceApiSseText(text: string) {
+  const events: ServiceApiSseEvent[] = []
+  const answers: string[] = []
+
+  for (const block of text.split(/\r?\n\r?\n/)) {
+    const lines = block.split(/\r?\n/)
+    const eventName = lines
+      .find(line => line.startsWith('event:'))
+      ?.slice('event:'.length)
+      .trim()
+    const dataText = lines
+      .filter(line => line.startsWith('data:'))
+      .map(line => line.slice('data:'.length).trimStart())
+      .join('\n')
+
+    if (!dataText)
+      continue
+
+    let data: unknown = dataText
+    try {
+      data = JSON.parse(dataText) as unknown
+    }
+    catch {
+      data = dataText
+    }
+
+    events.push({
+      data,
+      ...(eventName ? { event: eventName } : {}),
+    })
+
+    if (
+      data
+      && typeof data === 'object'
+      && !Array.isArray(data)
+      && 'answer' in data
+      && typeof data.answer === 'string'
+    ) {
+      answers.push(data.answer)
+    }
+  }
+
+  return {
+    answer: answers.join(''),
+    events,
+    raw: text,
+  }
+}
+
 export async function setAgentSiteAccessAndGetURL(
   agentId: string,
   enabled: boolean,
@@ -77,7 +156,7 @@ export async function sendAgentServiceApiChatMessage({
   const body = {
     inputs: {},
     query,
-    response_mode: 'blocking',
+    response_mode: 'streaming',
     user: 'e2e-agent-access-point',
   } satisfies ChatRequestPayloadWithUser
 
@@ -88,9 +167,7 @@ export async function sendAgentServiceApiChatMessage({
         Authorization: `Bearer ${apiKey}`,
       },
     })
-    const responseBody = await response.json().catch(async () => ({
-      message: await response.text().catch(() => ''),
-    }))
+    const responseBody = await parseServiceApiChatResponse(response)
 
     return {
       body: responseBody as PostChatMessagesResponse | unknown,
diff --git a/e2e/features/agent-v2/support/agent-builder-resources.ts b/e2e/features/agent-v2/support/agent-builder-resources.ts
index 0b6a70853a3..1ce25f836a5 100644
--- a/e2e/features/agent-v2/support/agent-builder-resources.ts
+++ b/e2e/features/agent-v2/support/agent-builder-resources.ts
@@ -5,10 +5,12 @@ export const agentBuilderPreseededResources = {
   tavilySearchTool: 'Tavily / Tavily Search',
   agentKnowledgeBase: 'E2E Agent Knowledge Base',
   indexingKnowledgeBase: 'E2E Agent Knowledge Base Indexing',
+  agentDecisionChatModel: 'E2E Agent Decision Chat Model',
   brokenModelProvider: 'E2E Broken Model Provider',
   brokenModel: 'e2e-broken-model',
   fullConfigAgent: 'E2E New Agent Builder Full Config',
   toolStatesAgent: 'E2E New Agent Builder Tool States',
+  oauthToolAgent: 'E2E Agent With OAuth Tool',
   fileTreeAgent: 'E2E Agent With File Tree',
   dualRetrievalAgent: 'E2E Agent With Dual Retrieval',
   publishedWebAppAgent: 'E2E Agent Published Web App',
@@ -23,7 +25,8 @@ export const agentBuilderFixedInputs = {
   missingSkillSearch: 'E2E_NOT_EXIST_SKILL',
   missingToolSearch: 'E2E_NOT_EXIST_TOOL',
   missingToolSearchWithSuffix: 'E2E_NOT_EXIST_TOOL_12345',
-  customKnowledgeQuery: 'Dify Agent E2E 测试暗号',
+  customKnowledgeQuery: 'Dify Agent E2E knowledge marker',
+  knowledgeRuntimeQuery: 'Use the connected knowledge source to find the Dify Agent E2E knowledge marker.',
   envPlainKey: 'E2E_AGENT_FLAG',
   envPlainValue: 'enabled',
   envModeKey: 'E2E_AGENT_MODE',
diff --git a/e2e/features/agent-v2/support/agent-soul.ts b/e2e/features/agent-v2/support/agent-soul.ts
index 1cdf59037e4..f3b3c7c6bb1 100644
--- a/e2e/features/agent-v2/support/agent-soul.ts
+++ b/e2e/features/agent-v2/support/agent-soul.ts
@@ -42,6 +42,11 @@ export const updatedAgentSoulConfig: AgentSoulConfig = {
   },
 }
 
+const stableAgentModelSettings = {
+  max_tokens: 4096,
+  temperature: 0,
+}
+
 const getAgentModelPluginId = (provider: string) => {
   const [organization, pluginName] = provider.split('/').filter(Boolean)
 
@@ -71,10 +76,7 @@ export function createAgentSoulConfigWithModel(
       plugin_id: getAgentModelPluginId(model.provider),
       model_provider: model.provider,
       model: model.name,
-      model_settings: {
-        temperature: 0,
-        max_tokens: 512,
-      },
+      model_settings: stableAgentModelSettings,
     },
   }
 }
@@ -103,3 +105,19 @@ export function createAgentSoulConfigWithKnowledgeDataset(
     },
   }
 }
+
+export function createAgentSoulConfigWithDifyTool(
+  agentSoul: AgentSoulConfig,
+  tool: NonNullable<NonNullable<AgentSoulConfig['tools']>['dify_tools']>[number],
+): AgentSoulConfig {
+  return {
+    ...agentSoul,
+    tools: {
+      ...agentSoul.tools,
+      dify_tools: [
+        ...(agentSoul.tools?.dify_tools ?? []),
+        tool,
+      ],
+    },
+  }
+}
diff --git a/e2e/features/agent-v2/support/preflight/agent-backend.ts b/e2e/features/agent-v2/support/preflight/agent-backend.ts
new file mode 100644
index 00000000000..4513a83a5b8
--- /dev/null
+++ b/e2e/features/agent-v2/support/preflight/agent-backend.ts
@@ -0,0 +1,120 @@
+import type { DifyWorld } from '../../../support/world'
+import { skipBlockedPrecondition } from './common'
+
+const isTruthyEnv = (value: string | undefined) => value === '1' || value === 'true'
+
+const getDefaultAgentBackendURL = () => {
+  const port = process.env.E2E_AGENT_BACKEND_PORT?.trim() || '5050'
+
+  return `http://127.0.0.1:${port}`
+}
+
+const getShellctlURL = () => {
+  const explicitE2EURL = process.env.E2E_SHELLCTL_URL?.trim()
+  if (explicitE2EURL)
+    return explicitE2EURL.replace(/\/$/, '')
+
+  const explicitAgentURL = process.env.DIFY_AGENT_SHELLCTL_ENTRYPOINT?.trim()
+  if (explicitAgentURL)
+    return explicitAgentURL.replace(/\/$/, '')
+
+  if (isTruthyEnv(process.env.E2E_START_AGENT_BACKEND)) {
+    const port = process.env.E2E_SHELLCTL_PORT?.trim() || '5004'
+    return `http://127.0.0.1:${port}`
+  }
+
+  return undefined
+}
+
+const getAgentBackendURL = () => {
+  const explicitE2EURL = process.env.E2E_AGENT_BACKEND_URL?.trim()
+  if (explicitE2EURL)
+    return explicitE2EURL.replace(/\/$/, '')
+
+  const explicitAPIURL = process.env.AGENT_BACKEND_BASE_URL?.trim()
+  if (explicitAPIURL)
+    return explicitAPIURL.replace(/\/$/, '')
+
+  if (isTruthyEnv(process.env.E2E_START_AGENT_BACKEND))
+    return getDefaultAgentBackendURL()
+
+  return undefined
+}
+
+const checkRuntimeOpenApi = async ({
+  remediation,
+  title,
+  url,
+  world,
+}: {
+  remediation: string
+  title: string
+  url: string
+  world: DifyWorld
+}) => {
+  const healthURL = `${url}/openapi.json`
+  try {
+    const response = await fetch(healthURL)
+    if (response.ok)
+      return undefined
+
+    return skipBlockedPrecondition(
+      world,
+      `${title} did not respond successfully at ${healthURL}: ${response.status} ${response.statusText}.`,
+      {
+        owner: 'e2e/runtime',
+        remediation,
+      },
+    )
+  }
+  catch (error) {
+    const message = error instanceof Error ? error.message : String(error)
+
+    return skipBlockedPrecondition(
+      world,
+      `${title} is unreachable at ${healthURL}: ${message}.`,
+      {
+        owner: 'e2e/runtime',
+        remediation,
+      },
+    )
+  }
+}
+
+export async function skipMissingAgentBackendRuntime(world: DifyWorld) {
+  const agentBackendURL = getAgentBackendURL()
+
+  if (!agentBackendURL) {
+    return skipBlockedPrecondition(
+      world,
+      'Agent v2 runtime backend is not configured. This scenario needs the standalone dify-agent run server, not just an active model provider.',
+      {
+        owner: 'e2e/runtime',
+        remediation: 'Run with E2E_START_AGENT_BACKEND=1 to let E2E start dify-agent, or set E2E_AGENT_BACKEND_URL/AGENT_BACKEND_BASE_URL to an existing dify-agent server.',
+      },
+    )
+  }
+
+  const agentBackendBlock = await checkRuntimeOpenApi({
+    remediation: 'Start a healthy dify-agent server and make sure AGENT_BACKEND_BASE_URL points to it before running Agent v2 runtime scenarios.',
+    title: 'Agent v2 runtime backend',
+    url: agentBackendURL,
+    world,
+  })
+  if (agentBackendBlock)
+    return agentBackendBlock
+
+  const shellctlURL = getShellctlURL()
+  if (shellctlURL) {
+    const shellctlBlock = await checkRuntimeOpenApi({
+      remediation: 'Start the shellctl local sandbox, or run with E2E_START_AGENT_BACKEND=1 so E2E starts it together with dify-agent.',
+      title: 'Agent v2 shellctl sandbox',
+      url: shellctlURL,
+      world,
+    })
+    if (shellctlBlock)
+      return shellctlBlock
+  }
+
+  return agentBackendURL
+}
diff --git a/e2e/features/agent-v2/support/preflight/agents.ts b/e2e/features/agent-v2/support/preflight/agents.ts
index 8fcb9519aa0..1d4ad2d7970 100644
--- a/e2e/features/agent-v2/support/preflight/agents.ts
+++ b/e2e/features/agent-v2/support/preflight/agents.ts
@@ -2,6 +2,7 @@ import type {
   AgentAppComposerResponse,
   AgentDriveListResponse,
   AgentDriveSkillListResponse,
+  AgentSoulConfig,
 } from '@dify/contracts/api/console/agent/types.gen'
 import type { DifyWorld } from '../../../support/world'
 import type { PreseededResource } from './common'
@@ -33,8 +34,11 @@ import {
   hasUnauthorizedToolCredentialState,
   skipMissingPreseededTool,
   splitToolDisplayName,
+  splitToolResourceId,
 } from './tools'
 
+type AgentSoulDifyToolConfig = NonNullable<NonNullable<AgentSoulConfig['tools']>['dify_tools']>[number]
+
 const hasKnowledgeDataset = (
   soul: Record<string, unknown>,
   dataset: PreseededResource,
@@ -84,6 +88,14 @@ const hasKnowledgeSet = (
   })
 }
 
+const getOAuth2ToolCredentialFixture = (toolItems: unknown[]) =>
+  toolItems.find((item) => {
+    const record = asRecord(item)
+
+    return record.credential_type === 'oauth2'
+      && Boolean(asString(asRecord(record.credential_ref).id))
+  }) as AgentSoulDifyToolConfig | undefined
+
 export async function skipMissingPreseededAgent(
   world: DifyWorld,
   resourceName: string,
@@ -223,7 +235,7 @@ export async function skipMissingPreseededFullConfigAgentCoreConfiguration(
     if (!hasNamedOrKeyedEntry(skills, agentBuilderPreseededResources.summarySkill))
       missing.push(agentBuilderPreseededResources.summarySkill)
 
-    const [providerName = '', toolName = ''] = jsonTool.id.split('/')
+    const { providerName, toolName } = splitToolResourceId(jsonTool.id)
     const parsedTool = splitToolDisplayName(agentBuilderPreseededResources.jsonReplaceTool)
     if (
       parsedTool.ok
@@ -297,7 +309,7 @@ export async function skipMissingPreseededToolStatesAgentConfiguration(
     if (!hasNamedOrKeyedEntry(skills, agentBuilderPreseededResources.summarySkill))
       missing.push(agentBuilderPreseededResources.summarySkill)
 
-    const [jsonProviderName = '', jsonToolName = ''] = jsonTool.id.split('/')
+    const { providerName: jsonProviderName, toolName: jsonToolName } = splitToolResourceId(jsonTool.id)
     const parsedJsonTool = splitToolDisplayName(agentBuilderPreseededResources.jsonReplaceTool)
     if (
       parsedJsonTool.ok
@@ -311,7 +323,7 @@ export async function skipMissingPreseededToolStatesAgentConfiguration(
       missing.push(agentBuilderPreseededResources.jsonReplaceTool)
     }
 
-    const [tavilyProviderName = '', tavilyToolName = ''] = tavilyTool.id.split('/')
+    const { providerName: tavilyProviderName, toolName: tavilyToolName } = splitToolResourceId(tavilyTool.id)
     const parsedTavilyTool = splitToolDisplayName(agentBuilderPreseededResources.tavilySearchTool)
     const tavilyEntry = parsedTavilyTool.ok
       ? findToolEntry(toolItems, {
@@ -343,6 +355,59 @@ export async function skipMissingPreseededToolStatesAgentConfiguration(
   }
 }
 
+export async function skipMissingPreseededOAuthToolAgentConfiguration(
+  world: DifyWorld,
+  agentName: string,
+): Promise<'skipped' | PreseededResource> {
+  const agent = await skipMissingPreseededAgent(world, agentName)
+  if (agent === 'skipped')
+    return agent
+
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.get(`/console/api/agent/${agent.id}/composer`)
+    await expectApiResponseOK(response, `Check preseeded Agent OAuth2 tool configuration ${agentName}`)
+    const body = (await response.json()) as AgentAppComposerResponse
+    const toolItems = asArray(asRecord(body.agent_soul?.tools).dify_tools)
+    const oauthTool = getOAuth2ToolCredentialFixture(toolItems)
+
+    if (!oauthTool) {
+      return skipBlockedPrecondition(
+        world,
+        `Preseeded Agent "${agentName}" is missing an OAuth2 tool with a credential reference.`,
+        {
+          owner: 'seed',
+          remediation: 'Seed an Agent v2 fixture that includes a built-in OAuth2 tool with credential_type oauth2 and credential_ref.id.',
+        },
+      )
+    }
+
+    return agent
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+export async function getPreseededOAuthToolConfig(agentId: string): Promise<AgentSoulDifyToolConfig> {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.get(`/console/api/agent/${agentId}/composer`)
+    await expectApiResponseOK(response, `Get preseeded Agent OAuth2 tool config ${agentId}`)
+    const body = (await response.json()) as AgentAppComposerResponse
+    const toolItems = asArray(asRecord(body.agent_soul?.tools).dify_tools)
+    const oauthTool = getOAuth2ToolCredentialFixture(toolItems)
+
+    if (!oauthTool)
+      throw new Error(`Preseeded Agent ${agentId} does not include an OAuth2 tool credential fixture.`)
+
+    return oauthTool
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
 export async function skipMissingPreseededDualRetrievalAgentConfiguration(
   world: DifyWorld,
   agentName: string,
diff --git a/e2e/features/agent-v2/support/preflight/datasets.ts b/e2e/features/agent-v2/support/preflight/datasets.ts
index b344ded7940..44841af3eb1 100644
--- a/e2e/features/agent-v2/support/preflight/datasets.ts
+++ b/e2e/features/agent-v2/support/preflight/datasets.ts
@@ -9,6 +9,7 @@ import type { PreseededResource } from './common'
 import { createApiContext, expectApiResponseOK } from '../../../../support/api'
 import {
   agentBuilderExpectedTokens,
+  agentBuilderFixedInputs,
   agentBuilderPreseededResources,
 } from '../agent-builder-resources'
 import {
@@ -84,10 +85,10 @@ const getDatasetDocuments = async (datasetId: string, resourceName: string) => {
   }
 }
 
-const datasetHasEnabledSegmentContainingToken = async (
+const datasetHasEnabledSegmentContainingTokens = async (
   datasetId: string,
   resourceName: string,
-  expectedToken: string,
+  expectedTokens: string[],
 ) => {
   const documents = await getDatasetDocuments(datasetId, resourceName)
   const ctx = await createApiContext()
@@ -95,7 +96,7 @@ const datasetHasEnabledSegmentContainingToken = async (
     for (const document of documents) {
       const query = buildQuery({
         enabled: 'true',
-        keyword: expectedToken,
+        keyword: agentBuilderExpectedTokens.knowledgeReply,
         limit: '20',
         page: '1',
       })
@@ -110,9 +111,9 @@ const datasetHasEnabledSegmentContainingToken = async (
       const matchingSegment = body.data.find(
         segment =>
           segment.enabled
-          && (
+          && expectedTokens.every(expectedToken =>
             segment.content.includes(expectedToken)
-            || segment.keywords?.some(keyword => keyword.includes(expectedToken))
+            || segment.keywords?.some(keyword => keyword.includes(expectedToken)),
           ),
       )
 
@@ -184,18 +185,23 @@ export async function skipMissingReadyPreseededDataset(
   }
 
   if (resourceName === agentBuilderPreseededResources.agentKnowledgeBase) {
-    const hasExpectedToken = await datasetHasEnabledSegmentContainingToken(
+    const requiredTokens = [
+      agentBuilderFixedInputs.customKnowledgeQuery,
+      agentBuilderFixedInputs.knowledgeRuntimeQuery,
+      agentBuilderExpectedTokens.knowledgeReply,
+    ]
+    const hasExpectedToken = await datasetHasEnabledSegmentContainingTokens(
       resource.id,
       resourceName,
-      agentBuilderExpectedTokens.knowledgeReply,
+      requiredTokens,
     )
 
     if (!hasExpectedToken) {
       return skipBlockedPrecondition(
         world,
-        `Preseeded dataset "${resourceName}" has no enabled segment containing "${agentBuilderExpectedTokens.knowledgeReply}".`,
+        `Preseeded dataset "${resourceName}" has no enabled segment containing "${requiredTokens.join('" and "')}".`,
         {
-          remediation: `Seed the dataset from the Agent Builder knowledge fixture and wait until an enabled segment contains "${agentBuilderExpectedTokens.knowledgeReply}".`,
+          remediation: `Seed the dataset from the Agent Builder knowledge fixture and wait until an enabled segment contains "${requiredTokens.join('" and "')}".`,
         },
       )
     }
diff --git a/e2e/features/agent-v2/support/preflight/models.ts b/e2e/features/agent-v2/support/preflight/models.ts
index 330c61c9d48..5c05c4187bf 100644
--- a/e2e/features/agent-v2/support/preflight/models.ts
+++ b/e2e/features/agent-v2/support/preflight/models.ts
@@ -7,13 +7,19 @@ import { skipBlockedPrecondition } from './common'
 const stableChatModelProviderEnv = 'E2E_STABLE_MODEL_PROVIDER'
 const stableChatModelNameEnv = 'E2E_STABLE_MODEL_NAME'
 const stableChatModelTypeEnv = 'E2E_STABLE_MODEL_TYPE'
+const agentDecisionChatModelProviderEnv = 'E2E_AGENT_DECISION_MODEL_PROVIDER'
+const agentDecisionChatModelNameEnv = 'E2E_AGENT_DECISION_MODEL_NAME'
+const agentDecisionChatModelTypeEnv = 'E2E_AGENT_DECISION_MODEL_TYPE'
 const brokenChatModelProviderEnv = 'E2E_BROKEN_MODEL_PROVIDER'
 const brokenChatModelNameEnv = 'E2E_BROKEN_MODEL_NAME'
 const brokenChatModelTypeEnv = 'E2E_BROKEN_MODEL_TYPE'
 const activeModelStatus = 'active'
 const defaultStableChatModelProvider = 'openai'
-const defaultStableChatModelName = 'gpt-5.4-mini'
+const defaultStableChatModelName = 'gpt-5-nano'
 const defaultStableChatModelType = 'llm'
+const defaultAgentDecisionChatModelProvider = 'openai'
+const defaultAgentDecisionChatModelName = 'gpt-5.5'
+const defaultAgentDecisionChatModelType = 'llm'
 const defaultBrokenChatModelName = agentBuilderPreseededResources.brokenModel
 
 const getProviderAlias = (provider: string) => provider.split('/').filter(Boolean).at(-1) ?? provider
@@ -48,6 +54,23 @@ export function readAgentBuilderStableChatModelConfig(): ModelPreflightConfig {
   }
 }
 
+export function readAgentBuilderAgentDecisionChatModelConfig(): ModelPreflightConfig {
+  const provider = process.env[agentDecisionChatModelProviderEnv]?.trim()
+    || defaultAgentDecisionChatModelProvider
+  const name = process.env[agentDecisionChatModelNameEnv]?.trim()
+    || defaultAgentDecisionChatModelName
+  const type = process.env[agentDecisionChatModelTypeEnv]?.trim()
+    || defaultAgentDecisionChatModelType
+
+  return {
+    ok: true,
+    provider,
+    resourceName: agentBuilderPreseededResources.agentDecisionChatModel,
+    type,
+    value: name,
+  }
+}
+
 export function readAgentBuilderBrokenChatModelConfig(): ModelPreflightConfig {
   const provider = process.env[brokenChatModelProviderEnv]?.trim()
   const name = process.env[brokenChatModelNameEnv]?.trim() || defaultBrokenChatModelName
@@ -129,6 +152,14 @@ export async function skipMissingAgentBuilderStableChatModel(
   })
 }
 
+export async function skipMissingAgentBuilderAgentDecisionChatModel(
+  world: DifyWorld,
+): Promise<'skipped' | NonNullable<DifyWorld['agentBuilder']['preflight']['stableModel']>> {
+  return skipMissingAgentBuilderModel(world, readAgentBuilderAgentDecisionChatModelConfig(), {
+    requireActive: true,
+  })
+}
+
 export async function skipMissingAgentBuilderBrokenChatModel(
   world: DifyWorld,
 ): Promise<'skipped' | NonNullable<DifyWorld['agentBuilder']['preflight']['stableModel']>> {
diff --git a/e2e/features/agent-v2/support/preflight/tools.ts b/e2e/features/agent-v2/support/preflight/tools.ts
index 9f20ee0e8e5..634a6318afa 100644
--- a/e2e/features/agent-v2/support/preflight/tools.ts
+++ b/e2e/features/agent-v2/support/preflight/tools.ts
@@ -36,6 +36,17 @@ export const splitToolDisplayName = (resourceName: string) => {
   }
 }
 
+export const splitToolResourceId = (resourceId: string) => {
+  const parts = resourceId.split('/').map(item => item.trim()).filter(Boolean)
+  const toolName = parts.at(-1)
+  const providerName = parts.slice(0, -1).join('/')
+
+  return {
+    providerName,
+    toolName: toolName ?? '',
+  }
+}
+
 export const findToolEntry = (
   items: unknown[],
   {
diff --git a/e2e/features/agent-v2/support/seed.ts b/e2e/features/agent-v2/support/seed.ts
new file mode 100644
index 00000000000..e3a3eb40bc1
--- /dev/null
+++ b/e2e/features/agent-v2/support/seed.ts
@@ -0,0 +1,1030 @@
+import type { AgentKnowledgeDatasetConfig, AgentSoulConfig } from '@dify/contracts/api/console/agent/types.gen'
+import type {
+  ConsoleSegmentListResponse,
+  DatasetListItemResponse,
+  DocumentStatusListResponse,
+  DocumentWithSegmentsListResponse,
+  KnowledgeConfig,
+} from '@dify/contracts/api/console/datasets/types.gen'
+import type {
+  ModelProviderListResponse,
+  ProviderWithModelsDataResponse,
+} from '@dify/contracts/api/console/workspaces/types.gen'
+import type { SeedContext, SeedResource, SeedTask } from '../../../support/seed'
+import type { UploadedConsoleFile } from './agent-drive'
+import { readFile } from 'node:fs/promises'
+import {
+  createApiContext,
+  createTestApp,
+  expectApiResponseOK,
+  publishWorkflowApp,
+  syncAgentV2WorkflowDraft,
+} from '../../../support/api'
+import { bootstrapMarketplacePlugins } from '../../../support/marketplace-plugins'
+import { sleep } from '../../../support/process'
+import {
+  blocked,
+  created,
+  skipped,
+  updated,
+  verified,
+} from '../../../support/seed'
+import {
+  createAgentApiKey,
+  setAgentApiAccess,
+  setAgentSiteAccessAndGetURL,
+} from './access-point'
+import {
+  createTestAgent,
+  publishAgent,
+  saveAgentComposerDraft,
+} from './agent'
+import {
+  agentBuilderExpectedTokens,
+  agentBuilderFixedInputs,
+  agentBuilderPreseededResources,
+} from './agent-builder-resources'
+import {
+  getAgentDriveSkills,
+  uploadAgentConfigFileToDraft,
+  uploadAgentConfigSkillToDraft,
+  uploadAgentDriveSkill,
+} from './agent-drive'
+import {
+  createAgentSoulConfigWithKnowledgeDataset,
+  createAgentSoulConfigWithModel,
+  normalAgentSoulConfig,
+} from './agent-soul'
+import {
+  buildQuery,
+  findConsoleResourceByName,
+  isRecord,
+  matchesNameOrLabel,
+} from './preflight/common'
+import { splitToolDisplayName } from './preflight/tools'
+import {
+  agentBuilderTestMaterials,
+  getAgentBuilderTestMaterialPath,
+} from './test-materials'
+
+type StableModel = {
+  name: string
+  provider: string
+  type: string
+}
+
+type ToolResource = SeedResource & {
+  providerName: string
+  toolName: string
+}
+
+const modelCredentialEnv = 'E2E_MODEL_PROVIDER_CREDENTIALS_JSON'
+const marketplacePluginIdsEnv = 'E2E_MARKETPLACE_PLUGIN_IDS'
+const marketplacePluginUniqueIdentifiersEnv = 'E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS'
+const oauthToolCredentialIdEnv = 'E2E_OAUTH_TOOL_CREDENTIAL_ID'
+const oauthToolProviderEnv = 'E2E_OAUTH_TOOL_PROVIDER'
+const oauthToolNameEnv = 'E2E_OAUTH_TOOL_NAME'
+const activeModelStatus = 'active'
+const stableModelCredentialName = 'E2E Stable Model'
+const agentV2MarketplacePluginIds = [
+  'langgenius/openai',
+  'langgenius/json_process',
+  'langgenius/tavily',
+]
+
+const getProviderAlias = (provider: string) => provider.split('/').filter(Boolean).at(-1) ?? provider
+
+const matchesProvider = (actual: string, expected: string) =>
+  actual === expected || getProviderAlias(actual) === getProviderAlias(expected)
+
+const matchesProviderLabel = (provider: { label?: { en_US?: string | null, zh_Hans?: string | null } | null, provider: string }, expected: string) =>
+  matchesProvider(provider.provider, expected)
+  || provider.label?.en_US === expected
+  || provider.label?.zh_Hans === expected
+
+const stableModelConfig = (): StableModel => ({
+  name: process.env.E2E_STABLE_MODEL_NAME?.trim() || 'gpt-5-nano',
+  provider: process.env.E2E_STABLE_MODEL_PROVIDER?.trim() || 'openai',
+  type: process.env.E2E_STABLE_MODEL_TYPE?.trim() || 'llm',
+})
+
+const agentDecisionModelConfig = (): StableModel => ({
+  name: process.env.E2E_AGENT_DECISION_MODEL_NAME?.trim() || 'gpt-5.5',
+  provider: process.env.E2E_AGENT_DECISION_MODEL_PROVIDER?.trim() || 'openai',
+  type: process.env.E2E_AGENT_DECISION_MODEL_TYPE?.trim() || 'llm',
+})
+
+const parseJsonEnv = (envName: string) => {
+  const raw = process.env[envName]?.trim()
+  if (!raw)
+    return { ok: false as const, reason: `${envName} is required.` }
+
+  try {
+    const value = JSON.parse(raw) as unknown
+    if (!isRecord(value))
+      return { ok: false as const, reason: `${envName} must be a JSON object.` }
+
+    return { ok: true as const, value }
+  }
+  catch (error) {
+    return {
+      ok: false as const,
+      reason: `${envName} must be valid JSON: ${error instanceof Error ? error.message : String(error)}`,
+    }
+  }
+}
+
+const findChatModel = async (config: StableModel, title: string) => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.get(`/console/api/workspaces/current/models/model-types/${config.type}`)
+    await expectApiResponseOK(response, `Check ${title}`)
+    const body = (await response.json()) as ProviderWithModelsDataResponse
+    const provider = body.data.find(item => matchesProvider(item.provider, config.provider))
+    const model = provider?.models.find(
+      item =>
+        item.model === config.name
+        || item.label?.en_US === config.name
+        || item.label?.zh_Hans === config.name,
+    )
+
+    if (!provider || !model)
+      return undefined
+
+    return {
+      name: model.model,
+      provider: provider.provider,
+      status: model.status,
+      type: config.type,
+    }
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const resolveProvider = async (config: StableModel) => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.get(`/console/api/workspaces/current/model-providers?${buildQuery({ model_type: config.type })}`)
+    await expectApiResponseOK(response, `Resolve model provider ${config.provider}`)
+    const body = (await response.json()) as ModelProviderListResponse
+    const provider = body.data.find(item => matchesProviderLabel(item, config.provider))
+
+    return {
+      availableProviders: body.data.map(provider => provider.provider),
+      credential: provider?.custom_configuration.available_credentials?.find(
+        credential => credential.credential_name === stableModelCredentialName,
+      ),
+      provider: provider?.provider,
+    }
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const selectCustomProviderCredential = async (provider: string, credentialId?: string) => {
+  const ctx = await createApiContext()
+  try {
+    if (credentialId) {
+      const switchResponse = await ctx.post(
+        `/console/api/workspaces/current/model-providers/${provider}/credentials/switch`,
+        {
+          data: { credential_id: credentialId },
+        },
+      )
+      await expectApiResponseOK(switchResponse, `Switch model provider credential for ${provider}`)
+    }
+
+    const preferredResponse = await ctx.post(
+      `/console/api/workspaces/current/model-providers/${provider}/preferred-provider-type`,
+      {
+        data: { preferred_provider_type: 'custom' },
+      },
+    )
+    await expectApiResponseOK(preferredResponse, `Select custom provider credential for ${provider}`)
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const upsertStableProviderCredential = async (
+  provider: string,
+  credentials: Record<string, unknown>,
+  credentialId?: string,
+) => {
+  const ctx = await createApiContext()
+  try {
+    if (credentialId) {
+      const updateResponse = await ctx.put(
+        `/console/api/workspaces/current/model-providers/${provider}/credentials`,
+        {
+          data: {
+            credential_id: credentialId,
+            credentials,
+            name: stableModelCredentialName,
+          },
+        },
+      )
+      await expectApiResponseOK(updateResponse, `Update model provider credential for ${provider}`)
+      return
+    }
+
+    const createResponse = await ctx.post(
+      `/console/api/workspaces/current/model-providers/${provider}/credentials`,
+      {
+        data: {
+          credentials,
+          name: stableModelCredentialName,
+        },
+      },
+    )
+    await expectApiResponseOK(createResponse, `Create model provider credential for ${provider}`)
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const seedChatModel = async (context: SeedContext, {
+  config,
+  title,
+}: {
+  config: StableModel
+  title: string
+}) => {
+  const existing = await findChatModel(config, title)
+  const resource = {
+    id: `${existing?.provider ?? config.provider}/${existing?.name ?? config.name}`,
+    kind: 'model',
+    name: title,
+  }
+
+  if (existing?.status === activeModelStatus)
+    return verified(title, resource)
+
+  if (context.dryRun)
+    return skipped(title, `Would configure ${config.provider}/${config.name} using ${modelCredentialEnv}.`)
+
+  const credentials = parseJsonEnv(modelCredentialEnv)
+  if (!credentials.ok)
+    return blocked(title, `${config.provider}/${config.name} is not active; ${credentials.reason}`)
+
+  const { availableProviders, credential, provider } = await resolveProvider(config)
+  if (!provider) {
+    const available = availableProviders.length > 0
+      ? availableProviders.join(', ')
+      : 'none'
+    return blocked(
+      title,
+      `Provider ${config.provider} was not found in available model providers for ${config.type}. Available providers: ${available}.`,
+    )
+  }
+
+  try {
+    await upsertStableProviderCredential(provider, credentials.value, credential?.credential_id)
+    await selectCustomProviderCredential(provider, credential?.credential_id)
+  }
+  catch (error) {
+    const message = error instanceof Error ? error.message : String(error)
+    if (!message.includes(`Credential with name '${stableModelCredentialName}' already exists.`))
+      return blocked(title, message)
+
+    const refreshed = await resolveProvider(config)
+    if (!refreshed.provider || !refreshed.credential) {
+      return blocked(
+        title,
+        `Credential ${stableModelCredentialName} already exists for ${provider}, but the seed could not resolve its credential id.`,
+      )
+    }
+
+    try {
+      await upsertStableProviderCredential(
+        refreshed.provider,
+        credentials.value,
+        refreshed.credential.credential_id,
+      )
+      await selectCustomProviderCredential(refreshed.provider, refreshed.credential.credential_id)
+    }
+    catch (retryError) {
+      return blocked(title, retryError instanceof Error ? retryError.message : String(retryError))
+    }
+  }
+
+  const seeded = await findChatModel(config, title)
+  if (seeded?.status !== activeModelStatus) {
+    return blocked(
+      title,
+      `${config.provider}/${config.name} is ${seeded?.status ?? 'missing'} after credential setup.`,
+    )
+  }
+
+  return updated(title, {
+    id: `${seeded.provider}/${seeded.name}`,
+    kind: 'model',
+    name: title,
+  })
+}
+
+const seedStableModel = async (context: SeedContext) => seedChatModel(context, {
+  config: stableModelConfig(),
+  title: agentBuilderPreseededResources.stableChatModel,
+})
+
+const seedAgentDecisionModel = async (context: SeedContext) => seedChatModel(context, {
+  config: agentDecisionModelConfig(),
+  title: agentBuilderPreseededResources.agentDecisionChatModel,
+})
+
+type BuiltinToolProvider = {
+  label?: { en_US?: string, zh_Hans?: string }
+  name: string
+  tools: Array<{
+    label?: { en_US?: string, zh_Hans?: string }
+    name: string
+  }>
+}
+
+const findBuiltinTool = async (displayName: string) => {
+  const parsed = splitToolDisplayName(displayName)
+  if (!parsed.ok)
+    return { ok: false as const, reason: parsed.reason }
+
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.get('/console/api/workspaces/current/tools/builtin')
+    await expectApiResponseOK(response, `Check built-in tool ${displayName}`)
+    const providers = (await response.json()) as BuiltinToolProvider[]
+    const provider = providers.find(item =>
+      matchesNameOrLabel(parsed.providerName, item.name, item.label))
+    const tool = provider?.tools.find(item =>
+      matchesNameOrLabel(parsed.toolName, item.name, item.label))
+
+    if (!provider || !tool)
+      return { ok: false as const, reason: `Built-in tool "${displayName}" was not found.` }
+
+    return {
+      ok: true as const,
+      resource: {
+        id: `${provider.name}/${tool.name}`,
+        kind: 'tool',
+        name: displayName,
+        providerName: provider.name,
+        toolName: tool.name,
+      } satisfies ToolResource,
+    }
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const seedTool = (displayName: string): SeedTask => ({
+  id: `tool:${displayName}`,
+  title: displayName,
+  async run() {
+    const result = await findBuiltinTool(displayName)
+    if (!result.ok)
+      return blocked(displayName, result.reason)
+
+    return verified(displayName, result.resource)
+  },
+})
+
+const uploadConsoleFile = async (fileName: string, filePath: string): Promise<UploadedConsoleFile> => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/files/upload', {
+      multipart: {
+        file: {
+          buffer: await readFile(filePath),
+          mimeType: 'text/plain',
+          name: fileName,
+        },
+      },
+    })
+    await expectApiResponseOK(response, `Upload seed file ${fileName}`)
+    return (await response.json()) as UploadedConsoleFile
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const findDataset = (name: string) => {
+  const query = buildQuery({ keyword: name, limit: '20', page: '1' })
+  return findConsoleResourceByName<DatasetListItemResponse>({
+    action: `Find seed dataset ${name}`,
+    path: `/console/api/datasets?${query}`,
+    resourceName: name,
+  })
+}
+
+const getDatasetDocuments = async (datasetId: string) => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.get(`/console/api/datasets/${datasetId}/documents?${buildQuery({ limit: '100', page: '1' })}`)
+    await expectApiResponseOK(response, `List dataset documents ${datasetId}`)
+    const body = (await response.json()) as DocumentWithSegmentsListResponse
+    return body.data
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const requiredKnowledgeSegmentTokens = [
+  agentBuilderFixedInputs.customKnowledgeQuery,
+  agentBuilderFixedInputs.knowledgeRuntimeQuery,
+  agentBuilderExpectedTokens.knowledgeReply,
+]
+
+const datasetHasKnowledgeSegment = async (datasetId: string) => {
+  const documents = await getDatasetDocuments(datasetId)
+  const ctx = await createApiContext()
+  try {
+    for (const document of documents) {
+      const response = await ctx.get(
+        `/console/api/datasets/${datasetId}/documents/${document.id}/segments?${buildQuery({
+          enabled: 'true',
+          keyword: agentBuilderExpectedTokens.knowledgeReply,
+          limit: '20',
+          page: '1',
+        })}`,
+      )
+      await expectApiResponseOK(response, `Check dataset knowledge segment ${agentBuilderExpectedTokens.knowledgeReply}`)
+      const body = (await response.json()) as ConsoleSegmentListResponse
+      if (body.data.some(segment =>
+        segment.enabled
+        && requiredKnowledgeSegmentTokens.every(token => segment.content.includes(token)),
+      )) {
+        return true
+      }
+    }
+
+    return false
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const waitForDatasetCompleted = async (datasetId: string) => {
+  const deadline = Date.now() + 180_000
+  let status = 'missing'
+
+  while (Date.now() < deadline) {
+    const ctx = await createApiContext()
+    try {
+      const response = await ctx.get(`/console/api/datasets/${datasetId}/indexing-status`)
+      await expectApiResponseOK(response, `Check dataset indexing ${datasetId}`)
+      const body = (await response.json()) as DocumentStatusListResponse
+      status = body.data.length < 1
+        ? 'missing'
+        : body.data.every(item => item.indexing_status === 'completed')
+          ? 'completed'
+          : body.data.map(item => item.indexing_status ?? 'missing').join(',')
+
+      if (status === 'completed')
+        return { ok: true as const }
+    }
+    finally {
+      await ctx.dispose()
+    }
+
+    await sleep(1_000)
+  }
+
+  return { ok: false as const, status }
+}
+
+const addKnowledgeDocument = async (datasetId: string) => {
+  const uploadedFile = await uploadConsoleFile(
+    agentBuilderTestMaterials.knowledgeSource,
+    getAgentBuilderTestMaterialPath('knowledgeSource'),
+  )
+  const body = {
+    data_source: {
+      info_list: {
+        data_source_type: 'upload_file',
+        file_info_list: {
+          file_ids: [uploadedFile.id],
+        },
+      },
+    },
+    doc_form: 'text_model',
+    doc_language: 'English',
+    indexing_technique: 'economy',
+    process_rule: {
+      mode: 'automatic',
+    },
+    retrieval_model: {
+      reranking_enable: false,
+      score_threshold_enabled: false,
+      search_method: 'keyword_search',
+      top_k: 4,
+    },
+  } satisfies KnowledgeConfig
+
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post(`/console/api/datasets/${datasetId}/documents`, { data: body })
+    await expectApiResponseOK(response, `Seed knowledge document ${datasetId}`)
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const createDataset = async (name: string) => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/datasets', {
+      data: {
+        indexing_technique: 'economy',
+        name,
+        permission: 'only_me',
+        provider: 'vendor',
+      },
+    })
+    await expectApiResponseOK(response, `Create dataset ${name}`)
+    return (await response.json()) as DatasetListItemResponse
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const seedReadyKnowledge = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.agentKnowledgeBase
+  let dataset = await findDataset(title)
+
+  if (context.dryRun) {
+    return dataset
+      ? verified(title, { id: dataset.id, kind: 'dataset', name: title })
+      : skipped(title, `Would create dataset "${title}".`)
+  }
+
+  const wasCreated = !dataset
+  dataset ??= await createDataset(title)
+
+  const hasKnowledgeSegment = await datasetHasKnowledgeSegment(dataset.id)
+  if (!hasKnowledgeSegment)
+    await addKnowledgeDocument(dataset.id)
+
+  const indexing = await waitForDatasetCompleted(dataset.id)
+  if (!indexing.ok) {
+    return blocked(
+      title,
+      `Dataset "${title}" indexing did not complete in seed timeout; last status: ${indexing.status}.`,
+    )
+  }
+
+  return datasetHasKnowledgeSegment(dataset.id)
+    .then((ready) => {
+      if (!ready) {
+        return blocked(
+          title,
+          `Dataset "${title}" does not expose an indexed segment containing ${requiredKnowledgeSegmentTokens.join(' and ')} after indexing.`,
+        )
+      }
+
+      const resource = { id: dataset.id, kind: 'dataset', name: title }
+      return wasCreated ? created(title, resource) : verified(title, resource)
+    })
+}
+
+const ensureAgent = async (name: string) => {
+  const query = buildQuery({ limit: '20', name, page: '1' })
+  const existing = await findConsoleResourceByName({
+    action: `Find seed Agent ${name}`,
+    path: `/console/api/agent?${query}`,
+    resourceName: name,
+  })
+
+  if (existing)
+    return { agent: existing, created: false }
+
+  const agent = await createTestAgent({
+    description: 'Created by Dify E2E seed.',
+    name,
+    role: 'E2E seeded assistant',
+  })
+
+  return { agent, created: true }
+}
+
+const getStableModelResource = (context: SeedContext): StableModel | undefined => {
+  const resource = context.resources.get('stable-model')
+  if (!resource?.id)
+    return undefined
+
+  const separatorIndex = resource.id.lastIndexOf('/')
+  if (separatorIndex === -1)
+    return undefined
+
+  return {
+    name: resource.id.slice(separatorIndex + 1),
+    provider: resource.id.slice(0, separatorIndex),
+    type: stableModelConfig().type,
+  }
+}
+
+const getToolResource = (context: SeedContext, displayName: string) =>
+  context.resources.get(`tool:${displayName}`) as ToolResource | undefined
+
+const toolConfig = (tool: ToolResource) => ({
+  credential_type: 'unauthorized' as const,
+  enabled: true,
+  provider_id: tool.providerName,
+  provider_type: 'builtin',
+  runtime_parameters: {},
+  tool_name: tool.toolName,
+})
+
+const saveSeededAgentComposer = async ({
+  agentId,
+  config,
+  shouldPublish = false,
+}: {
+  agentId: string
+  config: AgentSoulConfig
+  shouldPublish?: boolean
+}) => {
+  await saveAgentComposerDraft(agentId, config)
+  if (shouldPublish)
+    await publishAgent(agentId, 'E2E seed')
+}
+
+const ensureDriveSkill = async (agentId: string) => {
+  const skills = await getAgentDriveSkills(agentId)
+  if (skills.some(skill => skill.name === agentBuilderPreseededResources.summarySkill))
+    return
+
+  await uploadAgentDriveSkill({
+    agentId,
+    fileName: agentBuilderTestMaterials.summarySkill,
+    filePath: getAgentBuilderTestMaterialPath('summarySkill'),
+  })
+}
+
+const seedFullConfigAgent = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.fullConfigAgent
+  const model = getStableModelResource(context)
+  const jsonTool = getToolResource(context, agentBuilderPreseededResources.jsonReplaceTool)
+  const dataset = context.resources.get('ready-knowledge')
+  if (!model)
+    return blocked(title, `${agentBuilderPreseededResources.stableChatModel} is not ready.`)
+  if (!jsonTool)
+    return blocked(title, `${agentBuilderPreseededResources.jsonReplaceTool} is not ready.`)
+  if (!dataset?.id)
+    return blocked(title, `${agentBuilderPreseededResources.agentKnowledgeBase} is not ready.`)
+
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}".`)
+
+  const { agent, created: wasCreated } = await ensureAgent(title)
+  const agentId = agent.id
+  const smallFile = await uploadAgentConfigFileToDraft({
+    agentId,
+    fileName: agentBuilderTestMaterials.smallFile,
+    filePath: getAgentBuilderTestMaterialPath('smallFile'),
+  })
+  const specialFile = await uploadAgentConfigFileToDraft({
+    agentId,
+    fileName: agentBuilderTestMaterials.specialFilename,
+    filePath: getAgentBuilderTestMaterialPath('specialFilename'),
+  })
+  const summarySkill = await uploadAgentConfigSkillToDraft({
+    agentId,
+    fileName: agentBuilderTestMaterials.summarySkill,
+    filePath: getAgentBuilderTestMaterialPath('summarySkill'),
+  })
+  await ensureDriveSkill(agentId)
+
+  await saveSeededAgentComposer({
+    agentId,
+    config: createAgentSoulConfigWithKnowledgeDataset(
+      createAgentSoulConfigWithModel({
+        ...normalAgentSoulConfig,
+        config_files: [smallFile, specialFile],
+        config_skills: [summarySkill],
+        tools: {
+          dify_tools: [toolConfig(jsonTool)],
+        },
+      }, model),
+      {
+        id: dataset.id,
+        name: dataset.name,
+      } satisfies AgentKnowledgeDatasetConfig,
+    ),
+  })
+
+  const resource = { id: agentId, kind: 'agent', name: title }
+  return wasCreated ? created(title, resource) : updated(title, resource)
+}
+
+const seedToolStatesAgent = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.toolStatesAgent
+  const jsonTool = getToolResource(context, agentBuilderPreseededResources.jsonReplaceTool)
+  const tavilyTool = getToolResource(context, agentBuilderPreseededResources.tavilySearchTool)
+  if (!jsonTool)
+    return blocked(title, `${agentBuilderPreseededResources.jsonReplaceTool} is not ready.`)
+  if (!tavilyTool)
+    return blocked(title, `${agentBuilderPreseededResources.tavilySearchTool} is not ready.`)
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}".`)
+
+  const { agent, created: wasCreated } = await ensureAgent(title)
+  const summarySkill = await uploadAgentConfigSkillToDraft({
+    agentId: agent.id,
+    fileName: agentBuilderTestMaterials.summarySkill,
+    filePath: getAgentBuilderTestMaterialPath('summarySkill'),
+  })
+  await ensureDriveSkill(agent.id)
+  await saveSeededAgentComposer({
+    agentId: agent.id,
+    config: {
+      ...normalAgentSoulConfig,
+      config_skills: [summarySkill],
+      tools: {
+        dify_tools: [toolConfig(jsonTool), toolConfig(tavilyTool)],
+      },
+    },
+  })
+
+  const resource = { id: agent.id, kind: 'agent', name: title }
+  return wasCreated ? created(title, resource) : updated(title, resource)
+}
+
+const seedDualRetrievalAgent = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.dualRetrievalAgent
+  const dataset = context.resources.get('ready-knowledge')
+  if (!dataset?.id)
+    return blocked(title, `${agentBuilderPreseededResources.agentKnowledgeBase} is not ready.`)
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}".`)
+
+  const { agent, created: wasCreated } = await ensureAgent(title)
+  const datasetConfig = { id: dataset.id, name: dataset.name } satisfies AgentKnowledgeDatasetConfig
+  await saveSeededAgentComposer({
+    agentId: agent.id,
+    config: {
+      ...normalAgentSoulConfig,
+      knowledge: {
+        sets: [
+          {
+            datasets: [datasetConfig],
+            id: 'e2e-dual-retrieval-agent-decide',
+            name: 'Retrieval 1',
+            query: { mode: 'generated_query' },
+            retrieval: { mode: 'multiple', top_k: 4 },
+          },
+          {
+            datasets: [datasetConfig],
+            id: 'e2e-dual-retrieval-custom-query',
+            name: 'Retrieval 2',
+            query: {
+              mode: 'user_query',
+              value: agentBuilderFixedInputs.customKnowledgeQuery,
+            },
+            retrieval: { mode: 'multiple', top_k: 4 },
+          },
+        ],
+      },
+    },
+  })
+
+  const resource = { id: agent.id, kind: 'agent', name: title }
+  return wasCreated ? created(title, resource) : updated(title, resource)
+}
+
+const seedPublishedWebAppAgent = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.publishedWebAppAgent
+  const model = getStableModelResource(context)
+  if (!model)
+    return blocked(title, `${agentBuilderPreseededResources.stableChatModel} is not ready.`)
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}".`)
+
+  const { agent, created: wasCreated } = await ensureAgent(title)
+  await saveSeededAgentComposer({
+    agentId: agent.id,
+    config: createAgentSoulConfigWithModel(normalAgentSoulConfig, model),
+    shouldPublish: true,
+  })
+  await setAgentSiteAccessAndGetURL(agent.id, true)
+
+  const resource = { id: agent.id, kind: 'agent', name: title }
+  return wasCreated ? created(title, resource) : updated(title, resource)
+}
+
+const seedBackendApiAgent = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.backendApiEnabledAgent
+  const model = getStableModelResource(context)
+  if (!model)
+    return blocked(title, `${agentBuilderPreseededResources.stableChatModel} is not ready.`)
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}".`)
+
+  const { agent, created: wasCreated } = await ensureAgent(title)
+  await saveSeededAgentComposer({
+    agentId: agent.id,
+    config: createAgentSoulConfigWithModel(normalAgentSoulConfig, model),
+    shouldPublish: true,
+  })
+  const access = await setAgentApiAccess(agent.id, true)
+  if (access.api_key_count < 1)
+    await createAgentApiKey(agent.id)
+
+  const resource = { id: agent.id, kind: 'agent', name: title }
+  return wasCreated ? created(title, resource) : updated(title, resource)
+}
+
+const findWorkflow = (name: string) => {
+  const query = buildQuery({ limit: '20', mode: 'workflow', name, page: '1' })
+  return findConsoleResourceByName({
+    action: `Find seed workflow ${name}`,
+    path: `/console/api/apps?${query}`,
+    resourceName: name,
+  })
+}
+
+const seedWorkflowReference = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.workflowReferenceAgent
+  const workflowName = agentBuilderPreseededResources.referenceWorkflow
+  const model = getStableModelResource(context)
+  if (!model)
+    return blocked(title, `${agentBuilderPreseededResources.stableChatModel} is not ready.`)
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}" and workflow "${workflowName}".`)
+
+  const { agent, created: wasAgentCreated } = await ensureAgent(title)
+  await saveSeededAgentComposer({
+    agentId: agent.id,
+    config: createAgentSoulConfigWithModel(normalAgentSoulConfig, model),
+    shouldPublish: true,
+  })
+
+  let workflow = await findWorkflow(workflowName)
+  let wasWorkflowCreated = false
+  if (!workflow) {
+    workflow = await createTestApp(workflowName, 'workflow')
+    wasWorkflowCreated = true
+  }
+  await syncAgentV2WorkflowDraft(workflow.id, agent.id)
+  await publishWorkflowApp(workflow.id)
+
+  const resource = { id: workflow.id, kind: 'workflow', name: workflowName }
+  return wasAgentCreated || wasWorkflowCreated
+    ? created(`${title} / ${workflowName}`, resource)
+    : updated(`${title} / ${workflowName}`, resource)
+}
+
+const seedOAuthToolAgent = async (context: SeedContext) => {
+  const title = agentBuilderPreseededResources.oauthToolAgent
+  const credentialId = process.env[oauthToolCredentialIdEnv]?.trim()
+  const providerName = process.env[oauthToolProviderEnv]?.trim()
+  const toolName = process.env[oauthToolNameEnv]?.trim()
+  if (!credentialId || !providerName || !toolName) {
+    return blocked(
+      title,
+      `${oauthToolCredentialIdEnv}, ${oauthToolProviderEnv}, and ${oauthToolNameEnv} are required for OAuth2 tool seed.`,
+    )
+  }
+  if (context.dryRun)
+    return skipped(title, `Would create or update Agent "${title}" with OAuth2 tool ${providerName}/${toolName}.`)
+
+  const { agent, created: wasCreated } = await ensureAgent(title)
+  await saveSeededAgentComposer({
+    agentId: agent.id,
+    config: {
+      ...normalAgentSoulConfig,
+      tools: {
+        dify_tools: [
+          {
+            credential_ref: {
+              id: credentialId,
+              provider: providerName,
+              type: 'provider',
+            },
+            credential_type: 'oauth2',
+            enabled: true,
+            provider_id: providerName,
+            provider_type: 'builtin',
+            runtime_parameters: {},
+            tool_name: toolName,
+          },
+        ],
+      },
+    },
+  })
+
+  const resource = { id: agent.id, kind: 'agent', name: title }
+  return wasCreated ? created(title, resource) : updated(title, resource)
+}
+
+const agentV2BaseSeedTasks = (): SeedTask[] => [
+  {
+    id: 'marketplace-plugins',
+    title: 'Agent V2 marketplace plugins',
+    run: context => bootstrapMarketplacePlugins(context, {
+      defaultPluginIds: agentV2MarketplacePluginIds,
+      pluginIdsEnv: marketplacePluginIdsEnv,
+      pluginUniqueIdentifiersEnv: marketplacePluginUniqueIdentifiersEnv,
+      title: 'Agent V2 marketplace plugins',
+    }),
+  },
+  {
+    id: 'stable-model',
+    title: agentBuilderPreseededResources.stableChatModel,
+    run: seedStableModel,
+  },
+  {
+    id: 'agent-decision-model',
+    title: agentBuilderPreseededResources.agentDecisionChatModel,
+    run: seedAgentDecisionModel,
+  },
+  seedTool(agentBuilderPreseededResources.jsonReplaceTool),
+  seedTool(agentBuilderPreseededResources.tavilySearchTool),
+  {
+    id: 'ready-knowledge',
+    title: agentBuilderPreseededResources.agentKnowledgeBase,
+    run: seedReadyKnowledge,
+  },
+]
+
+const agentV2FullSeedTasks = (): SeedTask[] => [
+  ...agentV2BaseSeedTasks(),
+  {
+    id: 'indexing-knowledge',
+    title: agentBuilderPreseededResources.indexingKnowledgeBase,
+    run: async () => blocked(
+      agentBuilderPreseededResources.indexingKnowledgeBase,
+      'A deterministic long-lived "currently indexing" dataset seed is not implemented yet.',
+    ),
+  },
+  {
+    id: 'broken-model',
+    title: agentBuilderPreseededResources.brokenModelProvider,
+    run: async () => blocked(
+      agentBuilderPreseededResources.brokenModelProvider,
+      'Broken model fixture is validation-only for now; provide E2E_BROKEN_MODEL_PROVIDER and keep the model entry externally.',
+    ),
+  },
+  {
+    id: 'full-config-agent',
+    title: agentBuilderPreseededResources.fullConfigAgent,
+    run: seedFullConfigAgent,
+  },
+  {
+    id: 'tool-states-agent',
+    title: agentBuilderPreseededResources.toolStatesAgent,
+    run: seedToolStatesAgent,
+  },
+  {
+    id: 'oauth-tool-agent',
+    title: agentBuilderPreseededResources.oauthToolAgent,
+    run: seedOAuthToolAgent,
+  },
+  {
+    id: 'file-tree-agent',
+    title: agentBuilderPreseededResources.fileTreeAgent,
+    run: async () => blocked(
+      agentBuilderPreseededResources.fileTreeAgent,
+      'Agent drive arbitrary file upload does not have a stable public seed helper yet.',
+    ),
+  },
+  {
+    id: 'dual-retrieval-agent',
+    title: agentBuilderPreseededResources.dualRetrievalAgent,
+    run: seedDualRetrievalAgent,
+  },
+  {
+    id: 'published-web-app-agent',
+    title: agentBuilderPreseededResources.publishedWebAppAgent,
+    run: seedPublishedWebAppAgent,
+  },
+  {
+    id: 'backend-api-agent',
+    title: agentBuilderPreseededResources.backendApiEnabledAgent,
+    run: seedBackendApiAgent,
+  },
+  {
+    id: 'workflow-reference',
+    title: `${agentBuilderPreseededResources.workflowReferenceAgent} / ${agentBuilderPreseededResources.referenceWorkflow}`,
+    run: seedWorkflowReference,
+  },
+]
+
+export const createAgentV2SeedTasks = (profile: string = 'full'): SeedTask[] => {
+  if (profile === 'full')
+    return agentV2FullSeedTasks()
+
+  if (profile === 'external-runtime')
+    return agentV2BaseSeedTasks()
+
+  throw new Error(`Unknown Agent V2 seed profile "${profile}".`)
+}
diff --git a/e2e/features/agent-v2/support/tools.ts b/e2e/features/agent-v2/support/tools.ts
index d7f6eb60fe5..ccae6f92e69 100644
--- a/e2e/features/agent-v2/support/tools.ts
+++ b/e2e/features/agent-v2/support/tools.ts
@@ -1,5 +1,5 @@
 import type { DifyWorld } from '../../support/world'
-import { splitToolDisplayName } from './preflight/tools'
+import { splitToolDisplayName, splitToolResourceId } from './preflight/tools'
 
 export const getPreseededToolContract = (world: DifyWorld, resourceName: string) => {
   const resource = world.agentBuilder.preflight.preseededResources[resourceName]
@@ -10,11 +10,11 @@ export const getPreseededToolContract = (world: DifyWorld, resourceName: string)
   }
 
   const parsedDisplayName = splitToolDisplayName(resource.name)
-  const parsedToolId = splitToolDisplayName(resource.id)
+  const parsedToolId = splitToolResourceId(resource.id)
   if (!parsedDisplayName.ok)
     throw new Error(parsedDisplayName.reason)
-  if (!parsedToolId.ok)
-    throw new Error(parsedToolId.reason)
+  if (!parsedToolId.providerName || !parsedToolId.toolName)
+    throw new Error(`Preseeded tool "${resource.id}" must include provider and tool id segments.`)
 
   return {
     providerDisplayName: parsedDisplayName.providerName,
diff --git a/e2e/features/agent-v2/tools.feature b/e2e/features/agent-v2/tools.feature
index 305bf64fbc0..5feaac1ecc8 100644
--- a/e2e/features/agent-v2/tools.feature
+++ b/e2e/features/agent-v2/tools.feature
@@ -12,15 +12,34 @@ Feature: Agent v2 tools
     When I refresh the current page
     Then I should see the Agent v2 JSON Replace tool in the Tools section
 
-  @service-api-runtime @stable-model @tool-fixture
+  @core @oauth-tool-agent
+  Scenario: OAuth2 tool credentials stay authorized after Configure autosaves
+    Given I am signed in as the default E2E admin
+    And the Agent Builder preseeded Agent "E2E Agent With OAuth Tool" includes an OAuth2 tool credential
+    And an Agent v2 test agent with the OAuth2 tool credential fixture has been created via API
+    When I open the Agent v2 configure page
+    Then I should see the Agent v2 OAuth2 tool authorized in the Tools section
+    When I fill the Agent v2 prompt editor with the updated E2E prompt
+    Then the Agent v2 configuration should be saved automatically
+    And the Agent v2 OAuth2 tool credential should remain saved in the Agent v2 draft
+    When I refresh the current page
+    Then I should see the Agent v2 OAuth2 tool authorized in the Tools section
+    And the Agent v2 OAuth2 tool credential should remain saved in the Agent v2 draft
+
+  @service-api-runtime @external-model @agent-backend-runtime @stable-model @tool-fixture
   Scenario: JSON Replace tool runtime returns the replacement marker
     Given I am signed in as the default E2E admin
-    And Agent v2 JSON Replace runtime verification is available
+    And the Agent v2 runtime backend is available
     And the Agent Builder stable chat model is available
     And the Agent Builder preseeded tool "JSON Process / JSON Replace" is available
-    And a runnable Agent v2 test agent has been created via API
+    And a runnable Agent v2 test agent with the JSON Replace tool has been created via API
+    And Agent v2 Backend service API access has been enabled with a key via API
     When I open the Agent v2 configure page
-    Then Agent v2 JSON Replace runtime verification should be available
+    Then the Agent v2 JSON Replace tool should be saved in the Agent v2 draft
+    When I publish the Agent v2 draft
+    Then the Agent v2 draft should be published and up to date
+    When I send the Agent v2 Backend service API JSON Replace request
+    Then the Agent v2 Backend service API response should include the JSON Replace E2E marker
 
   @core
   Scenario: Tool selector shows an empty state for a missing tool search
diff --git a/e2e/features/step-definitions/agent-v2/access-point-helpers.ts b/e2e/features/step-definitions/agent-v2/access-point-helpers.ts
index 0dc2f9ee2d6..18e249c71d9 100644
--- a/e2e/features/step-definitions/agent-v2/access-point-helpers.ts
+++ b/e2e/features/step-definitions/agent-v2/access-point-helpers.ts
@@ -26,11 +26,16 @@ export const getPreseededResource = (
 export const getAccessRegion = (world: DifyWorld) =>
   world.getPage().getByRole('region', { name: 'Access Point' })
 
+export type AccessSurfaceName = 'Web app' | 'Backend service API'
+
+export const getAccessSurfaceCard = (world: DifyWorld, surface: AccessSurfaceName) =>
+  getAccessRegion(world).getByRole('article', { name: surface }).first()
+
 export const getWebAppCard = (world: DifyWorld) =>
-  getAccessRegion(world).locator('article').filter({ hasText: 'Web app' }).first()
+  getAccessSurfaceCard(world, 'Web app')
 
 export const getServiceApiCard = (world: DifyWorld) =>
-  getAccessRegion(world).locator('article').filter({ hasText: 'Backend service API' }).first()
+  getAccessSurfaceCard(world, 'Backend service API')
 
 export const getDialog = (world: DifyWorld, name: string | RegExp) =>
   world.getPage().getByRole('dialog', { name })
diff --git a/e2e/features/step-definitions/agent-v2/access-point-service-api.steps.ts b/e2e/features/step-definitions/agent-v2/access-point-service-api.steps.ts
index 0ea9e23a251..d23afca303d 100644
--- a/e2e/features/step-definitions/agent-v2/access-point-service-api.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/access-point-service-api.steps.ts
@@ -18,15 +18,6 @@ async function enableAgentApiAccessWithKey(world: DifyWorld) {
   world.agentBuilder.accessPoint.generatedApiKey = apiKey.token
 }
 
-Given(
-  'Agent v2 Backend service API access has been enabled via API',
-  async function (this: DifyWorld) {
-    const apiAccess = await setAgentApiAccess(getCurrentAgentId(this), true)
-
-    this.agentBuilder.accessPoint.serviceApiBaseURL = apiAccess.service_api_base_url
-  },
-)
-
 Given(
   'Agent v2 Backend service API access has been enabled with a key via API',
   async function (this: DifyWorld) {
@@ -181,26 +172,6 @@ Then('the Agent v2 API Reference should open in a new tab', async function (this
   this.agentBuilder.accessPoint.apiReferencePage = undefined
 })
 
-When('I disable Agent v2 Backend service API access', async function (this: DifyWorld) {
-  await getServiceApiCard(this).getByLabel('Toggle Backend service API access').click()
-})
-
-Then('Agent v2 Backend service API access should be out of service', async function (this: DifyWorld) {
-  const serviceApiCard = getServiceApiCard(this)
-
-  await expect(serviceApiCard.getByText('Out of service')).toBeVisible({ timeout: 30_000 })
-})
-
-When('I enable Agent v2 Backend service API access', async function (this: DifyWorld) {
-  await getServiceApiCard(this).getByLabel('Toggle Backend service API access').click()
-})
-
-Then('Agent v2 Backend service API access should be in service', async function (this: DifyWorld) {
-  const serviceApiCard = getServiceApiCard(this)
-
-  await expect(serviceApiCard.getByText('In service')).toBeVisible({ timeout: 30_000 })
-})
-
 When('I send the Agent v2 Backend service API minimal request', async function (this: DifyWorld) {
   const serviceApiBaseURL = this.agentBuilder.accessPoint.serviceApiBaseURL
   const apiKey = this.agentBuilder.accessPoint.generatedApiKey
@@ -225,11 +196,42 @@ When('I send the Agent v2 Backend service API knowledge request', async function
 
   this.agentBuilder.accessPoint.serviceApiResponse = await sendAgentServiceApiChatMessage({
     apiKey,
-    query: agentBuilderFixedInputs.customKnowledgeQuery,
+    query: agentBuilderFixedInputs.knowledgeRuntimeQuery,
     serviceApiBaseURL,
   })
 })
 
+const stringifyServiceApiBody = (body: unknown) => {
+  try {
+    return JSON.stringify(body)
+  }
+  catch {
+    return String(body)
+  }
+}
+
+const expectServiceApiResponseOK = (
+  response: NonNullable<DifyWorld['agentBuilder']['accessPoint']['serviceApiResponse']>,
+  action: string,
+) => {
+  if (response.ok)
+    return
+
+  throw new Error(`${action} failed with ${response.status}: ${stringifyServiceApiBody(response.body)}`)
+}
+
+const expectServiceApiResponseIncludes = (
+  response: NonNullable<DifyWorld['agentBuilder']['accessPoint']['serviceApiResponse']>,
+  expectedToken: string,
+  action: string,
+) => {
+  expectServiceApiResponseOK(response, action)
+
+  const body = stringifyServiceApiBody(response.body)
+  if (!body.includes(expectedToken))
+    throw new Error(`${action} response did not include ${expectedToken}: ${body}`)
+}
+
 Then(
   'the Agent v2 Backend service API request should be rejected while disabled',
   async function (this: DifyWorld) {
@@ -250,8 +252,11 @@ Then(
     if (!response)
       throw new Error('No Agent v2 Backend service API response was recorded.')
 
-    expect(response.ok).toBe(true)
-    expect(JSON.stringify(response.body)).toContain(agentBuilderExpectedTokens.knowledgeReply)
+    expectServiceApiResponseIncludes(
+      response,
+      agentBuilderExpectedTokens.knowledgeReply,
+      'Agent v2 Backend service API knowledge request',
+    )
   },
 )
 
@@ -262,7 +267,10 @@ Then(
     if (!response)
       throw new Error('No Agent v2 Backend service API response was recorded.')
 
-    expect(response.ok).toBe(true)
-    expect(JSON.stringify(response.body)).toContain(agentBuilderExpectedTokens.agentReply)
+    expectServiceApiResponseIncludes(
+      response,
+      agentBuilderExpectedTokens.agentReply,
+      'Agent v2 Backend service API request',
+    )
   },
 )
diff --git a/e2e/features/step-definitions/agent-v2/access-point-web-app.steps.ts b/e2e/features/step-definitions/agent-v2/access-point-web-app.steps.ts
index 34534edcbc8..d9d85e83772 100644
--- a/e2e/features/step-definitions/agent-v2/access-point-web-app.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/access-point-web-app.steps.ts
@@ -1,7 +1,6 @@
 import type { DifyWorld } from '../../support/world'
 import { Given, Then, When } from '@cucumber/cucumber'
 import { expect } from '@playwright/test'
-import { setAgentSiteAccessAndGetURL } from '../../agent-v2/support/access-point'
 import { getAgentComposerDraft } from '../../agent-v2/support/agent'
 import { agentBuilderExpectedTokens } from '../../agent-v2/support/agent-builder-resources'
 import { skipBlockedPrecondition } from '../../agent-v2/support/preflight/common'
@@ -11,15 +10,7 @@ import {
   getWebAppCard,
 } from './access-point-helpers'
 
-Given(
-  'Agent v2 Web app access has been enabled via API',
-  async function (this: DifyWorld) {
-    this.agentBuilder.accessPoint.webAppURL = await setAgentSiteAccessAndGetURL(
-      getCurrentAgentId(this),
-      true,
-    )
-  },
-)
+const WEB_APP_RUNTIME_RESPONSE_STEP_TIMEOUT_MS = 180_000
 
 Then('I should see the Agent v2 Web app access URL', async function (this: DifyWorld) {
   const webAppCard = getWebAppCard(this)
@@ -101,6 +92,7 @@ Then('the Agent v2 Web app should open in a new tab', async function (this: Dify
 
 Then(
   'the Agent v2 Web app response should include the updated E2E marker',
+  { timeout: WEB_APP_RUNTIME_RESPONSE_STEP_TIMEOUT_MS },
   async function (this: DifyWorld) {
     const webAppPage = this.agentBuilder.accessPoint.webAppPage
     if (!webAppPage)
@@ -113,6 +105,7 @@ Then(
 
 Then(
   'the Agent v2 Web app response should include the normal E2E marker',
+  { timeout: WEB_APP_RUNTIME_RESPONSE_STEP_TIMEOUT_MS },
   async function (this: DifyWorld) {
     const webAppPage = this.agentBuilder.accessPoint.webAppPage
     if (!webAppPage)
@@ -195,25 +188,6 @@ Then(
   },
 )
 
-When('I disable Agent v2 Web app access', async function (this: DifyWorld) {
-  const webAppCard = getWebAppCard(this)
-  const launchLink = webAppCard.getByRole('link', { name: 'Launch' })
-  const href = await launchLink.getAttribute('href')
-  if (!href)
-    throw new Error('Agent v2 Web app Launch link does not expose an href.')
-
-  this.agentBuilder.accessPoint.webAppURL = href
-
-  await webAppCard.getByLabel('Toggle Web app access').click()
-})
-
-Then('Agent v2 Web app access should be out of service', async function (this: DifyWorld) {
-  const webAppCard = getWebAppCard(this)
-
-  await expect(webAppCard.getByText('Out of service')).toBeVisible()
-  await expect(webAppCard.getByRole('button', { name: 'Launch' })).toBeDisabled()
-})
-
 Given(
   'Agent v2 disabled Web app public unavailable state is available',
   async function (this: DifyWorld) {
@@ -253,17 +227,6 @@ Then('the disabled Agent v2 Web app should show an unavailable state', async fun
   this.agentBuilder.accessPoint.webAppPage = undefined
 })
 
-When('I enable Agent v2 Web app access', async function (this: DifyWorld) {
-  await getWebAppCard(this).getByLabel('Toggle Web app access').click()
-})
-
-Then('Agent v2 Web app access should be in service', async function (this: DifyWorld) {
-  const webAppCard = getWebAppCard(this)
-
-  await expect(webAppCard.getByText('In service')).toBeVisible()
-  await expect(webAppCard.getByRole('link', { name: 'Launch' })).toBeVisible()
-})
-
 When('I open the restored Agent v2 Web app URL', async function (this: DifyWorld) {
   const webAppURL = this.agentBuilder.accessPoint.webAppURL
   if (!webAppURL)
diff --git a/e2e/features/step-definitions/agent-v2/access-point.steps.ts b/e2e/features/step-definitions/agent-v2/access-point.steps.ts
index 00f99593719..40c7e852aa3 100644
--- a/e2e/features/step-definitions/agent-v2/access-point.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/access-point.steps.ts
@@ -1,9 +1,15 @@
 import type { DifyWorld } from '../../support/world'
+import type { AccessSurfaceName } from './access-point-helpers'
 import { Given, Then, When } from '@cucumber/cucumber'
 import { expect } from '@playwright/test'
+import {
+  setAgentApiAccess,
+  setAgentSiteAccessAndGetURL,
+} from '../../agent-v2/support/access-point'
 import { getAgentAccessPath, publishAgent } from '../../agent-v2/support/agent'
 import {
   getAccessRegion,
+  getAccessSurfaceCard,
   getCurrentAgentId,
   getPreseededResource,
 } from './access-point-helpers'
@@ -12,6 +18,22 @@ Given('the Agent v2 draft has been published via API', async function (this: Dif
   await publishAgent(getCurrentAgentId(this))
 })
 
+Given(
+  /^Agent v2 (Web app|Backend service API) access has been enabled via API$/,
+  async function (this: DifyWorld, surface: AccessSurfaceName) {
+    if (surface === 'Web app') {
+      this.agentBuilder.accessPoint.webAppURL = await setAgentSiteAccessAndGetURL(
+        getCurrentAgentId(this),
+        true,
+      )
+      return
+    }
+
+    const apiAccess = await setAgentApiAccess(getCurrentAgentId(this), true)
+    this.agentBuilder.accessPoint.serviceApiBaseURL = apiAccess.service_api_base_url
+  },
+)
+
 When('I open the Agent v2 Access Point page', async function (this: DifyWorld) {
   await this.getPage().goto(getAgentAccessPath(getCurrentAgentId(this)))
 })
@@ -70,3 +92,50 @@ Then('I should see the Agent v2 Access Point overview', async function (this: Di
   await expect(accessRegion.getByRole('columnheader', { name: 'Actions' })).toBeVisible()
   await expect(accessRegion.getByText('No workflow references yet.')).toBeVisible()
 })
+
+When(
+  /^I disable Agent v2 (Web app|Backend service API) access$/,
+  async function (this: DifyWorld, surface: AccessSurfaceName) {
+    const accessSurfaceCard = getAccessSurfaceCard(this, surface)
+
+    if (surface === 'Web app') {
+      const launchLink = accessSurfaceCard.getByRole('link', { name: 'Launch' })
+      const href = await launchLink.getAttribute('href')
+      if (!href)
+        throw new Error('Agent v2 Web app Launch link does not expose an href.')
+
+      this.agentBuilder.accessPoint.webAppURL = href
+    }
+
+    await accessSurfaceCard.getByLabel(`Toggle ${surface} access`).click()
+  },
+)
+
+When(
+  /^I enable Agent v2 (Web app|Backend service API) access$/,
+  async function (this: DifyWorld, surface: AccessSurfaceName) {
+    await getAccessSurfaceCard(this, surface).getByLabel(`Toggle ${surface} access`).click()
+  },
+)
+
+Then(
+  /^Agent v2 (Web app|Backend service API) access should be out of service$/,
+  async function (this: DifyWorld, surface: AccessSurfaceName) {
+    const accessSurfaceCard = getAccessSurfaceCard(this, surface)
+
+    await expect(accessSurfaceCard.getByText('Out of service')).toBeVisible({ timeout: 30_000 })
+    if (surface === 'Web app')
+      await expect(accessSurfaceCard.getByRole('button', { name: 'Launch' })).toBeDisabled()
+  },
+)
+
+Then(
+  /^Agent v2 (Web app|Backend service API) access should be in service$/,
+  async function (this: DifyWorld, surface: AccessSurfaceName) {
+    const accessSurfaceCard = getAccessSurfaceCard(this, surface)
+
+    await expect(accessSurfaceCard.getByText('In service')).toBeVisible({ timeout: 30_000 })
+    if (surface === 'Web app')
+      await expect(accessSurfaceCard.getByRole('link', { name: 'Launch' })).toBeVisible()
+  },
+)
diff --git a/e2e/features/step-definitions/agent-v2/agent-roster.steps.ts b/e2e/features/step-definitions/agent-v2/agent-roster.steps.ts
index 99e69c1ba13..70fd03195e3 100644
--- a/e2e/features/step-definitions/agent-v2/agent-roster.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/agent-roster.steps.ts
@@ -16,8 +16,8 @@ When('I create an Agent v2 test agent from the Agent Roster', async function (th
   const dialog = page.getByRole('dialog', { name: 'Create agent' })
   await expect(dialog).toBeVisible()
   await dialog.getByRole('textbox', { name: 'Name' }).fill(agentName)
-  await dialog.getByRole('textbox', { name: 'Role' }).fill(agentRole)
-  await dialog.getByRole('textbox', { name: 'Description' }).fill(agentDescription)
+  await dialog.getByRole('textbox', { name: /Role.*optional/ }).fill(agentRole)
+  await dialog.getByRole('textbox', { name: /Description.*optional/ }).fill(agentDescription)
 
   const createResponsePromise = page.waitForResponse(response => (
     response.request().method() === 'POST'
diff --git a/e2e/features/step-definitions/agent-v2/build-draft.steps.ts b/e2e/features/step-definitions/agent-v2/build-draft.steps.ts
index 163932deb10..b9e5ff7335b 100644
--- a/e2e/features/step-definitions/agent-v2/build-draft.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/build-draft.steps.ts
@@ -27,6 +27,8 @@ import {
   uploadSummaryConfigSkillForBuildDraft,
 } from './configure-helpers'
 
+const BUILD_DRAFT_RUNTIME_STEP_TIMEOUT_MS = 180_000
+
 Given(
   'an Agent v2 Build draft adds the supported E2E files, skills, and env',
   async function (this: DifyWorld) {
@@ -108,30 +110,34 @@ Given(
   },
 )
 
-When('I generate an Agent v2 Build draft from the fixed instruction', async function (this: DifyWorld) {
-  const page = this.getPage()
-  const agentId = getCurrentAgentId(this)
-  const instruction = (await readFile(getAgentBuilderTestMaterialPath('buildInstruction'), 'utf8')).trim()
+When(
+  'I generate an Agent v2 Build draft from the fixed instruction',
+  { timeout: 180_000 },
+  async function (this: DifyWorld) {
+    const page = this.getPage()
+    const agentId = getCurrentAgentId(this)
+    const instruction = (await readFile(getAgentBuilderTestMaterialPath('buildInstruction'), 'utf8')).trim()
 
-  await page.getByRole('button', { name: 'Build' }).click()
-  await page.getByPlaceholder('Describe what your agent should do').fill(instruction)
+    await page.getByRole('button', { exact: true, name: 'Build' }).click()
+    await page.getByPlaceholder('Describe what your agent should do').fill(instruction)
 
-  const checkoutResponsePromise = page.waitForResponse(response => (
-    response.request().method() === 'POST'
-    && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/build-draft/checkout`)
-  ))
-  const chatResponsePromise = page.waitForResponse(response => (
-    response.request().method() === 'POST'
-    && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/chat-messages`)
-  ))
+    const checkoutResponsePromise = page.waitForResponse(response => (
+      response.request().method() === 'POST'
+      && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/build-draft/checkout`)
+    ))
+    const chatResponsePromise = page.waitForResponse(response => (
+      response.request().method() === 'POST'
+      && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/chat-messages`)
+    ))
 
-  await page.getByRole('button', { name: 'Start build' }).click()
-  expect((await checkoutResponsePromise).ok()).toBe(true)
-  expect((await chatResponsePromise).ok()).toBe(true)
-  await expect(page.getByText('Build draft')).toBeVisible({ timeout: 120_000 })
-  await expect(page.getByRole('button', { name: 'Apply' })).toBeEnabled({ timeout: 120_000 })
-  await expect(page.getByRole('button', { name: 'Discard' })).toBeEnabled()
-})
+    await page.getByRole('button', { name: 'Start build' }).click()
+    expect((await checkoutResponsePromise).ok()).toBe(true)
+    expect((await chatResponsePromise).ok()).toBe(true)
+    await expect(page.getByText('Build draft')).toBeVisible({ timeout: 120_000 })
+    await expect(page.getByRole('button', { exact: true, name: 'Apply' })).toBeEnabled({ timeout: 120_000 })
+    await expect(page.getByRole('button', { exact: true, name: 'Discard' })).toBeEnabled()
+  },
+)
 
 const expectPageResponseOK = async (response: Response, action: string) => {
   if (response.ok())
@@ -150,34 +156,38 @@ const expectPageResponseOK = async (response: Response, action: string) => {
 }
 
 When('I discard the Agent v2 Build draft', async function (this: DifyWorld) {
-  await this.getPage().getByRole('button', { name: 'Discard' }).click()
+  await this.getPage().getByRole('button', { exact: true, name: 'Discard' }).click()
 })
 
-When('I apply the Agent v2 Build draft', async function (this: DifyWorld) {
-  const page = this.getPage()
-  const agentId = getCurrentAgentId(this)
-  const applyButton = page.getByRole('button', { name: 'Apply' })
+When(
+  'I apply the Agent v2 Build draft',
+  { timeout: BUILD_DRAFT_RUNTIME_STEP_TIMEOUT_MS },
+  async function (this: DifyWorld) {
+    const page = this.getPage()
+    const agentId = getCurrentAgentId(this)
+    const applyButton = page.getByRole('button', { exact: true, name: 'Apply' })
 
-  await expect(applyButton).toBeEnabled({ timeout: 30_000 })
-  const finalizeResponsePromise = page.waitForResponse(response => (
-    response.request().method() === 'POST'
-    && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/build-chat/finalize`)
-  ), { timeout: 120_000 })
-  const applyResponsePromise = page.waitForResponse(response => (
-    response.request().method() === 'POST'
-    && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/build-draft/apply`)
-  ), { timeout: 120_000 })
+    await expect(applyButton).toBeEnabled({ timeout: 30_000 })
+    const finalizeResponsePromise = page.waitForResponse(response => (
+      response.request().method() === 'POST'
+      && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/build-chat/finalize`)
+    ), { timeout: 120_000 })
+    const applyResponsePromise = page.waitForResponse(response => (
+      response.request().method() === 'POST'
+      && new URL(response.url()).pathname.endsWith(`/console/api/agent/${agentId}/build-draft/apply`)
+    ), { timeout: 120_000 })
 
-  await applyButton.click()
-  await expectPageResponseOK(await finalizeResponsePromise, 'Finalize Agent v2 Build draft')
-  await expectPageResponseOK(await applyResponsePromise, 'Apply Agent v2 Build draft')
-  await expect(page.getByText('Action succeeded')).toBeVisible()
-})
+    await applyButton.click()
+    await expectPageResponseOK(await finalizeResponsePromise, 'Finalize Agent v2 Build draft')
+    await expectPageResponseOK(await applyResponsePromise, 'Apply Agent v2 Build draft')
+    await expect(page.getByText('Action succeeded')).toBeVisible()
+  },
+)
 
 async function skipBuildDraftToolWriteback(world: DifyWorld) {
   return skipBlockedPrecondition(
     world,
-    'Build draft Dify Tool writeback is not available: Build draft currently supports files, skills, and env only.',
+    'Build chat Dify Tool writeback is not available: finalize/config mutation paths do not support tools; current passing coverage only verifies API-seeded Build draft apply for files, skills, and env.',
     {
       owner: 'product',
       remediation: 'Define and implement Build draft Tool writeback before enabling this scenario.',
@@ -216,8 +226,8 @@ Then('I should see the Agent v2 Build draft pending changes', async function (th
   const page = this.getPage()
 
   await expect(page.getByText('Build draft')).toBeVisible({ timeout: 30_000 })
-  await expect(page.getByRole('button', { name: 'Apply' })).toBeEnabled()
-  await expect(page.getByRole('button', { name: 'Discard' })).toBeEnabled()
+  await expect(page.getByRole('button', { exact: true, name: 'Apply' })).toBeEnabled()
+  await expect(page.getByRole('button', { exact: true, name: 'Discard' })).toBeEnabled()
 })
 
 Then('I should see the Agent v2 Build mode confirmation state', async function (this: DifyWorld) {
diff --git a/e2e/features/step-definitions/agent-v2/configure.steps.ts b/e2e/features/step-definitions/agent-v2/configure.steps.ts
index 7d2af762d5d..bf9d4f0dfce 100644
--- a/e2e/features/step-definitions/agent-v2/configure.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/configure.steps.ts
@@ -100,6 +100,27 @@ Given('a runnable Agent v2 test agent has been created via API', async function
   this.lastCreatedAgentRole = agent.role ?? undefined
 })
 
+Given(
+  'a runnable Agent v2 test agent using the agent-decision model has been created via API',
+  async function (this: DifyWorld) {
+    if (!this.agentBuilder.preflight.agentDecisionModel) {
+      throw new Error(
+        'Create an agent-decision Agent v2 test agent after agent-decision model preflight.',
+      )
+    }
+
+    const agent = await createConfiguredTestAgent({
+      agentSoul: createAgentSoulConfigWithModel(
+        normalAgentSoulConfig,
+        this.agentBuilder.preflight.agentDecisionModel,
+      ),
+    })
+    this.createdAgentIds.push(agent.id)
+    this.lastCreatedAgentName = agent.name
+    this.lastCreatedAgentRole = agent.role ?? undefined
+  },
+)
+
 Given('a minimal Agent v2 composer draft has been synced', async function (this: DifyWorld) {
   const agentId = getCurrentAgentId(this)
 
diff --git a/e2e/features/step-definitions/agent-v2/files.steps.ts b/e2e/features/step-definitions/agent-v2/files.steps.ts
index e8780f5a032..76f8eaf6617 100644
--- a/e2e/features/step-definitions/agent-v2/files.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/files.steps.ts
@@ -2,7 +2,10 @@ import type { DifyWorld } from '../../support/world'
 import { Given, Then, When } from '@cucumber/cucumber'
 import { expect } from '@playwright/test'
 import { skipBlockedPrecondition } from '../../agent-v2/support/preflight/common'
-import { agentBuilderFileTreeFixtureFileNames } from '../../agent-v2/support/test-materials'
+import {
+  agentBuilderFileTreeFixtureFileNames,
+  agentBuilderTestMaterials,
+} from '../../agent-v2/support/test-materials'
 import {
   expectAgentConfigFileHidden,
   expectAgentConfigFileSaved,
@@ -22,6 +25,51 @@ When('I upload the special-name Agent v2 file from the Files section', async fun
   await uploadAgentConfigFile(this, 'specialFilename')
 })
 
+When('I drop multiple Agent v2 files into the Files upload dialog', async function (this: DifyWorld) {
+  const page = this.getPage()
+
+  await page.getByRole('button', { name: 'Add file' }).click()
+  const dialog = page.getByRole('dialog', { name: 'Upload file' })
+  await expect(dialog).toBeVisible()
+
+  const dropZone = dialog.getByRole('group', { name: 'Upload file' })
+  await expect(dropZone).toBeVisible()
+
+  const droppedFileNames: [string, string] = [
+    agentBuilderTestMaterials.smallFile,
+    agentBuilderTestMaterials.emptyFile,
+  ]
+  const dataTransfer = await page.evaluateHandle(([smallFileName, emptyFileName]: [string, string]) => {
+    const transfer = new DataTransfer()
+    transfer.items.add(new File(['small agent file'], smallFileName, { type: 'text/plain' }))
+    transfer.items.add(new File([''], emptyFileName, { type: 'text/plain' }))
+    return transfer
+  }, droppedFileNames)
+
+  await dropZone.dispatchEvent('dragenter', { dataTransfer })
+  await dropZone.dispatchEvent('dragover', { dataTransfer })
+  await dropZone.dispatchEvent('drop', { dataTransfer })
+  await dataTransfer.dispose()
+})
+
+Then(
+  'the Agent v2 Files upload dialog should reject the multiple-file drop',
+  async function (this: DifyWorld) {
+    const page = this.getPage()
+    const dialog = page.getByRole('dialog', { name: 'Upload file' })
+
+    await expect(page.getByText('Upload one file.')).toBeVisible()
+    await expect(dialog.getByRole('button', { name: 'Upload' })).toBeDisabled()
+    await expect(dialog.getByText(agentBuilderTestMaterials.smallFile, { exact: true })).not.toBeVisible()
+    await expect(dialog.getByText(agentBuilderTestMaterials.emptyFile, { exact: true })).not.toBeVisible()
+  },
+)
+
+Then('I should not see the dropped Agent v2 files in the Files section', async function (this: DifyWorld) {
+  await expectAgentConfigFileHidden(this, 'smallFile')
+  await expectAgentConfigFileHidden(this, 'emptyFile')
+})
+
 Then(
   'I should see the Agent v2 file fixture entries in the current flat Files list',
   async function (this: DifyWorld) {
@@ -120,25 +168,6 @@ Then('Agent v2 oversized file rejection should be available', async function (th
   return skipOversizedFileRejection(this)
 })
 
-async function skipSingleBatchFileCountLimits(world: DifyWorld) {
-  return skipBlockedPrecondition(
-    world,
-    'Agent v2 single-batch file count limits are not reachable: the current Agent config file upload dialog accepts one file per upload.',
-    {
-      owner: 'product',
-      remediation: 'Define multi-file upload behavior and its count-limit error before enabling this scenario.',
-    },
-  )
-}
-
-Given('Agent v2 single-batch file count limits are available', async function (this: DifyWorld) {
-  return skipSingleBatchFileCountLimits(this)
-})
-
-Then('Agent v2 single-batch file count limits should be available', async function (this: DifyWorld) {
-  return skipSingleBatchFileCountLimits(this)
-})
-
 async function skipTotalFileCountLimits(world: DifyWorld) {
   return skipBlockedPrecondition(
     world,
diff --git a/e2e/features/step-definitions/agent-v2/knowledge.steps.ts b/e2e/features/step-definitions/agent-v2/knowledge.steps.ts
index fad39fac655..9cdf210248c 100644
--- a/e2e/features/step-definitions/agent-v2/knowledge.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/knowledge.steps.ts
@@ -107,7 +107,7 @@ const expectKnowledgeRetrievalDraft = async (
         mode: query.mode,
         name: knowledgeSet.name,
         retrievalMode: retrieval.mode,
-        value: query.value,
+        value: query.value ?? undefined,
       }
     },
     { timeout: 30_000 },
@@ -149,6 +149,8 @@ When(
     await selectPreseededKnowledgeBase(this, dialog)
     await expect(dialog.getByText(getPreseededKnowledgeBase(this).name, { exact: true }))
       .toBeVisible()
+    await this.getPage().keyboard.press('Escape')
+    await expect(dialog).toBeHidden()
   },
 )
 
@@ -164,6 +166,8 @@ When(
     await selectPreseededKnowledgeBase(this, dialog)
     await expect(dialog.getByText(getPreseededKnowledgeBase(this).name, { exact: true }))
       .toBeVisible()
+    await this.getPage().keyboard.press('Escape')
+    await expect(dialog).toBeHidden()
   },
 )
 
diff --git a/e2e/features/step-definitions/agent-v2/preflight.steps.ts b/e2e/features/step-definitions/agent-v2/preflight.steps.ts
index 09080d7a873..bd62425ba34 100644
--- a/e2e/features/step-definitions/agent-v2/preflight.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/preflight.steps.ts
@@ -5,6 +5,7 @@ import {
   skipMissingPreseededAgentPublishedWebApp,
   skipMissingPreseededAgentWorkflowReference,
 } from '../../agent-v2/support/preflight/access'
+import { skipMissingAgentBackendRuntime } from '../../agent-v2/support/preflight/agent-backend'
 import {
   skipMissingPreseededAgent,
   skipMissingPreseededAgentDriveSkill,
@@ -12,6 +13,7 @@ import {
   skipMissingPreseededAgentFlatFileFixtureConfiguration,
   skipMissingPreseededDualRetrievalAgentConfiguration,
   skipMissingPreseededFullConfigAgentCoreConfiguration,
+  skipMissingPreseededOAuthToolAgentConfiguration,
   skipMissingPreseededToolStatesAgentConfiguration,
   skipMissingPreseededWorkflow,
 } from '../../agent-v2/support/preflight/agents'
@@ -21,6 +23,7 @@ import {
   skipMissingReadyPreseededDataset,
 } from '../../agent-v2/support/preflight/datasets'
 import {
+  skipMissingAgentBuilderAgentDecisionChatModel,
   skipMissingAgentBuilderBrokenChatModel,
   skipMissingAgentBuilderStableChatModel,
 } from '../../agent-v2/support/preflight/models'
@@ -34,6 +37,14 @@ Given('the Agent Builder stable chat model is available', async function (this:
   this.agentBuilder.preflight.stableModel = stableModel
 })
 
+Given('the Agent Builder agent-decision chat model is available', async function (this: DifyWorld) {
+  const agentDecisionModel = await skipMissingAgentBuilderAgentDecisionChatModel(this)
+  if (agentDecisionModel === 'skipped')
+    return agentDecisionModel
+
+  this.agentBuilder.preflight.agentDecisionModel = agentDecisionModel
+})
+
 Given('the Agent Builder broken chat model is available', async function (this: DifyWorld) {
   const brokenModel = await skipMissingAgentBuilderBrokenChatModel(this)
   if (brokenModel === 'skipped')
@@ -42,6 +53,12 @@ Given('the Agent Builder broken chat model is available', async function (this:
   this.agentBuilder.preflight.brokenModel = brokenModel
 })
 
+Given('the Agent v2 runtime backend is available', async function (this: DifyWorld) {
+  const runtimeBackend = await skipMissingAgentBackendRuntime(this)
+  if (runtimeBackend === 'skipped')
+    return runtimeBackend
+})
+
 Given(
   'the Agent Builder preseeded Agent {string} is available',
   async function (this: DifyWorld, resourceName: string) {
@@ -142,6 +159,18 @@ Given(
   },
 )
 
+Given(
+  'the Agent Builder preseeded Agent {string} includes an OAuth2 tool credential',
+  async function (this: DifyWorld, agentName: string) {
+    const resource = await skipMissingPreseededOAuthToolAgentConfiguration(this, agentName)
+    if (resource === 'skipped')
+      return resource
+
+    this.agentBuilder.preflight.preseededResources[`${agentName} / OAuth2 tool credential`]
+      = resource
+  },
+)
+
 Given(
   'the Agent Builder preseeded Agent {string} includes the dual retrieval fixture configuration',
   async function (this: DifyWorld, agentName: string) {
diff --git a/e2e/features/step-definitions/agent-v2/tools.steps.ts b/e2e/features/step-definitions/agent-v2/tools.steps.ts
index 889d4b2a3a1..78eab7a488a 100644
--- a/e2e/features/step-definitions/agent-v2/tools.steps.ts
+++ b/e2e/features/step-definitions/agent-v2/tools.steps.ts
@@ -1,9 +1,13 @@
+import type { AgentSoulDifyToolConfig } from '@dify/contracts/api/console/agent/types.gen'
 import type { DifyWorld } from '../../support/world'
 import { Given, Then, When } from '@cucumber/cucumber'
 import { expect } from '@playwright/test'
-import { getAgentComposerDraft } from '../../agent-v2/support/agent'
-import { agentBuilderFixedInputs, agentBuilderPreseededResources } from '../../agent-v2/support/agent-builder-resources'
-import { asArray, asRecord, skipBlockedPrecondition } from '../../agent-v2/support/preflight/common'
+import { sendAgentServiceApiChatMessage } from '../../agent-v2/support/access-point'
+import { createConfiguredTestAgent, getAgentComposerDraft } from '../../agent-v2/support/agent'
+import { agentBuilderExpectedTokens, agentBuilderFixedInputs, agentBuilderPreseededResources } from '../../agent-v2/support/agent-builder-resources'
+import { createAgentSoulConfigWithDifyTool, createAgentSoulConfigWithModel, normalAgentSoulConfig } from '../../agent-v2/support/agent-soul'
+import { getPreseededOAuthToolConfig } from '../../agent-v2/support/preflight/agents'
+import { asArray, asRecord, asString } from '../../agent-v2/support/preflight/common'
 import { hasToolEntry } from '../../agent-v2/support/preflight/tools'
 import { getPreseededToolContract } from '../../agent-v2/support/tools'
 import { expectProviderToolActionVisible, getCurrentAgentId } from './configure-helpers'
@@ -14,6 +18,14 @@ const getToolsSection = (world: DifyWorld) =>
 const getToolSelectorSearch = (world: DifyWorld) =>
   world.getPage().getByRole('textbox', { name: 'Search integrations...' })
 
+const jsonReplaceRuntimePrompt = [
+  'You are a Dify Agent E2E JSON tool verifier.',
+  'When the user asks to verify JSON Replace, call the JSON Replace tool exactly once.',
+  `Use content {"marker":"${agentBuilderExpectedTokens.jsonToolBefore}","nested":{"status":"keep"}}.`,
+  'Use query $.marker and replace_value E2E_AFTER.',
+  'After the tool returns, answer with the resulting JSON and include E2E_AFTER.',
+].join(' ')
+
 const expectJsonReplaceToolDraft = async (world: DifyWorld) => {
   const agentId = getCurrentAgentId(world)
   const tool = getPreseededToolContract(world, agentBuilderPreseededResources.jsonReplaceTool)
@@ -29,17 +41,176 @@ const expectJsonReplaceToolDraft = async (world: DifyWorld) => {
   ).toBe(true)
 }
 
-async function skipJsonReplaceRuntimeVerification(world: DifyWorld) {
-  return skipBlockedPrecondition(
-    world,
-    'Agent v2 JSON Replace runtime verification is blocked: the suite needs the JSON Process / JSON Replace runtime parameter contract and a deterministic published-runtime prompt before asserting tool execution.',
-    {
-      owner: 'test/seed',
-      remediation: 'Seed the JSON Replace tool runtime contract, then verify execution through published Web app or Backend service API instead of Builder Preview.',
-    },
-  )
+const getOAuth2ToolEntries = async (agentId: string) => {
+  const draft = await getAgentComposerDraft(agentId)
+
+  return asArray(asRecord(draft.agent_soul?.tools).dify_tools).filter((item) => {
+    const record = asRecord(item)
+
+    return record.credential_type === 'oauth2'
+      && Boolean(asString(asRecord(record.credential_ref).id))
+  })
 }
 
+const getOAuth2ToolDisplayName = async (world: DifyWorld) => {
+  const [tool] = await getOAuth2ToolEntries(getCurrentAgentId(world))
+  const record = asRecord(tool)
+  const providerName = asString(record.provider) || asString(record.provider_id) || asString(record.plugin_id)
+  const toolName = asString(record.name) || asString(record.tool_name)
+
+  if (!providerName || !toolName)
+    throw new Error('Agent v2 OAuth2 tool fixture must include provider and tool names.')
+
+  return `${providerName} / ${toolName}`
+}
+
+const getPreseededOAuthToolAgent = (world: DifyWorld) => {
+  const resource = world.agentBuilder.preflight.preseededResources[
+    `${agentBuilderPreseededResources.oauthToolAgent} / OAuth2 tool credential`
+  ]
+  if (!resource || resource.kind !== 'agent') {
+    throw new Error(
+      `Preseeded Agent "${agentBuilderPreseededResources.oauthToolAgent}" OAuth2 tool credential fixture is not available. Run the matching preflight step first.`,
+    )
+  }
+
+  return resource
+}
+
+const jsonReplaceToolConfig = (world: DifyWorld): AgentSoulDifyToolConfig => {
+  const tool = getPreseededToolContract(world, agentBuilderPreseededResources.jsonReplaceTool)
+
+  return {
+    credential_type: 'unauthorized',
+    enabled: true,
+    provider: tool.providerDisplayName,
+    provider_id: tool.providerName,
+    provider_type: 'builtin',
+    runtime_parameters: {
+      ensure_ascii: true,
+      replace_model: 'value',
+      value_decode: false,
+    },
+    tool_name: tool.toolName,
+  }
+}
+
+const getServiceApiSseEvents = (body: unknown) =>
+  asArray(asRecord(body).events).map(asRecord)
+
+const getServiceApiEventData = (event: Record<string, unknown>) => asRecord(event.data)
+
+const getServiceApiEventName = (event: Record<string, unknown>) => {
+  const data = getServiceApiEventData(event)
+
+  return asString(data.event) || asString(event.event)
+}
+
+const summarizeServiceApiRuntimeEvents = (body: unknown) =>
+  getServiceApiSseEvents(body).map((event, index) => {
+    const data = getServiceApiEventData(event)
+    const observation = asString(data.observation)
+    const answer = asString(data.answer)
+
+    return {
+      index,
+      event: getServiceApiEventName(event),
+      tool: asString(data.tool),
+      hasToolInput: Boolean(asString(data.tool_input)),
+      hasObservation: Boolean(observation),
+      observation: observation.slice(0, 180),
+      answer: answer.slice(0, 180),
+    }
+  })
+
+const findJsonReplaceRuntimeThought = (body: unknown) =>
+  getServiceApiSseEvents(body).find((event) => {
+    const data = getServiceApiEventData(event)
+    const toolInput = asString(data.tool_input)
+    const observation = asString(data.observation)
+
+    return getServiceApiEventName(event) === 'agent_thought'
+      && asString(data.tool) === 'json_replace'
+      && toolInput.includes(agentBuilderExpectedTokens.jsonToolBefore)
+      && toolInput.includes('$.marker')
+      && observation.includes(agentBuilderExpectedTokens.jsonToolAfter)
+  })
+
+const expectOAuth2CredentialPreserved = async (world: DifyWorld) => {
+  const preseededAgent = getPreseededOAuthToolAgent(world)
+  const expectedTool = await getPreseededOAuthToolConfig(preseededAgent.id)
+  const expected = asRecord(expectedTool)
+  const expectedCredentialRef = asRecord(expected.credential_ref)
+  const expectedCredentialId = asString(expectedCredentialRef.id)
+  const expectedProvider = asString(expected.provider_id) || asString(expected.provider) || asString(expected.plugin_id)
+  const expectedToolName = asString(expected.tool_name) || asString(expected.name)
+
+  await expect.poll(
+    async () => {
+      const tools = await getOAuth2ToolEntries(getCurrentAgentId(world))
+      const matchingTool = tools.find((item) => {
+        const record = asRecord(item)
+        const provider = asString(record.provider_id) || asString(record.provider) || asString(record.plugin_id)
+        const toolName = asString(record.tool_name) || asString(record.name)
+
+        return provider === expectedProvider && toolName === expectedToolName
+      })
+      const record = asRecord(matchingTool)
+
+      return {
+        credentialId: asString(asRecord(record.credential_ref).id),
+        credentialType: asString(record.credential_type),
+      }
+    },
+    { timeout: 30_000 },
+  ).toEqual({
+    credentialId: expectedCredentialId,
+    credentialType: 'oauth2',
+  })
+}
+
+Given(
+  'an Agent v2 test agent with the OAuth2 tool credential fixture has been created via API',
+  async function (this: DifyWorld) {
+    const preseededAgent = getPreseededOAuthToolAgent(this)
+    const oauthTool = await getPreseededOAuthToolConfig(preseededAgent.id)
+    const agent = await createConfiguredTestAgent({
+      agentSoul: createAgentSoulConfigWithDifyTool(normalAgentSoulConfig, oauthTool),
+    })
+
+    this.createdAgentIds.push(agent.id)
+    this.lastCreatedAgentName = agent.name
+    this.lastCreatedAgentRole = agent.role ?? undefined
+  },
+)
+
+Given(
+  'a runnable Agent v2 test agent with the JSON Replace tool has been created via API',
+  async function (this: DifyWorld) {
+    if (!this.agentBuilder.preflight.stableModel)
+      throw new Error('Create a JSON Replace runtime Agent after stable model preflight.')
+
+    const agent = await createConfiguredTestAgent({
+      agentSoul: createAgentSoulConfigWithDifyTool(
+        createAgentSoulConfigWithModel(
+          {
+            ...normalAgentSoulConfig,
+            prompt: {
+              system_prompt: jsonReplaceRuntimePrompt,
+            },
+          },
+          this.agentBuilder.preflight.stableModel,
+        ),
+        jsonReplaceToolConfig(this),
+      ),
+    })
+
+    this.createdAgentIds.push(agent.id)
+    this.lastCreatedAgentName = agent.name
+    this.lastCreatedAgentRole = agent.role ?? undefined
+  },
+)
+
 When(
   'I add the Agent Builder JSON Replace tool from the Tools selector',
   async function (this: DifyWorld) {
@@ -83,14 +254,6 @@ When('I clear the Agent v2 tool selector search', async function (this: DifyWorl
   await search.fill('')
 })
 
-Given('Agent v2 JSON Replace runtime verification is available', async function (this: DifyWorld) {
-  return skipJsonReplaceRuntimeVerification(this)
-})
-
-Then('Agent v2 JSON Replace runtime verification should be available', async function (this: DifyWorld) {
-  return skipJsonReplaceRuntimeVerification(this)
-})
-
 Then(
   'the Agent v2 JSON Replace tool should be saved in the Agent v2 draft',
   async function (this: DifyWorld) {
@@ -109,6 +272,72 @@ Then(
   },
 )
 
+When('I send the Agent v2 Backend service API JSON Replace request', async function (this: DifyWorld) {
+  const serviceApiBaseURL = this.agentBuilder.accessPoint.serviceApiBaseURL
+  const apiKey = this.agentBuilder.accessPoint.generatedApiKey
+  if (!serviceApiBaseURL)
+    throw new Error('No Agent v2 service API endpoint found. Enable Backend service API first.')
+  if (!apiKey)
+    throw new Error('No Agent v2 API key found. Create a Backend service API key first.')
+
+  this.agentBuilder.accessPoint.serviceApiResponse = await sendAgentServiceApiChatMessage({
+    apiKey,
+    query: [
+      'Verify JSON Replace now.',
+      `Replace ${agentBuilderExpectedTokens.jsonToolBefore} with ${agentBuilderExpectedTokens.jsonToolAfter}.`,
+      `Return the JSON result and include ${agentBuilderExpectedTokens.jsonToolAfter}.`,
+    ].join(' '),
+    serviceApiBaseURL,
+  })
+})
+
+Then(
+  'the Agent v2 Backend service API response should include the JSON Replace E2E marker',
+  async function (this: DifyWorld) {
+    const response = this.agentBuilder.accessPoint.serviceApiResponse
+    if (!response)
+      throw new Error('No Agent v2 Backend service API response was recorded.')
+    if (!response.ok)
+      throw new Error(`Agent v2 Backend service API JSON Replace request failed with ${response.status}: ${JSON.stringify(response.body)}`)
+
+    const jsonReplaceThought = findJsonReplaceRuntimeThought(response.body)
+    if (!jsonReplaceThought) {
+      throw new Error(
+        [
+          'Agent v2 Backend service API did not emit a JSON Replace agent_thought with matching tool input and observation.',
+          `Received events: ${JSON.stringify(summarizeServiceApiRuntimeEvents(response.body))}`,
+        ].join('\n'),
+      )
+    }
+
+    const thought = getServiceApiEventData(jsonReplaceThought)
+    expect(asString(thought.tool_input)).toContain(agentBuilderExpectedTokens.jsonToolBefore)
+    expect(asString(thought.tool_input)).toContain('$.marker')
+    expect(asString(thought.observation)).toContain(agentBuilderExpectedTokens.jsonToolAfter)
+    expect(asString(asRecord(response.body).answer)).toContain(agentBuilderExpectedTokens.jsonToolAfter)
+  },
+)
+
+Then(
+  'I should see the Agent v2 OAuth2 tool authorized in the Tools section',
+  async function (this: DifyWorld) {
+    const toolsSection = getToolsSection(this)
+    const displayName = await getOAuth2ToolDisplayName(this)
+
+    await expectProviderToolActionVisible(toolsSection, displayName)
+    await expect(toolsSection.getByRole('button', { exact: true, name: 'Not authorized' }))
+      .not
+      .toBeVisible()
+  },
+)
+
+Then(
+  'the Agent v2 OAuth2 tool credential should remain saved in the Agent v2 draft',
+  async function (this: DifyWorld) {
+    await expectOAuth2CredentialPreserved(this)
+  },
+)
+
 Then('I should see the Agent v2 tool selector empty state', async function (this: DifyWorld) {
   const page = this.getPage()
 
diff --git a/e2e/features/step-definitions/apps/share-app.steps.ts b/e2e/features/step-definitions/apps/share-app.steps.ts
index 20d9c05522e..eacc49f4ead 100644
--- a/e2e/features/step-definitions/apps/share-app.steps.ts
+++ b/e2e/features/step-definitions/apps/share-app.steps.ts
@@ -9,6 +9,8 @@ import {
 } from '../../../support/api'
 import { createE2EResourceName } from '../../../support/naming'
 
+const escapeRegExp = (value: string) => value.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')
+
 When('I enable the Web App share', async function (this: DifyWorld) {
   const page = this.getPage()
   const appName = this.lastCreatedAppName
@@ -18,7 +20,7 @@ When('I enable the Web App share', async function (this: DifyWorld) {
     )
   }
 
-  await page.locator('button').filter({ hasText: appName }).filter({ hasText: 'Workflow' }).click()
+  await page.getByRole('button', { name: new RegExp(escapeRegExp(appName)) }).click()
   await expect(page.getByRole('switch').first()).toBeEnabled({ timeout: 15_000 })
   await page.getByRole('switch').first().click()
 })
@@ -47,7 +49,7 @@ When('I open the shared app URL', async function (this: DifyWorld) {
 
 Then('the shared app page should be accessible', async function (this: DifyWorld) {
   await expect(this.getPage()).toHaveURL(/\/(workflow|chat)\/[a-zA-Z0-9]+/, { timeout: 15_000 })
-  await expect(this.getPage().locator('body')).toBeVisible({ timeout: 10_000 })
+  await expect(this.getPage().getByRole('button', { name: 'Execute' })).toBeVisible({ timeout: 10_000 })
 })
 
 When('I run the shared workflow app', async function (this: DifyWorld) {
@@ -59,5 +61,5 @@ When('I run the shared workflow app', async function (this: DifyWorld) {
 })
 
 Then('the shared workflow run should succeed', async function (this: DifyWorld) {
-  await expect(this.getPage().getByTestId('status-icon-success')).toBeVisible({ timeout: 55_000 })
+  await expect(this.getPage().getByRole('img', { name: 'Workflow Process succeeded' })).toBeVisible({ timeout: 55_000 })
 })
diff --git a/e2e/features/support/world.ts b/e2e/features/support/world.ts
index 284e8806f75..689275c9ccf 100644
--- a/e2e/features/support/world.ts
+++ b/e2e/features/support/world.ts
@@ -22,7 +22,7 @@ export type CreatedBuiltinToolCredential = {
   credentialId: string
   provider: string
 }
-export type AgentBuilderStableChatModel = {
+export type AgentBuilderChatModel = {
   name: string
   provider: string
   type: string
@@ -39,9 +39,10 @@ export type AgentV2WorkflowOutputVariable = {
 
 export const createAgentBuilderWorldState = () => ({
   preflight: {
-    brokenModel: undefined as AgentBuilderStableChatModel | undefined,
+    agentDecisionModel: undefined as AgentBuilderChatModel | undefined,
+    brokenModel: undefined as AgentBuilderChatModel | undefined,
     preseededResources: {} as Record<string, AgentBuilderPreseededResource>,
-    stableModel: undefined as AgentBuilderStableChatModel | undefined,
+    stableModel: undefined as AgentBuilderChatModel | undefined,
   },
   accessPoint: {
     apiReferencePage: undefined as Page | undefined,
diff --git a/e2e/fixtures/test-materials/agent-build-instruction.txt b/e2e/fixtures/test-materials/agent-build-instruction.txt
index d5cf5f83e5e..82ebf01d3a7 100644
--- a/e2e/fixtures/test-materials/agent-build-instruction.txt
+++ b/e2e/fixtures/test-materials/agent-build-instruction.txt
@@ -1,3 +1,3 @@
-Ask the Agent to use e2e-summary-skill to summarize user input.
-When replacing JSON strings, use JSON Process / JSON Replace.
-After applying, include these capabilities in the Agent instructions.
+Update only the Agent instructions.
+After applying, every response should briefly mention E2E_BUILD_DRAFT_PASS.
+Do not add files, skills, tools, knowledge, environment variables, or other capabilities.
diff --git a/e2e/fixtures/test-materials/agent-knowledge-source.txt b/e2e/fixtures/test-materials/agent-knowledge-source.txt
index 2275246ac5b..fd5f4299175 100644
--- a/e2e/fixtures/test-materials/agent-knowledge-source.txt
+++ b/e2e/fixtures/test-materials/agent-knowledge-source.txt
@@ -1,2 +1,3 @@
 Dify Agent E2E knowledge source.
-Dify Agent E2E test passphrase is AGENT_KNOWLEDGE_PASS.
+The Dify Agent E2E knowledge marker is AGENT_KNOWLEDGE_PASS.
+Use the connected knowledge source to find the Dify Agent E2E knowledge marker.
diff --git a/e2e/package.json b/e2e/package.json
index 9210fc6a81b..c0804956f31 100644
--- a/e2e/package.json
+++ b/e2e/package.json
@@ -4,6 +4,8 @@
   "private": true,
   "scripts": {
     "e2e": "tsx ./scripts/run-cucumber.ts",
+    "e2e:external": "tsx ./scripts/run-external-runtime.ts",
+    "e2e:external:prepare": "tsx ./scripts/prepare-external-runtime.ts",
     "e2e:full": "tsx ./scripts/run-cucumber.ts --full",
     "e2e:full:headed": "tsx ./scripts/run-cucumber.ts --full --headed",
     "e2e:headed": "tsx ./scripts/run-cucumber.ts --headed",
@@ -11,6 +13,7 @@
     "e2e:middleware:down": "tsx ./scripts/setup.ts middleware-down",
     "e2e:middleware:up": "tsx ./scripts/setup.ts middleware-up",
     "e2e:reset": "tsx ./scripts/setup.ts reset",
+    "seed": "tsx ./scripts/seed.ts",
     "type-check": "tsgo"
   },
   "devDependencies": {
@@ -18,11 +21,13 @@
     "@dify/contracts": "workspace:*",
     "@dify/tsconfig": "workspace:*",
     "@playwright/test": "catalog:",
+    "@t3-oss/env-core": "catalog:",
     "@types/node": "catalog:",
     "@typescript/native-preview": "catalog:",
     "tsx": "catalog:",
     "typescript": "catalog:",
     "vite": "catalog:",
-    "vite-plus": "catalog:"
+    "vite-plus": "catalog:",
+    "zod": "catalog:"
   }
 }
diff --git a/e2e/scripts/common.ts b/e2e/scripts/common.ts
index 5e5edca70ce..be507b62abf 100644
--- a/e2e/scripts/common.ts
+++ b/e2e/scripts/common.ts
@@ -32,6 +32,7 @@ type ForegroundProcessOptions = {
 export const rootDir = fileURLToPath(new URL('../..', import.meta.url))
 export const e2eDir = path.join(rootDir, 'e2e')
 export const apiDir = path.join(rootDir, 'api')
+export const difyAgentDir = path.join(rootDir, 'dify-agent')
 export const dockerDir = path.join(rootDir, 'docker')
 export const webDir = path.join(rootDir, 'web')
 
@@ -207,7 +208,7 @@ export const getWebEnvLocalHash = async () => {
     .digest('hex')
 }
 
-export const readSimpleDotenv = async (filePath: string) => {
+export const readSimpleDotenv = async (filePath: string): Promise<Record<string, string>> => {
   const fileContent = await readFile(filePath, 'utf8')
   const entries = fileContent
     .split(/\r?\n/)
diff --git a/e2e/scripts/env-register.ts b/e2e/scripts/env-register.ts
new file mode 100644
index 00000000000..b4e69651d8b
--- /dev/null
+++ b/e2e/scripts/env-register.ts
@@ -0,0 +1,5 @@
+import { loadE2eEnv, validateE2eEnv } from './env'
+
+// Entry points import this before test-env.ts so e2e/.env.local can affect base URLs and runner flags.
+loadE2eEnv()
+validateE2eEnv()
diff --git a/e2e/scripts/env.ts b/e2e/scripts/env.ts
new file mode 100644
index 00000000000..381eada1ecc
--- /dev/null
+++ b/e2e/scripts/env.ts
@@ -0,0 +1,150 @@
+import { existsSync, readFileSync } from 'node:fs'
+import path from 'node:path'
+import { createEnv } from '@t3-oss/env-core'
+import * as z from 'zod'
+import { e2eDir } from './common'
+
+const booleanString = z.enum(['0', '1', 'false', 'true'])
+const jsonObjectString = z.string().refine((value) => {
+  try {
+    const parsed = JSON.parse(value) as unknown
+    return typeof parsed === 'object' && parsed !== null && !Array.isArray(parsed)
+  }
+  catch {
+    return false
+  }
+}, 'must be a JSON object string')
+
+const parseEnvLine = (line: string) => {
+  const trimmed = line.trim()
+  if (!trimmed || trimmed.startsWith('#'))
+    return undefined
+
+  const normalized = trimmed.startsWith('export ')
+    ? trimmed.slice('export '.length).trim()
+    : trimmed
+  const separatorIndex = normalized.indexOf('=')
+  if (separatorIndex === -1)
+    return undefined
+
+  const key = normalized.slice(0, separatorIndex).trim()
+  let value = normalized.slice(separatorIndex + 1).trim()
+  if (!key)
+    return undefined
+
+  if (
+    (value.startsWith('"') && value.endsWith('"'))
+    || (value.startsWith('\'') && value.endsWith('\''))
+  ) {
+    value = value.slice(1, -1)
+  }
+
+  return [key, value] as const
+}
+
+export const loadE2eEnv = () => {
+  const envFilePath = process.env.E2E_ENV_FILE
+    ? path.resolve(process.env.E2E_ENV_FILE)
+    : path.join(e2eDir, '.env.local')
+
+  if (!existsSync(envFilePath))
+    return
+
+  const content = readFileSync(envFilePath, 'utf8')
+  for (const line of content.split(/\r?\n/)) {
+    const entry = parseEnvLine(line)
+    if (!entry)
+      continue
+
+    const [key, value] = entry
+    process.env[key] ??= value
+  }
+}
+
+export const validateE2eEnv = () => createEnv({
+  client: {},
+  clientPrefix: 'NEXT_PUBLIC_',
+  emptyStringAsUndefined: true,
+  onValidationError: (issues) => {
+    const messages = issues.map((issue) => {
+      const path = Array.isArray(issue.path) && issue.path.length > 0
+        ? issue.path.join('.')
+        : '(root)'
+      return `- ${path}: ${issue.message}`
+    })
+
+    throw new Error(`Invalid E2E env:\n${messages.join('\n')}`)
+  },
+  runtimeEnv: {
+    CUCUMBER_HEADLESS: process.env.CUCUMBER_HEADLESS,
+    E2E_ADMIN_EMAIL: process.env.E2E_ADMIN_EMAIL,
+    E2E_ADMIN_NAME: process.env.E2E_ADMIN_NAME,
+    E2E_ADMIN_PASSWORD: process.env.E2E_ADMIN_PASSWORD,
+    E2E_AGENT_BACKEND_PORT: process.env.E2E_AGENT_BACKEND_PORT,
+    E2E_AGENT_BACKEND_URL: process.env.E2E_AGENT_BACKEND_URL,
+    E2E_API_URL: process.env.E2E_API_URL,
+    E2E_BASE_URL: process.env.E2E_BASE_URL,
+    E2E_BROKEN_MODEL_NAME: process.env.E2E_BROKEN_MODEL_NAME,
+    E2E_BROKEN_MODEL_PROVIDER: process.env.E2E_BROKEN_MODEL_PROVIDER,
+    E2E_BROKEN_MODEL_TYPE: process.env.E2E_BROKEN_MODEL_TYPE,
+    E2E_CUCUMBER_TAGS: process.env.E2E_CUCUMBER_TAGS,
+    E2E_EXTERNAL_RUNTIME_SEED_SPECS: process.env.E2E_EXTERNAL_RUNTIME_SEED_SPECS,
+    E2E_EXTERNAL_RUNTIME_TAGS: process.env.E2E_EXTERNAL_RUNTIME_TAGS,
+    E2E_FORCE_WEB_BUILD: process.env.E2E_FORCE_WEB_BUILD,
+    E2E_INIT_PASSWORD: process.env.E2E_INIT_PASSWORD,
+    E2E_MARKETPLACE_API_URL: process.env.E2E_MARKETPLACE_API_URL,
+    E2E_MARKETPLACE_PLUGIN_IDS: process.env.E2E_MARKETPLACE_PLUGIN_IDS,
+    E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS: process.env.E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS,
+    E2E_MODEL_PROVIDER_CREDENTIALS_JSON: process.env.E2E_MODEL_PROVIDER_CREDENTIALS_JSON,
+    E2E_OAUTH_TOOL_CREDENTIAL_ID: process.env.E2E_OAUTH_TOOL_CREDENTIAL_ID,
+    E2E_OAUTH_TOOL_NAME: process.env.E2E_OAUTH_TOOL_NAME,
+    E2E_OAUTH_TOOL_PROVIDER: process.env.E2E_OAUTH_TOOL_PROVIDER,
+    E2E_REUSE_WEB_SERVER: process.env.E2E_REUSE_WEB_SERVER,
+    E2E_SLOW_MO: process.env.E2E_SLOW_MO,
+    E2E_SHELLCTL_AUTH_TOKEN: process.env.E2E_SHELLCTL_AUTH_TOKEN,
+    E2E_SHELLCTL_CONTAINER_NAME: process.env.E2E_SHELLCTL_CONTAINER_NAME,
+    E2E_SHELLCTL_IMAGE: process.env.E2E_SHELLCTL_IMAGE,
+    E2E_SHELLCTL_PORT: process.env.E2E_SHELLCTL_PORT,
+    E2E_SHELLCTL_URL: process.env.E2E_SHELLCTL_URL,
+    E2E_START_AGENT_BACKEND: process.env.E2E_START_AGENT_BACKEND,
+    E2E_STABLE_MODEL_NAME: process.env.E2E_STABLE_MODEL_NAME,
+    E2E_STABLE_MODEL_PROVIDER: process.env.E2E_STABLE_MODEL_PROVIDER,
+    E2E_STABLE_MODEL_TYPE: process.env.E2E_STABLE_MODEL_TYPE,
+  },
+  server: {
+    CUCUMBER_HEADLESS: booleanString.optional(),
+    E2E_ADMIN_EMAIL: z.email().optional(),
+    E2E_ADMIN_NAME: z.string().min(1).optional(),
+    E2E_ADMIN_PASSWORD: z.string().min(1).optional(),
+    E2E_AGENT_BACKEND_PORT: z.coerce.number().int().positive().max(65535).optional(),
+    E2E_AGENT_BACKEND_URL: z.url().optional(),
+    E2E_API_URL: z.url().optional(),
+    E2E_BASE_URL: z.url().optional(),
+    E2E_BROKEN_MODEL_NAME: z.string().min(1).optional(),
+    E2E_BROKEN_MODEL_PROVIDER: z.string().min(1).optional(),
+    E2E_BROKEN_MODEL_TYPE: z.string().min(1).optional(),
+    E2E_CUCUMBER_TAGS: z.string().min(1).optional(),
+    E2E_EXTERNAL_RUNTIME_SEED_SPECS: z.string().min(1).optional(),
+    E2E_EXTERNAL_RUNTIME_TAGS: z.string().min(1).optional(),
+    E2E_FORCE_WEB_BUILD: z.literal('1').optional(),
+    E2E_INIT_PASSWORD: z.string().min(1).optional(),
+    E2E_MARKETPLACE_API_URL: z.url().optional(),
+    E2E_MARKETPLACE_PLUGIN_IDS: z.string().min(1).optional(),
+    E2E_MARKETPLACE_PLUGIN_UNIQUE_IDENTIFIERS: z.string().min(1).optional(),
+    E2E_MODEL_PROVIDER_CREDENTIALS_JSON: jsonObjectString.optional(),
+    E2E_OAUTH_TOOL_CREDENTIAL_ID: z.string().min(1).optional(),
+    E2E_OAUTH_TOOL_NAME: z.string().min(1).optional(),
+    E2E_OAUTH_TOOL_PROVIDER: z.string().min(1).optional(),
+    E2E_REUSE_WEB_SERVER: booleanString.optional(),
+    E2E_SLOW_MO: z.coerce.number().nonnegative().optional(),
+    E2E_SHELLCTL_AUTH_TOKEN: z.string().min(1).optional(),
+    E2E_SHELLCTL_CONTAINER_NAME: z.string().min(1).optional(),
+    E2E_SHELLCTL_IMAGE: z.string().min(1).optional(),
+    E2E_SHELLCTL_PORT: z.coerce.number().int().positive().max(65535).optional(),
+    E2E_SHELLCTL_URL: z.url().optional(),
+    E2E_START_AGENT_BACKEND: booleanString.optional(),
+    E2E_STABLE_MODEL_NAME: z.string().min(1).optional(),
+    E2E_STABLE_MODEL_PROVIDER: z.string().min(1).optional(),
+    E2E_STABLE_MODEL_TYPE: z.string().min(1).optional(),
+  },
+})
diff --git a/e2e/scripts/prepare-external-runtime.ts b/e2e/scripts/prepare-external-runtime.ts
new file mode 100644
index 00000000000..e1800f3de41
--- /dev/null
+++ b/e2e/scripts/prepare-external-runtime.ts
@@ -0,0 +1,38 @@
+import { e2eDir, isMainModule, runCommandOrThrow } from './common'
+import './env-register'
+
+const defaultExternalRuntimeSeedSpecs = 'agent-v2:external-runtime'
+
+const parseSeedSpecs = (value: string) =>
+  value
+    .split(',')
+    .map(entry => entry.trim())
+    .filter(Boolean)
+    .map((entry) => {
+      const [pack, profile = 'full'] = entry.split(':').map(part => part.trim())
+      if (!pack)
+        throw new Error(`Invalid external runtime seed spec "${entry}". Expected "pack" or "pack:profile".`)
+
+      return { pack, profile }
+    })
+
+const main = async () => {
+  const seedSpecs = parseSeedSpecs(
+    process.env.E2E_EXTERNAL_RUNTIME_SEED_SPECS?.trim() || defaultExternalRuntimeSeedSpecs,
+  )
+
+  for (const { pack, profile } of seedSpecs) {
+    await runCommandOrThrow({
+      command: 'npx',
+      args: ['tsx', './scripts/seed.ts', '--pack', pack, '--profile', profile],
+      cwd: e2eDir,
+    })
+  }
+}
+
+if (isMainModule(import.meta.url)) {
+  void main().catch((error) => {
+    console.error(error instanceof Error ? error.message : String(error))
+    process.exit(1)
+  })
+}
diff --git a/e2e/scripts/run-cucumber.ts b/e2e/scripts/run-cucumber.ts
index 5913450f14b..70b0c34902a 100644
--- a/e2e/scripts/run-cucumber.ts
+++ b/e2e/scripts/run-cucumber.ts
@@ -6,6 +6,7 @@ import { startWebServer, stopWebServer } from '../support/web-server'
 import { apiURL, baseURL, reuseExistingWebServer } from '../test-env'
 import { e2eDir, isMainModule, runCommand } from './common'
 import { resetState, startMiddleware, stopMiddleware } from './setup'
+import './env-register'
 
 type RunOptions = {
   forwardArgs: string[]
@@ -43,6 +44,20 @@ const parseArgs = (argv: string[]): RunOptions => {
 const hasCustomTags = (forwardArgs: string[]) =>
   forwardArgs.some(arg => arg === '--tags' || arg.startsWith('--tags='))
 
+const fullNonExternalTags = 'not @skip and not @preview and not @external-model and not @external-tool'
+
+const isTruthyEnv = (value: string | undefined) => value === '1' || value === 'true'
+
+const shouldStartAgentBackend = () => {
+  if (isTruthyEnv(process.env.E2E_START_AGENT_BACKEND))
+    return true
+
+  if (process.env.E2E_AGENT_BACKEND_URL || process.env.AGENT_BACKEND_BASE_URL)
+    return false
+
+  return false
+}
+
 const readLogTail = async (logFilePath: string) => {
   const content = await readFile(logFilePath, 'utf8').catch(() => '')
 
@@ -85,6 +100,7 @@ const main = async () => {
   const { forwardArgs, full, headed } = parseArgs(process.argv.slice(2))
   const startMiddlewareForRun = full
   const resetStateForRun = full
+  const startAgentBackendForRun = shouldStartAgentBackend()
 
   if (resetStateForRun)
     await resetState()
@@ -98,10 +114,38 @@ const main = async () => {
   await rm(cucumberReportDir, { force: true, recursive: true })
   await mkdir(logDir, { recursive: true })
 
+  const shellctlProcess = startAgentBackendForRun
+    ? await startLoggedProcess({
+        command: 'npx',
+        args: ['tsx', './scripts/setup.ts', 'shellctl-sandbox'],
+        cwd: e2eDir,
+        label: 'shellctl sandbox',
+        logFilePath: path.join(logDir, 'cucumber-shellctl-sandbox.log'),
+      })
+    : undefined
+
+  const difyAgentProcess = startAgentBackendForRun
+    ? await startLoggedProcess({
+        command: 'npx',
+        args: ['tsx', './scripts/setup.ts', 'agent-backend'],
+        cwd: e2eDir,
+        env: {
+          E2E_START_AGENT_BACKEND: '1',
+        },
+        label: 'agent backend',
+        logFilePath: path.join(logDir, 'cucumber-agent-backend.log'),
+      })
+    : undefined
+
   const apiProcess = await startLoggedProcess({
     command: 'npx',
     args: ['tsx', './scripts/setup.ts', 'api'],
     cwd: e2eDir,
+    env: startAgentBackendForRun
+      ? {
+          E2E_START_AGENT_BACKEND: '1',
+        }
+      : undefined,
     label: 'api server',
     logFilePath: path.join(logDir, 'cucumber-api.log'),
   })
@@ -121,6 +165,8 @@ const main = async () => {
         await stopWebServer()
         await stopManagedProcess(celeryProcess)
         await stopManagedProcess(apiProcess)
+        await stopManagedProcess(difyAgentProcess)
+        await stopManagedProcess(shellctlProcess)
 
         if (startMiddlewareForRun) {
           try {
@@ -146,6 +192,50 @@ const main = async () => {
   process.once('SIGTERM', onTerminate)
 
   try {
+    if (shellctlProcess) {
+      let waitingForShellctl = true
+      try {
+        const shellctlPort = process.env.E2E_SHELLCTL_PORT || '5004'
+        await Promise.race([
+          waitForUrl(`http://127.0.0.1:${shellctlPort}/openapi.json`, 180_000, 1_000),
+          waitForUnexpectedProcessExit(shellctlProcess, () => !waitingForShellctl),
+        ])
+      }
+      catch (error) {
+        if (error instanceof Error && error.message.includes('exited before becoming ready'))
+          throw error
+
+        throw new Error(
+          `Shellctl sandbox did not become ready. See ${shellctlProcess.logFilePath}.`,
+        )
+      }
+      finally {
+        waitingForShellctl = false
+      }
+    }
+
+    if (difyAgentProcess) {
+      let waitingForAgentBackend = true
+      try {
+        const agentBackendPort = process.env.E2E_AGENT_BACKEND_PORT || '5050'
+        await Promise.race([
+          waitForUrl(`http://127.0.0.1:${agentBackendPort}/openapi.json`, 180_000, 1_000),
+          waitForUnexpectedProcessExit(difyAgentProcess, () => !waitingForAgentBackend),
+        ])
+      }
+      catch (error) {
+        if (error instanceof Error && error.message.includes('exited before becoming ready'))
+          throw error
+
+        throw new Error(
+          `Agent backend did not become ready. See ${difyAgentProcess.logFilePath}.`,
+        )
+      }
+      finally {
+        waitingForAgentBackend = false
+      }
+    }
+
     let waitingForApi = true
     try {
       await Promise.race([
@@ -179,7 +269,7 @@ const main = async () => {
     }
 
     if (startMiddlewareForRun && !hasCustomTags(forwardArgs))
-      cucumberEnv.E2E_CUCUMBER_TAGS = 'not @skip and not @preview'
+      cucumberEnv.E2E_CUCUMBER_TAGS = fullNonExternalTags
 
     const result = await runCommand({
       command: 'npx',
diff --git a/e2e/scripts/run-external-runtime.ts b/e2e/scripts/run-external-runtime.ts
new file mode 100644
index 00000000000..d5ea4a5ecb4
--- /dev/null
+++ b/e2e/scripts/run-external-runtime.ts
@@ -0,0 +1,17 @@
+import { e2eDir, isMainModule, runForegroundProcess } from './common'
+import './env-register'
+
+const defaultExternalRuntimeTags = '(@external-model or @external-tool) and not @feature-gated and not @skip and not @preview'
+
+const main = async () => {
+  const tags = process.env.E2E_EXTERNAL_RUNTIME_TAGS?.trim() || defaultExternalRuntimeTags
+
+  await runForegroundProcess({
+    command: 'npx',
+    args: ['tsx', './scripts/run-cucumber.ts', '--', '--tags', tags],
+    cwd: e2eDir,
+  })
+}
+
+if (isMainModule(import.meta.url))
+  void main()
diff --git a/e2e/scripts/seed.ts b/e2e/scripts/seed.ts
new file mode 100644
index 00000000000..82d5dddf22c
--- /dev/null
+++ b/e2e/scripts/seed.ts
@@ -0,0 +1,167 @@
+import type { ManagedProcess } from '../support/process'
+import { mkdir } from 'node:fs/promises'
+import path from 'node:path'
+import { chromium } from '@playwright/test'
+import { createAgentV2SeedTasks } from '../features/agent-v2/support/seed'
+import { ensureAuthenticatedState } from '../fixtures/auth'
+import { startLoggedProcess, stopManagedProcess, waitForUrl } from '../support/process'
+import { runSeedTasks, writeSeedReport } from '../support/seed'
+import { startWebServer, stopWebServer } from '../support/web-server'
+import { apiURL, baseURL, reuseExistingWebServer } from '../test-env'
+import { e2eDir, isMainModule } from './common'
+import './env-register'
+
+type SeedOptions = {
+  allowBlocked: boolean
+  dryRun: boolean
+  pack: string
+  profile: string
+}
+
+const parseArgs = (argv: string[]): SeedOptions => {
+  const options: SeedOptions = {
+    allowBlocked: false,
+    dryRun: false,
+    pack: 'agent-v2',
+    profile: 'full',
+  }
+
+  for (const [index, arg] of argv.entries()) {
+    if (arg === '--pack') {
+      options.pack = argv[index + 1] || options.pack
+      continue
+    }
+    if (arg.startsWith('--pack=')) {
+      options.pack = arg.slice('--pack='.length)
+      continue
+    }
+    if (arg === '--dry-run')
+      options.dryRun = true
+    if (arg === '--allow-blocked')
+      options.allowBlocked = true
+    if (arg === '--profile') {
+      options.profile = argv[index + 1] || options.profile
+      continue
+    }
+    if (arg.startsWith('--profile=')) {
+      options.profile = arg.slice('--profile='.length)
+      continue
+    }
+  }
+
+  return options
+}
+
+const getTasks = (pack: string, profile: string) => {
+  if (pack === 'agent-v2')
+    return createAgentV2SeedTasks(profile)
+
+  throw new Error(`Unknown seed pack "${pack}".`)
+}
+
+const ensureAuth = async () => {
+  const browser = await chromium.launch({ headless: true })
+  try {
+    await ensureAuthenticatedState(browser, baseURL)
+  }
+  finally {
+    await browser.close()
+  }
+}
+
+const startApiProcess = async (logDir: string) => {
+  try {
+    await waitForUrl(`${apiURL}/health`, 1_000, 250, 1_000)
+    return undefined
+  }
+  catch {
+    // Start a local API process below.
+  }
+
+  const apiProcess = await startLoggedProcess({
+    command: 'npx',
+    args: ['tsx', './scripts/setup.ts', 'api'],
+    cwd: e2eDir,
+    label: 'api server',
+    logFilePath: path.join(logDir, 'seed-api.log'),
+  })
+
+  try {
+    await waitForUrl(`${apiURL}/health`, 180_000, 1_000)
+    return apiProcess
+  }
+  catch (error) {
+    await stopManagedProcess(apiProcess)
+    throw error
+  }
+}
+
+const startCeleryProcess = async (logDir: string) =>
+  startLoggedProcess({
+    command: 'npx',
+    args: [
+      'tsx',
+      './scripts/setup.ts',
+      'celery',
+      '--queues',
+      'dataset,priority_dataset,workflow_based_app_execution',
+    ],
+    cwd: e2eDir,
+    label: 'celery worker',
+    logFilePath: path.join(logDir, 'seed-celery.log'),
+  })
+
+const main = async () => {
+  const options = parseArgs(process.argv.slice(2))
+  const logDir = path.join(e2eDir, '.logs')
+  let apiProcess: ManagedProcess | undefined
+  let celeryProcess: ManagedProcess | undefined
+
+  await mkdir(logDir, { recursive: true })
+
+  try {
+    apiProcess = await startApiProcess(logDir)
+    celeryProcess = await startCeleryProcess(logDir)
+    await startWebServer({
+      baseURL,
+      command: 'npx',
+      args: ['tsx', './scripts/setup.ts', 'web'],
+      cwd: e2eDir,
+      logFilePath: path.join(logDir, 'seed-web.log'),
+      reuseExistingServer: reuseExistingWebServer,
+      timeoutMs: 300_000,
+    })
+
+    console.warn(`[seed] bootstrapping auth state against ${baseURL}`)
+    await ensureAuth()
+
+    const results = await runSeedTasks(getTasks(options.pack, options.profile), {
+      dryRun: options.dryRun,
+      resources: new Map(),
+    })
+    const reportName = options.profile === 'full'
+      ? options.pack
+      : `${options.pack}-${options.profile}`
+    const reportPath = await writeSeedReport(reportName, results)
+    const blockedCount = results.filter(result => result.status === 'blocked').length
+
+    console.warn(`[seed] report ${reportPath}`)
+    if (blockedCount > 0 && !options.allowBlocked) {
+      throw new Error(
+        `${blockedCount} seed task${blockedCount === 1 ? '' : 's'} blocked. Re-run with --allow-blocked only when partial readiness is intentional.`,
+      )
+    }
+  }
+  finally {
+    await stopWebServer()
+    await stopManagedProcess(celeryProcess)
+    await stopManagedProcess(apiProcess)
+  }
+}
+
+if (isMainModule(import.meta.url)) {
+  void main().catch((error) => {
+    console.error(error instanceof Error ? error.message : String(error))
+    process.exit(1)
+  })
+}
diff --git a/e2e/scripts/setup.ts b/e2e/scripts/setup.ts
index b1f623f923b..0654772142f 100644
--- a/e2e/scripts/setup.ts
+++ b/e2e/scripts/setup.ts
@@ -5,6 +5,7 @@ import { waitForUrl } from '../support/process'
 import {
   apiDir,
   apiEnvExampleFile,
+  difyAgentDir,
   dockerDir,
   e2eDir,
   e2eWebEnvOverrides,
@@ -30,6 +31,15 @@ const buildIdPath = path.join(webDir, '.next', 'BUILD_ID')
 const webBuildStampPath = path.join(webDir, '.next', 'e2e-web-build.sha256')
 const apiHost = '127.0.0.1'
 const apiPort = 5001
+const agentBackendHost = '127.0.0.1'
+const agentBackendPort = Number(process.env.E2E_AGENT_BACKEND_PORT || 5050)
+const shellctlHost = '127.0.0.1'
+const shellctlPort = Number(process.env.E2E_SHELLCTL_PORT || 5004)
+const shellctlContainerName = process.env.E2E_SHELLCTL_CONTAINER_NAME || 'dify-agent-e2e-shellctl'
+const shellctlImage = process.env.E2E_SHELLCTL_IMAGE || 'dify-agent-local-sandbox:e2e'
+const shellctlUrl = `http://${shellctlHost}:${shellctlPort}`
+const defaultPluginDaemonKey = 'lYkiYYT6owG+71oLerGzA7GXCgOT++6ovaezWAjpCjf+Sjc3ZtU+qUEi'
+const defaultInnerApiKeyForPlugin = 'QaHbTe77CtuXmsfyhR7+vRjI/+XbV1AaFy691iy+kGDv2Jvy0/eAh8Y1'
 
 const middlewareDataPaths = [
   path.join(dockerDir, 'volumes', 'db', 'data'),
@@ -56,15 +66,68 @@ const composeArgs = [
   'weaviate',
 ]
 
-const getApiEnvironment = async () => {
+const getApiEnvironment = async (): Promise<Record<string, string>> => {
   const envFromExample = await readSimpleDotenv(apiEnvExampleFile)
+  const agentBackendBaseUrl = getAgentBackendBaseUrl()
 
   return {
     ...envFromExample,
+    ...(agentBackendBaseUrl ? { AGENT_BACKEND_BASE_URL: agentBackendBaseUrl } : {}),
     FLASK_APP: 'app.py',
   }
 }
 
+function getAgentBackendBaseUrl() {
+  const explicitApiUrl = process.env.AGENT_BACKEND_BASE_URL?.trim()
+  if (explicitApiUrl)
+    return explicitApiUrl
+
+  const explicitE2EUrl = process.env.E2E_AGENT_BACKEND_URL?.trim()
+  if (explicitE2EUrl)
+    return explicitE2EUrl.replace(/\/$/, '')
+
+  if (process.env.E2E_START_AGENT_BACKEND === '1' || process.env.E2E_START_AGENT_BACKEND === 'true')
+    return `http://${agentBackendHost}:${agentBackendPort}`
+
+  return undefined
+}
+
+const getAgentBackendEnvironment = async () => {
+  const apiEnv = await getApiEnvironment()
+  const redisPassword = process.env.REDIS_PASSWORD || apiEnv.REDIS_PASSWORD || 'difyai123456'
+
+  return {
+    DIFY_AGENT_INNER_API_KEY:
+      process.env.DIFY_AGENT_INNER_API_KEY
+      || process.env.INNER_API_KEY_FOR_PLUGIN
+      || process.env.PLUGIN_DIFY_INNER_API_KEY
+      || apiEnv.INNER_API_KEY_FOR_PLUGIN
+      || defaultInnerApiKeyForPlugin,
+    DIFY_AGENT_INNER_API_URL: process.env.DIFY_AGENT_INNER_API_URL || `http://${apiHost}:${apiPort}`,
+    DIFY_AGENT_PLUGIN_DAEMON_API_KEY:
+      process.env.DIFY_AGENT_PLUGIN_DAEMON_API_KEY
+      || process.env.PLUGIN_DAEMON_KEY
+      || apiEnv.PLUGIN_DAEMON_KEY
+      || defaultPluginDaemonKey,
+    DIFY_AGENT_PLUGIN_DAEMON_URL:
+      process.env.DIFY_AGENT_PLUGIN_DAEMON_URL
+      || process.env.PLUGIN_DAEMON_URL
+      || 'http://127.0.0.1:5002',
+    DIFY_AGENT_REDIS_PREFIX: process.env.DIFY_AGENT_REDIS_PREFIX || 'dify-agent-e2e',
+    DIFY_AGENT_REDIS_URL:
+      process.env.DIFY_AGENT_REDIS_URL
+      || `redis://:${redisPassword}@127.0.0.1:6379/0`,
+    DIFY_AGENT_SHELLCTL_AUTH_TOKEN:
+      process.env.DIFY_AGENT_SHELLCTL_AUTH_TOKEN
+      || process.env.E2E_SHELLCTL_AUTH_TOKEN
+      || '',
+    DIFY_AGENT_SHELLCTL_ENTRYPOINT:
+      process.env.DIFY_AGENT_SHELLCTL_ENTRYPOINT
+      || process.env.E2E_SHELLCTL_URL
+      || shellctlUrl,
+  }
+}
+
 const getServiceContainerId = async (service: string) => {
   const result = await runCommandOrThrow({
     command: 'docker',
@@ -272,7 +335,103 @@ export const startApi = async () => {
   })
 }
 
-export const startCelery = async () => {
+export const startAgentBackend = async () => {
+  if (await isTcpPortReachable(agentBackendHost, agentBackendPort)) {
+    const listenerDescription = await getTcpPortListenerDescription(agentBackendPort)
+    const listenerMessage = listenerDescription
+      ? `\n\nPort listener:\n${listenerDescription}`
+      : ''
+
+    throw new Error(
+      `Cannot start the E2E Agent backend because ${agentBackendHost}:${agentBackendPort} is already in use.${listenerMessage}`,
+    )
+  }
+
+  await runForegroundProcess({
+    command: 'uv',
+    args: [
+      'run',
+      '--project',
+      '.',
+      '--extra',
+      'server',
+      'uvicorn',
+      'dify_agent.server.app:app',
+      '--host',
+      agentBackendHost,
+      '--port',
+      String(agentBackendPort),
+    ],
+    cwd: difyAgentDir,
+    env: await getAgentBackendEnvironment(),
+  })
+}
+
+const ensureShellctlSandboxImage = async () => {
+  const inspectResult = await runCommand({
+    command: 'docker',
+    args: ['image', 'inspect', shellctlImage],
+    cwd: rootDir,
+    stdio: 'pipe',
+  })
+
+  if (inspectResult.exitCode === 0 && process.env.E2E_FORCE_SHELLCTL_BUILD !== '1')
+    return
+
+  await runCommandOrThrow({
+    command: 'docker',
+    args: [
+      'build',
+      '-f',
+      path.join(difyAgentDir, 'docker', 'local-sandbox', 'Dockerfile'),
+      '-t',
+      shellctlImage,
+      '.',
+    ],
+    cwd: difyAgentDir,
+  })
+}
+
+export const startShellctlSandbox = async () => {
+  if (await isTcpPortReachable(shellctlHost, shellctlPort)) {
+    const listenerDescription = await getTcpPortListenerDescription(shellctlPort)
+    const listenerMessage = listenerDescription
+      ? `\n\nPort listener:\n${listenerDescription}`
+      : ''
+
+    throw new Error(
+      `Cannot start the E2E shellctl sandbox because ${shellctlHost}:${shellctlPort} is already in use.${listenerMessage}`,
+    )
+  }
+
+  await runCommand({
+    command: 'docker',
+    args: ['rm', '-f', shellctlContainerName],
+    cwd: rootDir,
+    stdio: 'pipe',
+  })
+
+  await ensureShellctlSandboxImage()
+
+  await runForegroundProcess({
+    command: 'docker',
+    args: [
+      'run',
+      '--rm',
+      '--name',
+      shellctlContainerName,
+      '-p',
+      `${shellctlHost}:${shellctlPort}:5004`,
+      ...(process.env.E2E_SHELLCTL_AUTH_TOKEN
+        ? ['-e', `SHELLCTL_AUTH_TOKEN=${process.env.E2E_SHELLCTL_AUTH_TOKEN}`]
+        : []),
+      shellctlImage,
+    ],
+    cwd: rootDir,
+  })
+}
+
+export const startCelery = async ({ queues = 'workflow_based_app_execution' } = {}) => {
   const env = await getApiEnvironment()
 
   await runForegroundProcess({
@@ -291,7 +450,7 @@ export const startCelery = async () => {
       '--loglevel',
       'INFO',
       '-Q',
-      'workflow_based_app_execution',
+      queues,
     ],
     cwd: apiDir,
     env,
@@ -404,18 +563,23 @@ export const startMiddleware = async () => {
 }
 
 const printUsage = () => {
-  console.log('Usage: tsx ./scripts/setup.ts <reset|middleware-up|middleware-down|api|celery|web>')
+  console.log('Usage: tsx ./scripts/setup.ts <reset|middleware-up|middleware-down|shellctl-sandbox|agent-backend|api|celery [--queues queues]|web>')
 }
 
 const main = async () => {
   const command = process.argv[2]
+  const queuesIndex = process.argv.indexOf('--queues')
+  const queues = queuesIndex === -1 ? undefined : process.argv[queuesIndex + 1]
 
   switch (command) {
+    case 'agent-backend':
+      await startAgentBackend()
+      return
     case 'api':
       await startApi()
       return
     case 'celery':
-      await startCelery()
+      await startCelery({ queues })
       return
     case 'middleware-down':
       await stopMiddleware()
@@ -426,6 +590,9 @@ const main = async () => {
     case 'reset':
       await resetState()
       return
+    case 'shellctl-sandbox':
+      await startShellctlSandbox()
+      return
     case 'web':
       await startWeb()
       return
diff --git a/e2e/support/marketplace-plugins.ts b/e2e/support/marketplace-plugins.ts
new file mode 100644
index 00000000000..cd12778b06b
--- /dev/null
+++ b/e2e/support/marketplace-plugins.ts
@@ -0,0 +1,333 @@
+import type { SeedContext, SeedResult } from './seed'
+import { Buffer } from 'node:buffer'
+import { createApiContext, expectApiResponseOK } from './api'
+import { sleep } from './process'
+import {
+  blocked,
+  created,
+  skipped,
+  verified,
+} from './seed'
+
+type LatestPlugin = {
+  unique_identifier?: string
+  version?: string
+}
+
+type PluginInstallation = {
+  plugin_id: string
+  plugin_unique_identifier: string
+}
+
+type PluginInstallTask = {
+  id?: string
+  plugins?: Array<{
+    message?: string
+    plugin_id?: string
+    plugin_unique_identifier?: string
+    status?: string
+  }>
+  status?: string
+}
+
+type PluginInstallStartResponse = {
+  all_installed?: boolean
+  task?: PluginInstallTask | null
+  task_id?: string
+}
+
+type MarketplacePluginBootstrapConfig = {
+  defaultPluginIds: string[]
+  pluginIdsEnv: string
+  pluginUniqueIdentifiersEnv: string
+  title: string
+}
+
+const pendingTaskStatuses = new Set(['pending', 'running'])
+const terminalSuccessTaskStatus = 'success'
+const terminalFailedTaskStatus = 'failed'
+const defaultMarketplaceApiUrl = 'https://marketplace.dify.ai'
+
+const parseListEnv = (envName: string) =>
+  process.env[envName]
+    ?.split(/[\n,]/)
+    .map(item => item.trim())
+    .filter(Boolean) ?? []
+
+const unique = (values: string[]) => Array.from(new Set(values))
+
+const getPluginId = (pluginUniqueIdentifier: string) =>
+  pluginUniqueIdentifier.split(':')[0]?.trim() || pluginUniqueIdentifier.trim()
+
+const findPlaceholderPluginIdentifier = (pluginUniqueIdentifiers: string[]) =>
+  pluginUniqueIdentifiers.find(identifier => identifier.includes('replace-with-'))
+
+const withoutPlaceholderPluginIdentifiers = (pluginUniqueIdentifiers: string[]) =>
+  pluginUniqueIdentifiers.filter(identifier => !identifier.includes('replace-with-'))
+
+const resolveLatestPluginIdentifiers = async (pluginIds: string[]) => {
+  if (pluginIds.length === 0)
+    return { identifiers: [] as string[], missing: [] as string[] }
+
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/workspaces/current/plugin/list/latest-versions', {
+      data: { plugin_ids: pluginIds },
+    })
+    await expectApiResponseOK(response, 'Resolve latest marketplace plugin versions')
+    const body = (await response.json()) as { versions?: Record<string, LatestPlugin | null> }
+    const identifiers: string[] = []
+    const missing: string[] = []
+
+    for (const pluginId of pluginIds) {
+      const latest = body.versions?.[pluginId]
+      if (latest?.unique_identifier)
+        identifiers.push(latest.unique_identifier)
+      else
+        missing.push(pluginId)
+    }
+
+    return { identifiers, missing }
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const listInstalledPlugins = async (pluginIds: string[]) => {
+  if (pluginIds.length === 0)
+    return [] as PluginInstallation[]
+
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/workspaces/current/plugin/list/installations/ids', {
+      data: { plugin_ids: pluginIds },
+    })
+    await expectApiResponseOK(response, 'List installed marketplace plugins')
+    const body = (await response.json()) as { plugins?: PluginInstallation[] }
+    return body.plugins ?? []
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const waitForPluginInstallTask = async (taskId: string, timeoutMs = 300_000) => {
+  const deadline = Date.now() + timeoutMs
+  let lastTask: PluginInstallTask | undefined
+
+  while (Date.now() < deadline) {
+    const ctx = await createApiContext()
+    try {
+      const response = await ctx.get(`/console/api/workspaces/current/plugin/tasks/${taskId}`)
+      await expectApiResponseOK(response, `Fetch marketplace plugin install task ${taskId}`)
+      const body = (await response.json()) as { task?: PluginInstallTask }
+      lastTask = body.task
+    }
+    finally {
+      await ctx.dispose()
+    }
+
+    if (lastTask?.status === terminalSuccessTaskStatus)
+      return { ok: true as const, task: lastTask }
+
+    if (lastTask?.status === terminalFailedTaskStatus) {
+      const details = lastTask.plugins
+        ?.filter(plugin => plugin.status === terminalFailedTaskStatus)
+        .map(plugin => `${plugin.plugin_id ?? plugin.plugin_unique_identifier ?? 'unknown'}: ${plugin.message ?? 'failed'}`)
+        .join('; ')
+      return { ok: false as const, reason: details || 'Plugin install task failed.' }
+    }
+
+    if (!lastTask?.status || pendingTaskStatuses.has(lastTask.status)) {
+      await sleep(2_000)
+      continue
+    }
+
+    return { ok: false as const, reason: `Plugin install task ended with status ${lastTask.status}.` }
+  }
+
+  return { ok: false as const, reason: `Plugin install task did not finish within ${timeoutMs}ms.` }
+}
+
+const installMarketplacePlugins = async (pluginUniqueIdentifiers: string[]) => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/workspaces/current/plugin/install/marketplace', {
+      data: { plugin_unique_identifiers: pluginUniqueIdentifiers },
+    })
+    await expectApiResponseOK(response, 'Install marketplace plugins')
+    return (await response.json()) as PluginInstallStartResponse
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const getMarketplaceDownloadUrl = (pluginUniqueIdentifier: string) => {
+  const url = new URL('/api/v1/plugins/download', process.env.E2E_MARKETPLACE_API_URL || defaultMarketplaceApiUrl)
+  url.searchParams.set('unique_identifier', pluginUniqueIdentifier)
+  return url.toString()
+}
+
+const downloadMarketplacePluginPackage = async (pluginUniqueIdentifier: string) => {
+  const response = await fetch(getMarketplaceDownloadUrl(pluginUniqueIdentifier))
+  if (!response.ok) {
+    throw new Error(
+      `Download marketplace package for ${getPluginId(pluginUniqueIdentifier)} failed with ${response.status} ${response.statusText}.`,
+    )
+  }
+
+  return Buffer.from(await response.arrayBuffer())
+}
+
+const uploadMarketplacePluginPackage = async (pluginUniqueIdentifier: string) => {
+  const pkg = await downloadMarketplacePluginPackage(pluginUniqueIdentifier)
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/workspaces/current/plugin/upload/pkg', {
+      multipart: {
+        pkg: {
+          buffer: pkg,
+          mimeType: 'application/octet-stream',
+          name: `${getPluginId(pluginUniqueIdentifier).replaceAll('/', '-')}.difypkg`,
+        },
+      },
+    })
+    await expectApiResponseOK(response, `Upload marketplace package ${getPluginId(pluginUniqueIdentifier)}`)
+    const body = (await response.json()) as { unique_identifier?: string }
+    if (!body.unique_identifier)
+      throw new Error(`Upload marketplace package ${getPluginId(pluginUniqueIdentifier)} did not return a unique identifier.`)
+
+    return body.unique_identifier
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const installLocalPluginPackages = async (pluginUniqueIdentifiers: string[]) => {
+  const ctx = await createApiContext()
+  try {
+    const response = await ctx.post('/console/api/workspaces/current/plugin/install/pkg', {
+      data: { plugin_unique_identifiers: pluginUniqueIdentifiers },
+    })
+    await expectApiResponseOK(response, 'Install uploaded plugin packages')
+    return (await response.json()) as PluginInstallStartResponse
+  }
+  finally {
+    await ctx.dispose()
+  }
+}
+
+const shouldFallbackToLocalPackageInstall = (error: string) =>
+  error.includes('/plugins/download') || error.includes('Reached maximum retries')
+
+const installMarketplacePluginsWithFallback = async (pluginUniqueIdentifiers: string[]) => {
+  try {
+    return await installMarketplacePlugins(pluginUniqueIdentifiers)
+  }
+  catch (error) {
+    const message = error instanceof Error ? error.message : String(error)
+    if (!shouldFallbackToLocalPackageInstall(message))
+      throw error
+
+    console.warn('[seed] marketplace install download failed in API process; falling back to local package upload.')
+    const uploadedPluginUniqueIdentifiers: string[] = []
+    for (const pluginUniqueIdentifier of pluginUniqueIdentifiers)
+      uploadedPluginUniqueIdentifiers.push(await uploadMarketplacePluginPackage(pluginUniqueIdentifier))
+
+    return await installLocalPluginPackages(uploadedPluginUniqueIdentifiers)
+  }
+}
+
+export const bootstrapMarketplacePlugins = async (
+  context: SeedContext,
+  config: MarketplacePluginBootstrapConfig,
+): Promise<SeedResult> => {
+  const requestedPluginIds = parseListEnv(config.pluginIdsEnv)
+  const rawExactPluginUniqueIdentifiers = unique(parseListEnv(config.pluginUniqueIdentifiersEnv))
+  const exactPluginUniqueIdentifiers = withoutPlaceholderPluginIdentifiers(rawExactPluginUniqueIdentifiers)
+  const placeholderPluginIdentifier = findPlaceholderPluginIdentifier(rawExactPluginUniqueIdentifiers)
+  if (placeholderPluginIdentifier) {
+    console.warn(`[seed] ignoring example marketplace package placeholder for ${getPluginId(placeholderPluginIdentifier)}.`)
+  }
+
+  const pluginIds = unique(
+    exactPluginUniqueIdentifiers.length > 0
+      ? []
+      : requestedPluginIds.length > 0
+        ? requestedPluginIds
+        : config.defaultPluginIds,
+  )
+
+  if (pluginIds.length > 0) {
+    const installedPlugins = await listInstalledPlugins(pluginIds)
+    const installedPluginIds = new Set(installedPlugins.map(plugin => plugin.plugin_id))
+    if (pluginIds.every(pluginId => installedPluginIds.has(pluginId))) {
+      return verified(config.title, {
+        id: pluginIds.join(','),
+        kind: 'marketplace-plugins',
+        name: config.title,
+      })
+    }
+  }
+
+  const resolved = await resolveLatestPluginIdentifiers(pluginIds)
+
+  if (resolved.missing.length > 0) {
+    return blocked(
+      config.title,
+      `Marketplace metadata was not found for plugin ids: ${resolved.missing.join(', ')}. Set ${config.pluginUniqueIdentifiersEnv} to exact package identifiers to bypass latest-version lookup.`,
+    )
+  }
+
+  const requiredPluginUniqueIdentifiers = unique([
+    ...resolved.identifiers,
+    ...exactPluginUniqueIdentifiers,
+  ])
+  const requiredPluginIds = unique(requiredPluginUniqueIdentifiers.map(getPluginId))
+
+  if (requiredPluginUniqueIdentifiers.length === 0)
+    return skipped(config.title, 'No marketplace plugins were requested.')
+
+  const installedPlugins = await listInstalledPlugins(requiredPluginIds)
+  const installedPluginIds = new Set(installedPlugins.map(plugin => plugin.plugin_id))
+  const missingPluginUniqueIdentifiers = requiredPluginUniqueIdentifiers.filter(
+    identifier => !installedPluginIds.has(getPluginId(identifier)),
+  )
+  const resource = {
+    id: requiredPluginIds.join(','),
+    kind: 'marketplace-plugins',
+    name: config.title,
+  }
+
+  if (missingPluginUniqueIdentifiers.length === 0)
+    return verified(config.title, resource)
+
+  if (context.dryRun) {
+    return skipped(
+      config.title,
+      `Would install marketplace plugins: ${missingPluginUniqueIdentifiers.map(getPluginId).join(', ')}.`,
+    )
+  }
+
+  const startedTask = await installMarketplacePluginsWithFallback(missingPluginUniqueIdentifiers).catch((error) => {
+    return { error: error instanceof Error ? error.message : String(error) }
+  })
+  if ('error' in startedTask)
+    return blocked(config.title, startedTask.error)
+
+  if (startedTask.all_installed)
+    return verified(config.title, resource)
+
+  const taskId = startedTask.task_id || startedTask.task?.id
+  if (!taskId)
+    return blocked(config.title, 'Marketplace plugin install did not return a task id.')
+
+  const taskResult = await waitForPluginInstallTask(taskId)
+  if (!taskResult.ok)
+    return blocked(config.title, taskResult.reason)
+
+  return created(config.title, resource)
+}
diff --git a/e2e/support/seed.ts b/e2e/support/seed.ts
new file mode 100644
index 00000000000..1e89cbcee55
--- /dev/null
+++ b/e2e/support/seed.ts
@@ -0,0 +1,93 @@
+import { mkdir, writeFile } from 'node:fs/promises'
+import path from 'node:path'
+import { e2eDir } from '../scripts/common'
+
+export type SeedStatus = 'blocked' | 'created' | 'skipped' | 'updated' | 'verified'
+
+export type SeedResource = {
+  id?: string
+  kind: string
+  name: string
+}
+
+export type SeedResult = {
+  reason?: string
+  resource?: SeedResource
+  status: SeedStatus
+  title: string
+}
+
+export type SeedContext = {
+  dryRun: boolean
+  resources: Map<string, SeedResource>
+}
+
+export type SeedTask = {
+  id: string
+  run: (context: SeedContext) => Promise<SeedResult>
+  title: string
+}
+
+const reportDir = path.join(e2eDir, 'seed-report')
+
+export const created = (title: string, resource?: SeedResource): SeedResult => ({
+  resource,
+  status: 'created',
+  title,
+})
+
+export const updated = (title: string, resource?: SeedResource): SeedResult => ({
+  resource,
+  status: 'updated',
+  title,
+})
+
+export const verified = (title: string, resource?: SeedResource): SeedResult => ({
+  resource,
+  status: 'verified',
+  title,
+})
+
+export const skipped = (title: string, reason: string): SeedResult => ({
+  reason,
+  status: 'skipped',
+  title,
+})
+
+export const blocked = (title: string, reason: string): SeedResult => ({
+  reason,
+  status: 'blocked',
+  title,
+})
+
+export async function runSeedTasks(tasks: SeedTask[], context: SeedContext) {
+  const results: SeedResult[] = []
+
+  for (const task of tasks) {
+    const result = await task.run(context)
+    results.push(result)
+    if (result.resource)
+      context.resources.set(task.id, result.resource)
+
+    const suffix = result.reason ? `: ${result.reason}` : ''
+    console.warn(`[seed] ${result.status.padEnd(8)} ${result.title}${suffix}`)
+  }
+
+  return results
+}
+
+export async function writeSeedReport(pack: string, results: SeedResult[]) {
+  await mkdir(reportDir, { recursive: true })
+  const reportPath = path.join(reportDir, `${pack}.json`)
+  await writeFile(
+    reportPath,
+    `${JSON.stringify({
+      generated_at: new Date().toISOString(),
+      pack,
+      results,
+    }, null, 2)}\n`,
+    'utf8',
+  )
+
+  return reportPath
+}
diff --git a/eslint-suppressions.json b/eslint-suppressions.json
index ef487d7e69e..d08aba6fad7 100644
--- a/eslint-suppressions.json
+++ b/eslint-suppressions.json
@@ -1170,12 +1170,6 @@
     }
   },
   "web/app/components/base/chat/chat/answer/workflow-process.tsx": {
-    "jsx-a11y/click-events-have-key-events": {
-      "count": 1
-    },
-    "jsx-a11y/no-static-element-interactions": {
-      "count": 1
-    },
     "react/set-state-in-effect": {
       "count": 1
     }
diff --git a/packages/contracts/generated/api/console/agent/types.gen.ts b/packages/contracts/generated/api/console/agent/types.gen.ts
index c7a35f003db..5a28df41715 100644
--- a/packages/contracts/generated/api/console/agent/types.gen.ts
+++ b/packages/contracts/generated/api/console/agent/types.gen.ts
@@ -18,7 +18,7 @@ export type AgentAppCreatePayload = {
   icon_background?: string | null
   icon_type?: IconType | null
   name: string
-  role: string
+  role?: string | null
 }
 
 export type AgentAppDetailWithSite = {
@@ -74,7 +74,7 @@ export type AgentAppUpdatePayload = {
   icon_type?: IconType | null
   max_active_requests?: number | null
   name: string
-  role: string
+  role?: string | null
   use_icon_as_answer_icon?: boolean | null
 }
 
@@ -846,7 +846,7 @@ export type AgentSensitiveWordAvoidanceFeatureConfig = {
 
 export type AgentSuggestedQuestionsAfterAnswerFeatureConfig = {
   enabled?: boolean
-  model?: AgentSoulModelConfig | null
+  model?: AgentSuggestedQuestionsAfterAnswerModelConfig | null
   prompt?: string | null
   [key: string]: unknown
 }
@@ -1327,6 +1327,16 @@ export type AgentModerationProviderConfig = {
   [key: string]: unknown
 }
 
+export type AgentSuggestedQuestionsAfterAnswerModelConfig = {
+  completion_params?: {
+    [key: string]: unknown
+  } | null
+  mode?: string | null
+  name: string
+  provider: string
+  [key: string]: unknown
+}
+
 export type SimpleAccount = {
   email: string
   id: string
diff --git a/packages/contracts/generated/api/console/agent/zod.gen.ts b/packages/contracts/generated/api/console/agent/zod.gen.ts
index 52f46a4ae1f..0ae9fd0dae6 100644
--- a/packages/contracts/generated/api/console/agent/zod.gen.ts
+++ b/packages/contracts/generated/api/console/agent/zod.gen.ts
@@ -236,7 +236,7 @@ export const zAgentAppCreatePayload = z.object({
   icon_background: z.string().nullish(),
   icon_type: zIconType.nullish(),
   name: z.string().min(1),
-  role: z.string().min(1).max(255),
+  role: z.string().max(255).nullish(),
 })
 
 /**
@@ -249,7 +249,7 @@ export const zAgentAppUpdatePayload = z.object({
   icon_type: zIconType.nullish(),
   max_active_requests: z.int().nullish(),
   name: z.string().min(1),
-  role: z.string().min(1).max(255),
+  role: z.string().max(255).nullish(),
   use_icon_as_answer_icon: z.boolean().nullish(),
 })
 
@@ -1327,6 +1327,27 @@ export const zAgentComposerDifyToolCandidateResponse = z.object({
   tools_count: z.int().nullish(),
 })
 
+/**
+ * AgentSuggestedQuestionsAfterAnswerModelConfig
+ *
+ * Legacy Chat App model config used only for follow-up question generation.
+ */
+export const zAgentSuggestedQuestionsAfterAnswerModelConfig = z.object({
+  completion_params: z.record(z.string(), z.unknown()).nullish(),
+  mode: z.string().max(64).nullish(),
+  name: z.string().min(1).max(255),
+  provider: z.string().min(1).max(255),
+})
+
+/**
+ * AgentSuggestedQuestionsAfterAnswerFeatureConfig
+ */
+export const zAgentSuggestedQuestionsAfterAnswerFeatureConfig = z.object({
+  enabled: z.boolean().optional().default(false),
+  model: zAgentSuggestedQuestionsAfterAnswerModelConfig.nullish(),
+  prompt: z.string().nullish(),
+})
+
 /**
  * SimpleAccount
  */
@@ -1760,7 +1781,7 @@ export const zAgentSecretRefConfig = z.object({
   provider_credential_id: z.string().max(255).nullish(),
   ref: z.string().max(255).nullish(),
   type: z.string().max(64).nullish(),
-  value: z.string().max(255).nullish(),
+  value: z.string().nullish(),
   variable: z.string().max(255).nullish(),
 })
 
@@ -1887,6 +1908,37 @@ export const zAgentSensitiveWordAvoidanceFeatureConfig = z.object({
   type: z.string().nullish(),
 })
 
+/**
+ * AgentAppFeaturesPayload
+ *
+ * Presentation features configurable on an Agent App.
+ *
+ * All fields are optional; an omitted field is reset to its disabled/empty
+ * default (the config form sends the full desired feature state on save).
+ */
+export const zAgentAppFeaturesPayload = z.object({
+  opening_statement: z.string().nullish(),
+  retriever_resource: zAgentFeatureToggleConfig.nullish(),
+  sensitive_word_avoidance: zAgentSensitiveWordAvoidanceFeatureConfig.nullish(),
+  speech_to_text: zAgentFeatureToggleConfig.nullish(),
+  suggested_questions: z.array(z.string()).nullish(),
+  suggested_questions_after_answer: zAgentSuggestedQuestionsAfterAnswerFeatureConfig.nullish(),
+  text_to_speech: zAgentTextToSpeechFeatureConfig.nullish(),
+})
+
+/**
+ * AgentSoulAppFeaturesConfig
+ */
+export const zAgentSoulAppFeaturesConfig = z.object({
+  opening_statement: z.string().nullish(),
+  retriever_resource: zAgentFeatureToggleConfig.nullish(),
+  sensitive_word_avoidance: zAgentSensitiveWordAvoidanceFeatureConfig.nullish(),
+  speech_to_text: zAgentFeatureToggleConfig.nullish(),
+  suggested_questions: z.array(z.string()).nullish(),
+  suggested_questions_after_answer: zAgentSuggestedQuestionsAfterAnswerFeatureConfig.nullish(),
+  text_to_speech: zAgentTextToSpeechFeatureConfig.nullish(),
+})
+
 export const zJsonValue2 = z.unknown()
 
 /**
@@ -1943,46 +1995,6 @@ export const zAgentSoulModelConfig = z.object({
   plugin_id: z.string().min(1).max(255),
 })
 
-/**
- * AgentSuggestedQuestionsAfterAnswerFeatureConfig
- */
-export const zAgentSuggestedQuestionsAfterAnswerFeatureConfig = z.object({
-  enabled: z.boolean().optional().default(false),
-  model: zAgentSoulModelConfig.nullish(),
-  prompt: z.string().nullish(),
-})
-
-/**
- * AgentAppFeaturesPayload
- *
- * Presentation features configurable on an Agent App.
- *
- * All fields are optional; an omitted field is reset to its disabled/empty
- * default (the config form sends the full desired feature state on save).
- */
-export const zAgentAppFeaturesPayload = z.object({
-  opening_statement: z.string().nullish(),
-  retriever_resource: zAgentFeatureToggleConfig.nullish(),
-  sensitive_word_avoidance: zAgentSensitiveWordAvoidanceFeatureConfig.nullish(),
-  speech_to_text: zAgentFeatureToggleConfig.nullish(),
-  suggested_questions: z.array(z.string()).nullish(),
-  suggested_questions_after_answer: zAgentSuggestedQuestionsAfterAnswerFeatureConfig.nullish(),
-  text_to_speech: zAgentTextToSpeechFeatureConfig.nullish(),
-})
-
-/**
- * AgentSoulAppFeaturesConfig
- */
-export const zAgentSoulAppFeaturesConfig = z.object({
-  opening_statement: z.string().nullish(),
-  retriever_resource: zAgentFeatureToggleConfig.nullish(),
-  sensitive_word_avoidance: zAgentSensitiveWordAvoidanceFeatureConfig.nullish(),
-  speech_to_text: zAgentFeatureToggleConfig.nullish(),
-  suggested_questions: z.array(z.string()).nullish(),
-  suggested_questions_after_answer: zAgentSuggestedQuestionsAfterAnswerFeatureConfig.nullish(),
-  text_to_speech: zAgentTextToSpeechFeatureConfig.nullish(),
-})
-
 /**
  * DeclaredOutputCheckConfig
  *
diff --git a/packages/contracts/generated/api/console/apps/types.gen.ts b/packages/contracts/generated/api/console/apps/types.gen.ts
index f880dbd9004..6c8ab2f2cf2 100644
--- a/packages/contracts/generated/api/console/apps/types.gen.ts
+++ b/packages/contracts/generated/api/console/apps/types.gen.ts
@@ -2635,7 +2635,7 @@ export type AgentSensitiveWordAvoidanceFeatureConfig = {
 
 export type AgentSuggestedQuestionsAfterAnswerFeatureConfig = {
   enabled?: boolean
-  model?: AgentSoulModelConfig | null
+  model?: AgentSuggestedQuestionsAfterAnswerModelConfig | null
   prompt?: string | null
   [key: string]: unknown
 }
@@ -2875,6 +2875,16 @@ export type AgentModerationProviderConfig = {
   [key: string]: unknown
 }
 
+export type AgentSuggestedQuestionsAfterAnswerModelConfig = {
+  completion_params?: {
+    [key: string]: unknown
+  } | null
+  mode?: string | null
+  name: string
+  provider: string
+  [key: string]: unknown
+}
+
 export type AgentKnowledgeDatasetConfig = {
   description?: string | null
   id?: string | null
diff --git a/packages/contracts/generated/api/console/apps/zod.gen.ts b/packages/contracts/generated/api/console/apps/zod.gen.ts
index 74185ad0dfa..dfa767fd22e 100644
--- a/packages/contracts/generated/api/console/apps/zod.gen.ts
+++ b/packages/contracts/generated/api/console/apps/zod.gen.ts
@@ -3339,7 +3339,7 @@ export const zAgentSecretRefConfig = z.object({
   provider_credential_id: z.string().max(255).nullish(),
   ref: z.string().max(255).nullish(),
   type: z.string().max(64).nullish(),
-  value: z.string().max(255).nullish(),
+  value: z.string().nullish(),
   variable: z.string().max(255).nullish(),
 })
 
@@ -3457,6 +3457,27 @@ export const zUserActionConfig = z.object({
   title: z.string().max(100),
 })
 
+/**
+ * AgentSuggestedQuestionsAfterAnswerModelConfig
+ *
+ * Legacy Chat App model config used only for follow-up question generation.
+ */
+export const zAgentSuggestedQuestionsAfterAnswerModelConfig = z.object({
+  completion_params: z.record(z.string(), z.unknown()).nullish(),
+  mode: z.string().max(64).nullish(),
+  name: z.string().min(1).max(255),
+  provider: z.string().min(1).max(255),
+})
+
+/**
+ * AgentSuggestedQuestionsAfterAnswerFeatureConfig
+ */
+export const zAgentSuggestedQuestionsAfterAnswerFeatureConfig = z.object({
+  enabled: z.boolean().optional().default(false),
+  model: zAgentSuggestedQuestionsAfterAnswerModelConfig.nullish(),
+  prompt: z.string().nullish(),
+})
+
 /**
  * AgentKnowledgeDatasetConfig
  */
@@ -3566,15 +3587,6 @@ export const zWorkflowNodeJobConfig = z.object({
   workflow_prompt: z.string().optional().default(''),
 })
 
-/**
- * AgentSuggestedQuestionsAfterAnswerFeatureConfig
- */
-export const zAgentSuggestedQuestionsAfterAnswerFeatureConfig = z.object({
-  enabled: z.boolean().optional().default(false),
-  model: zAgentSoulModelConfig.nullish(),
-  prompt: z.string().nullish(),
-})
-
 /**
  * AgentSoulDifyToolCredentialRef
  *
diff --git a/packages/dify-ui/src/file-tree/index.tsx b/packages/dify-ui/src/file-tree/index.tsx
index 64555d3ca48..e368364b7ec 100644
--- a/packages/dify-ui/src/file-tree/index.tsx
+++ b/packages/dify-ui/src/file-tree/index.tsx
@@ -288,7 +288,7 @@ export function FileTreeIcon({
             ? (
                 <React.Fragment>
                   <span className="size-4 i-ri-folder-line group-data-panel-open/file-tree-row:hidden" />
-                  <span className="hidden size-4 text-text-accent i-ri-folder-open-line group-data-panel-open/file-tree-row:block" />
+                  <span className="hidden size-4 text-text-secondary i-ri-folder-open-line group-data-panel-open/file-tree-row:block" />
                 </React.Fragment>
               )
             : <span className={cn('size-4', fileTreeIconClassNames[type])} />
@@ -319,7 +319,7 @@ export function FileTreeLabel({
   const defaultProps = {
     'data-label': labelText,
     'className': cn(
-      'min-w-0 truncate rounded-[5px] px-1 py-0.5',
+      'w-0 min-w-0 flex-1 truncate rounded-[5px] px-1 py-0.5',
       labelText && 'after:invisible after:block after:h-0 after:overflow-hidden after:system-sm-medium after:content-[attr(data-label)]',
       'system-sm-regular text-text-secondary group-data-[selected]/file-tree-row:system-sm-medium group-data-[selected]/file-tree-row:text-text-primary',
       className,
diff --git a/packages/dify-ui/src/tooltip/index.tsx b/packages/dify-ui/src/tooltip/index.tsx
index 359fb532e83..cd65da1727b 100644
--- a/packages/dify-ui/src/tooltip/index.tsx
+++ b/packages/dify-ui/src/tooltip/index.tsx
@@ -62,7 +62,7 @@ export function TooltipContent({
       >
         <BaseTooltip.Popup
           className={cn(
-            'max-w-[300px] rounded-md bg-components-panel-bg px-3 py-2 text-start system-xs-regular wrap-break-word text-text-tertiary shadow-lg',
+            'max-w-75 rounded-md bg-components-panel-bg px-3 py-2 text-start system-xs-regular wrap-break-word text-text-tertiary shadow-lg',
             'origin-(--transform-origin) transition-opacity data-ending-style:opacity-0 data-instant:transition-none data-starting-style:opacity-0 motion-reduce:transition-none',
             className,
           )}
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml
index 91e4c3d8e8a..d9ac763b741 100644
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -153,6 +153,9 @@ catalogs:
     '@svgdotjs/svg.js':
       specifier: 3.2.5
       version: 3.2.5
+    '@t3-oss/env-core':
+      specifier: 0.13.11
+      version: 0.13.11
     '@t3-oss/env-nextjs':
       specifier: 0.13.11
       version: 0.13.11
@@ -772,6 +775,9 @@ importers:
       '@playwright/test':
         specifier: 'catalog:'
         version: 1.61.1
+      '@t3-oss/env-core':
+        specifier: 'catalog:'
+        version: 0.13.11(typescript@6.0.3)(valibot@1.4.2(typescript@6.0.3))(zod@4.4.3)
       '@types/node':
         specifier: 'catalog:'
         version: 25.9.4
@@ -790,6 +796,9 @@ importers:
       vite-plus:
         specifier: 'catalog:'
         version: 0.2.1(@types/node@25.9.4)(@vitest/browser-playwright@4.1.9)(@vitest/coverage-v8@4.1.9)(@voidzero-dev/vite-plus-core@0.2.1(@types/node@25.9.4)(esbuild@0.28.1)(jiti@2.7.0)(tsx@4.22.4)(typescript@6.0.3)(yaml@2.9.0))(esbuild@0.28.1)(happy-dom@20.10.6)(jiti@2.7.0)(tsx@4.22.4)(typescript@6.0.3)(yaml@2.9.0)
+      zod:
+        specifier: 'catalog:'
+        version: 4.4.3
 
   packages/contracts:
     dependencies:
@@ -20092,6 +20101,7 @@ time:
   '@storybook/react@10.4.6': '2026-06-16T11:43:30.413Z'
   '@streamdown/math@1.0.2': '2026-02-09T17:31:31.085Z'
   '@svgdotjs/svg.js@3.2.5': '2025-09-15T16:22:12.771Z'
+  '@t3-oss/env-core@0.13.11': '2026-03-22T19:16:09.121Z'
   '@t3-oss/env-nextjs@0.13.11': '2026-03-22T19:16:09.026Z'
   '@tailwindcss/postcss@4.3.1': '2026-06-12T17:59:36.796Z'
   '@tailwindcss/typography@0.5.20': '2026-06-08T10:34:41.124Z'
diff --git a/pnpm-workspace.yaml b/pnpm-workspace.yaml
index ad5d91fbfff..62c36b70183 100644
--- a/pnpm-workspace.yaml
+++ b/pnpm-workspace.yaml
@@ -101,6 +101,7 @@ catalog:
   '@storybook/react-vite': 10.4.6
   '@streamdown/math': 1.0.2
   '@svgdotjs/svg.js': 3.2.5
+  '@t3-oss/env-core': 0.13.11
   '@t3-oss/env-nextjs': 0.13.11
   '@tailwindcss/postcss': 4.3.1
   '@tailwindcss/typography': 0.5.20
diff --git a/web/app/components/base/chat/chat/__tests__/index.spec.tsx b/web/app/components/base/chat/chat/__tests__/index.spec.tsx
index 440022ff23e..01c45e7bbb7 100644
--- a/web/app/components/base/chat/chat/__tests__/index.spec.tsx
+++ b/web/app/components/base/chat/chat/__tests__/index.spec.tsx
@@ -46,17 +46,21 @@ vi.mock('../chat-input-area', () => ({
     customPlaceholder,
     disabled,
     readonly,
+    footerNotice,
   }: {
     customPlaceholder?: string
     disabled?: boolean
     readonly?: boolean
+    footerNotice?: string
   }) => (
     <div
       data-testid="chat-input-area"
       data-custom-placeholder={customPlaceholder}
       data-disabled={String(!!disabled)}
       data-readonly={String(!!readonly)}
-    />
+    >
+      {footerNotice}
+    </div>
   ),
 }))
 
@@ -888,6 +892,15 @@ describe('Chat', () => {
       expect(screen.getByTestId('chat-input-area')).toBeInTheDocument()
     })
 
+    it('should pass footer notice to ChatInputArea', () => {
+      renderChat({
+        noChatInput: false,
+        footerNotice: 'Agent runs in a Linux sandbox.',
+      })
+
+      expect(screen.getByTestId('chat-input-area')).toHaveTextContent('Agent runs in a Linux sandbox.')
+    })
+
     it('should pass inputs and inputsForm to ChatInputArea', () => {
       const inputs = { field1: 'value1' }
       const inputsForm = [{ key: 'field1', type: 'text' }]
diff --git a/web/app/components/base/chat/chat/answer/__tests__/workflow-process.spec.tsx b/web/app/components/base/chat/chat/answer/__tests__/workflow-process.spec.tsx
index f8ae9aa3dc5..a1d49280a19 100644
--- a/web/app/components/base/chat/chat/answer/__tests__/workflow-process.spec.tsx
+++ b/web/app/components/base/chat/chat/answer/__tests__/workflow-process.spec.tsx
@@ -20,7 +20,7 @@ describe('WorkflowProcessItem', () => {
 
   it('should render the latest node title when collapsed', () => {
     render(<WorkflowProcessItem data={mockData as WorkflowProcess} expand={false} />)
-    expect(screen.getByTestId('workflow-process-title')).toHaveTextContent('End')
+    expect(screen.getByRole('button', { name: /End/ })).toBeInTheDocument()
     expect(screen.queryByTestId('tracing-panel')).not.toBeInTheDocument()
   })
 
@@ -36,7 +36,7 @@ describe('WorkflowProcessItem', () => {
       />,
     )
 
-    expect(screen.getByTestId('workflow-process-title')).toHaveTextContent('Invalid upload file')
+    expect(screen.getByRole('button', { name: /Invalid upload file/ })).toBeInTheDocument()
   })
 
   it('should render "Workflow Process" title and TracingPanel when expanded', () => {
@@ -65,7 +65,7 @@ describe('WorkflowProcessItem', () => {
     const user = userEvent.setup()
     render(<WorkflowProcessItem data={mockData as WorkflowProcess} expand={false} />)
 
-    const header = screen.getByTestId('workflow-process-header')
+    const header = screen.getByRole('button', { name: /End/ })
 
     // Expand
     await user.click(header)
@@ -73,9 +73,9 @@ describe('WorkflowProcessItem', () => {
     expect(screen.getByText(/workflowProcess/i)).toBeInTheDocument()
 
     // Collapse
-    await user.click(header)
+    await user.click(screen.getByRole('button', { name: /workflowProcess/ }))
     expect(screen.queryByTestId('tracing-panel')).not.toBeInTheDocument()
-    expect(screen.getByTestId('workflow-process-title')).toHaveTextContent('End')
+    expect(screen.getByRole('button', { name: /End/ })).toBeInTheDocument()
   })
 
   it('should render nothing if readonly is true', () => {
@@ -86,27 +86,27 @@ describe('WorkflowProcessItem', () => {
   describe('Status Icons', () => {
     it('should show running spinner when status is Running', () => {
       render(<WorkflowProcessItem data={{ ...mockData, status: WorkflowRunningStatus.Running } as WorkflowProcess} />)
-      expect(screen.getByTestId('status-icon-running')).toBeInTheDocument()
+      expect(screen.getByRole('img', { name: /workflowProcessRunning/ })).toBeInTheDocument()
     })
 
     it('should show success circle when status is Succeeded', () => {
       render(<WorkflowProcessItem data={{ ...mockData, status: WorkflowRunningStatus.Succeeded } as WorkflowProcess} />)
-      expect(screen.getByTestId('status-icon-success')).toBeInTheDocument()
+      expect(screen.getByRole('img', { name: /workflowProcessSucceeded/ })).toBeInTheDocument()
     })
 
     it('should show error warning when status is Failed', () => {
       render(<WorkflowProcessItem data={{ ...mockData, status: WorkflowRunningStatus.Failed } as WorkflowProcess} />)
-      expect(screen.getByTestId('status-icon-failed')).toBeInTheDocument()
+      expect(screen.getByRole('img', { name: /workflowProcessFailed/ })).toBeInTheDocument()
     })
 
     it('should show error warning when status is Stopped', () => {
       render(<WorkflowProcessItem data={{ ...mockData, status: WorkflowRunningStatus.Stopped } as WorkflowProcess} />)
-      expect(screen.getByTestId('status-icon-failed')).toBeInTheDocument()
+      expect(screen.getByRole('img', { name: /workflowProcessFailed/ })).toBeInTheDocument()
     })
 
     it('should show pause circle when status is Paused', () => {
       render(<WorkflowProcessItem data={{ ...mockData, status: WorkflowRunningStatus.Paused } as WorkflowProcess} />)
-      expect(screen.getByTestId('status-icon-paused')).toBeInTheDocument()
+      expect(screen.getByRole('img', { name: /workflowProcessPaused/ })).toBeInTheDocument()
     })
   })
 
diff --git a/web/app/components/base/chat/chat/answer/workflow-process.tsx b/web/app/components/base/chat/chat/answer/workflow-process.tsx
index 0427f9843b0..6b506221867 100644
--- a/web/app/components/base/chat/chat/answer/workflow-process.tsx
+++ b/web/app/components/base/chat/chat/answer/workflow-process.tsx
@@ -32,6 +32,15 @@ const WorkflowProcessItem = ({
   const paused = data.status === WorkflowRunningStatus.Paused
   const latestNode = data.tracing[data.tracing.length - 1]
   const fallbackTitle = t('common.workflowProcess', { ns: 'workflow' })
+  const statusLabel = running
+    ? t('common.workflowProcessRunning', { ns: 'workflow' })
+    : succeeded
+      ? t('common.workflowProcessSucceeded', { ns: 'workflow' })
+      : failed
+        ? t('common.workflowProcessFailed', { ns: 'workflow' })
+        : paused
+          ? t('common.workflowProcessPaused', { ns: 'workflow' })
+          : undefined
   const collapsedTitle = failed
     ? data.error || latestNode?.error || latestNode?.title || fallbackTitle
     : latestNode?.title || fallbackTitle
@@ -58,40 +67,45 @@ const WorkflowProcessItem = ({
       )}
       data-testid="workflow-process-item"
     >
-      <div
-        className={cn('flex cursor-pointer items-center', !collapse && 'px-1.5')}
+      <button
+        type="button"
+        className={cn('flex w-full cursor-pointer items-center border-0 bg-transparent p-0 text-left', !collapse && 'px-1.5')}
+        aria-expanded={!collapse}
         onClick={() => setCollapse(!collapse)}
-        data-testid="workflow-process-header"
       >
         {
           running && (
-            <div
+            <span
+              role="img"
+              aria-label={statusLabel}
               className="mr-1 i-ri-loader-2-line size-3.5 shrink-0 animate-spin text-text-tertiary"
-              data-testid="status-icon-running"
             />
           )
         }
         {
           succeeded && (
-            <div
+            <span
+              role="img"
+              aria-label={statusLabel}
               className="mr-1 i-custom-vender-solid-general-check-circle size-3.5 shrink-0 text-text-success"
-              data-testid="status-icon-success"
             />
           )
         }
         {
           failed && (
-            <div
+            <span
+              role="img"
+              aria-label={statusLabel}
               className="mr-1 i-ri-error-warning-fill size-3.5 shrink-0 text-text-destructive"
-              data-testid="status-icon-failed"
             />
           )
         }
         {
           paused && (
-            <div
+            <span
+              role="img"
+              aria-label={statusLabel}
               className="mr-1 i-ri-pause-circle-fill size-3.5 shrink-0 text-text-warning-secondary"
-              data-testid="status-icon-paused"
             />
           )
         }
@@ -100,20 +114,19 @@ const WorkflowProcessItem = ({
             'min-w-0 grow truncate system-xs-medium',
             collapse && failed && data.error ? 'text-text-destructive' : 'text-text-secondary',
           )}
-          data-testid="workflow-process-title"
         >
           {!collapse ? fallbackTitle : collapsedTitle}
         </div>
-        <div className={cn('ml-1 i-ri-arrow-right-s-line size-4 shrink-0 text-text-tertiary', !collapse && 'rotate-90')} />
-      </div>
+        <span aria-hidden className={cn('ml-1 i-ri-arrow-right-s-line size-4 shrink-0 text-text-tertiary', !collapse && 'rotate-90')} />
+      </button>
       {
         !collapse && (
           <div className="mt-1.5">
             {
               failed && data.error && (
                 <div
+                  role="alert"
                   className="mb-1.5 rounded-lg border-[0.5px] border-state-destructive-border bg-state-destructive-hover px-2 py-1.5 system-xs-regular text-text-destructive"
-                  data-testid="workflow-process-error"
                 >
                   {data.error}
                 </div>
diff --git a/web/app/components/base/chat/chat/chat-input-area/__tests__/index.spec.tsx b/web/app/components/base/chat/chat/chat-input-area/__tests__/index.spec.tsx
index 7fa4623b3a4..7db45d3f9b4 100644
--- a/web/app/components/base/chat/chat/chat-input-area/__tests__/index.spec.tsx
+++ b/web/app/components/base/chat/chat/chat-input-area/__tests__/index.spec.tsx
@@ -340,6 +340,27 @@ describe('ChatInputArea', () => {
       expect(screen.getByRole('button', { name: 'Start build' })).toBeInTheDocument()
       expect(screen.queryByRole('button', { name: 'common.operation.send' })).not.toBeInTheDocument()
     })
+
+    it('should render the send button loading state when provided', async () => {
+      const user = userEvent.setup({ delay: null })
+      const onSend = vi.fn()
+      render(
+        <ChatInputArea
+          visionConfig={mockVisionConfig}
+          onSend={onSend}
+          sendButtonLabel="Start build"
+          sendButtonLoading
+        />,
+      )
+
+      await user.type(getTextarea()!, 'Build an agent')
+      const startBuildButton = screen.getByRole('button', { name: 'Start build' })
+
+      expect(startBuildButton).toHaveAttribute('aria-disabled', 'true')
+      expect(startBuildButton.querySelector('[aria-hidden="true"]')).toBeInTheDocument()
+      await user.click(startBuildButton)
+      expect(onSend).not.toHaveBeenCalled()
+    })
   })
 
   // -------------------------------------------------------------------------
@@ -735,6 +756,20 @@ describe('ChatInputArea', () => {
 
   // -------------------------------------------------------------------------
   describe('Feature Bar', () => {
+    it('should render footer notice with tooltip when provided', async () => {
+      const user = userEvent.setup({ delay: null })
+      const footerNotice = 'Agent runs in a Linux sandbox.'
+      const footerNoticeTooltip = 'For Dify Community Edition, each of your agents runs in a Linux 7.0.0-10060-aws sandbox environment within your docker. Your edits to the environment via Build Chats are persistent.'
+      render(<ChatInputArea visionConfig={mockVisionConfig} footerNotice={footerNotice} footerNoticeTooltip={footerNoticeTooltip} />)
+
+      expect(screen.getByText(footerNotice)).toBeInTheDocument()
+      expect(screen.queryByText(footerNoticeTooltip)).not.toBeInTheDocument()
+
+      await user.hover(screen.getByRole('button', { name: footerNoticeTooltip }))
+
+      expect(await screen.findByText(footerNoticeTooltip)).toBeInTheDocument()
+    })
+
     it('should render feature bar when showFeatureBar is true', () => {
       render(<ChatInputArea visionConfig={mockVisionConfig} showFeatureBar />)
       expect(screen.getByText(/feature.bar.empty/i)).toBeTruthy()
diff --git a/web/app/components/base/chat/chat/chat-input-area/index.tsx b/web/app/components/base/chat/chat/chat-input-area/index.tsx
index a30dddd2b2b..54a579540a0 100644
--- a/web/app/components/base/chat/chat/chat-input-area/index.tsx
+++ b/web/app/components/base/chat/chat/chat-input-area/index.tsx
@@ -1,9 +1,11 @@
+import type { ReactNode } from 'react'
 import type { Theme } from '../../embedded-chatbot/theme/theme-context'
 import type { EnableType, OnSend } from '../../types'
 import type { InputForm } from '../type'
 import type { FileUpload } from '@/app/components/base/features/types'
 import { cn } from '@langgenius/dify-ui/cn'
 import { toast } from '@langgenius/dify-ui/toast'
+import { Tooltip, TooltipContent, TooltipTrigger } from '@langgenius/dify-ui/tooltip'
 import { noop } from 'es-toolkit/function'
 import { decode } from 'html-entities'
 import Recorder from 'js-audio-recorder'
@@ -44,6 +46,9 @@ type ChatInputAreaProps = {
   isResponding?: boolean
   disabled?: boolean
   sendButtonLabel?: string
+  sendButtonLoading?: boolean
+  footerNotice?: ReactNode
+  footerNoticeTooltip?: ReactNode
   /**
    * Controls whether pressing Enter sends the message.
    * - true (default): Enter sends, Shift+Enter inserts newline
@@ -52,7 +57,7 @@ type ChatInputAreaProps = {
    */
   sendOnEnter?: boolean
 }
-const ChatInputArea = ({ readonly, botName, customPlaceholder, showFeatureBar, showFileUpload, featureBarReadonly = readonly, featureBarDisabled, onFeatureBarClick, visionConfig, speechToTextConfig = { enabled: true }, onSend, inputs = {}, inputsForm = [], theme, isResponding, disabled, sendButtonLabel, sendOnEnter = true }: ChatInputAreaProps) => {
+const ChatInputArea = ({ readonly, botName, customPlaceholder, showFeatureBar, showFileUpload, featureBarReadonly = readonly, featureBarDisabled, onFeatureBarClick, visionConfig, speechToTextConfig = { enabled: true }, onSend, inputs = {}, inputsForm = [], theme, isResponding, disabled, sendButtonLabel, sendButtonLoading, footerNotice, footerNoticeTooltip, sendOnEnter = true }: ChatInputAreaProps) => {
   const { t } = useTranslation()
   const { wrapperRef, textareaRef, textValueRef, holdSpaceRef, handleTextareaResize, isMultipleLine } = useTextAreaHeight()
   const [query, setQuery] = useState('')
@@ -161,7 +166,9 @@ const ChatInputArea = ({ readonly, botName, customPlaceholder, showFeatureBar, s
       toast.error(t('voiceInput.notAllow', { ns: 'common' }))
     })
   }, [t])
-  const operation = (<Operation ref={holdSpaceRef} readonly={readonly} fileConfig={visionConfig} speechToTextConfig={speechToTextConfig} onShowVoiceInput={handleShowVoiceInput} onSend={handleSend} sendButtonLabel={sendButtonLabel} disabled={!canSend} theme={theme} />)
+  const operation = (<Operation ref={holdSpaceRef} readonly={readonly} fileConfig={visionConfig} speechToTextConfig={speechToTextConfig} onShowVoiceInput={handleShowVoiceInput} onSend={handleSend} sendButtonLabel={sendButtonLabel} sendButtonLoading={sendButtonLoading} disabled={!canSend} theme={theme} />)
+  const shouldShowFooterNotice = footerNotice !== undefined && footerNotice !== null
+  const shouldShowFooterNoticeTooltip = footerNoticeTooltip !== undefined && footerNoticeTooltip !== null
   return (
     <>
       <div className={cn('pointer-events-auto relative z-10 overflow-hidden rounded-xl border border-components-chat-input-border bg-components-panel-bg-blur pb-[9px] shadow-md', isDragActive && 'border border-dashed border-components-option-card-option-selected-border', disabled && 'pointer-events-none border-components-panel-border opacity-50 shadow-none')}>
@@ -201,6 +208,31 @@ const ChatInputArea = ({ readonly, botName, customPlaceholder, showFeatureBar, s
         </div>
         {isMultipleLine && (<div className="px-[9px]">{operation}</div>)}
       </div>
+      {shouldShowFooterNotice && (
+        <div className="m-1 mt-0 -translate-y-2 rounded-b-[10px] border-r border-b border-l border-components-panel-border-subtle bg-util-colors-indigo-indigo-50 px-2.5 py-2 pt-4">
+          <div className="flex items-center gap-1">
+            {shouldShowFooterNoticeTooltip && (
+              <Tooltip>
+                <TooltipTrigger
+                  render={(
+                    <button
+                      type="button"
+                      className="flex size-5 items-center justify-center rounded-md text-text-accent hover:bg-state-base-hover focus-visible:outline focus-visible:outline-2 focus-visible:outline-offset-2 focus-visible:outline-components-button-primary-bg"
+                      aria-label={typeof footerNoticeTooltip === 'string' ? footerNoticeTooltip : undefined}
+                    >
+                      <span aria-hidden className="i-ri-information-line size-3.5 shrink-0" />
+                    </button>
+                  )}
+                />
+                <TooltipContent className="max-w-80">
+                  {footerNoticeTooltip}
+                </TooltipContent>
+              </Tooltip>
+            )}
+            <div className="body-xs-medium text-text-accent">{footerNotice}</div>
+          </div>
+        </div>
+      )}
       {showFeatureBar && (
         <div className="pointer-events-auto">
           <FeatureBar showFileUpload={showFileUpload} disabled={featureBarDisabled} onFeatureBarClick={featureBarReadonly ? noop : onFeatureBarClick} hideEditEntrance={featureBarReadonly} />
diff --git a/web/app/components/base/chat/chat/chat-input-area/operation.tsx b/web/app/components/base/chat/chat/chat-input-area/operation.tsx
index 542665cb574..bf6c2aa1381 100644
--- a/web/app/components/base/chat/chat/chat-input-area/operation.tsx
+++ b/web/app/components/base/chat/chat/chat-input-area/operation.tsx
@@ -18,6 +18,7 @@ type OperationProps = {
   onShowVoiceInput?: () => void
   onSend: () => void
   sendButtonLabel?: string
+  sendButtonLoading?: boolean
   disabled?: boolean
   theme?: Theme | null
   ref?: Ref<HTMLDivElement>
@@ -30,6 +31,7 @@ const Operation: FC<OperationProps> = ({
   onShowVoiceInput,
   onSend,
   sendButtonLabel,
+  sendButtonLoading,
   disabled,
   theme,
 }) => {
@@ -68,6 +70,7 @@ const Operation: FC<OperationProps> = ({
           )}
           variant="primary"
           disabled={readonly || disabled}
+          loading={sendButtonLoading}
           onClick={onSend}
           style={
             theme
diff --git a/web/app/components/base/chat/chat/hooks.ts b/web/app/components/base/chat/chat/hooks.ts
index dbb770af849..df1a1619042 100644
--- a/web/app/components/base/chat/chat/hooks.ts
+++ b/web/app/components/base/chat/chat/hooks.ts
@@ -52,6 +52,7 @@ type SendCallback = {
   onGetConversationMessages?: (conversationId: string, getAbortController: GetAbortController) => Promise<any>
   onGetSuggestedQuestions?: (responseItemId: string, getAbortController: GetAbortController) => Promise<any>
   onConversationComplete?: (conversationId: string, workflowRunId?: string) => void
+  onUnhandledEvent?: IOtherOptions['onUnhandledEvent']
   onSendSettled?: (hasError?: boolean) => void
   isPublicAPI?: boolean
 }
@@ -684,6 +685,7 @@ export const useChat = (
       onGetConversationMessages,
       onGetSuggestedQuestions,
       onConversationComplete,
+      onUnhandledEvent,
       onSendSettled,
       isPublicAPI,
     }: SendCallback,
@@ -774,6 +776,7 @@ export const useChat = (
 
     const otherOptions: IOtherOptions = {
       isPublicAPI,
+      onUnhandledEvent,
       getAbortController: (abortController) => {
         workflowEventsAbortControllerRef.current = abortController
       },
diff --git a/web/app/components/base/chat/chat/index.tsx b/web/app/components/base/chat/chat/index.tsx
index b48973c2963..733abcae322 100644
--- a/web/app/components/base/chat/chat/index.tsx
+++ b/web/app/components/base/chat/chat/index.tsx
@@ -72,6 +72,9 @@ export type ChatProps = {
   inputPlaceholder?: string
   inputPlaceholderBotName?: string
   sendButtonLabel?: string
+  sendButtonLoading?: boolean
+  footerNotice?: ReactNode
+  footerNoticeTooltip?: ReactNode
   sidebarCollapseState?: boolean
   hideAvatar?: boolean
   sendOnEnter?: boolean
@@ -121,6 +124,9 @@ const Chat: FC<ChatProps> = ({
   inputPlaceholder,
   inputPlaceholderBotName,
   sendButtonLabel,
+  sendButtonLoading,
+  footerNotice,
+  footerNoticeTooltip,
   sidebarCollapseState,
   hideAvatar,
   sendOnEnter,
@@ -267,6 +273,9 @@ const Chat: FC<ChatProps> = ({
                   isResponding={isResponding}
                   readonly={readonly}
                   sendButtonLabel={sendButtonLabel}
+                  sendButtonLoading={sendButtonLoading}
+                  footerNotice={footerNotice}
+                  footerNoticeTooltip={footerNoticeTooltip}
                   sendOnEnter={sendOnEnter}
                 />
               )
diff --git a/web/app/components/header/account-setting/model-provider-page/model-selector/__tests__/popup-item.spec.tsx b/web/app/components/header/account-setting/model-provider-page/model-selector/__tests__/popup-item.spec.tsx
index 291b903f62a..6fd2876c9cb 100644
--- a/web/app/components/header/account-setting/model-provider-page/model-selector/__tests__/popup-item.spec.tsx
+++ b/web/app/components/header/account-setting/model-provider-page/model-selector/__tests__/popup-item.spec.tsx
@@ -3,6 +3,7 @@ import type { DefaultModel, Model, ModelItem } from '../../declarations'
 import { Combobox } from '@langgenius/dify-ui/combobox'
 import { createPreviewCardHandle } from '@langgenius/dify-ui/preview-card'
 import { fireEvent, render, screen } from '@testing-library/react'
+import userEvent from '@testing-library/user-event'
 import {
   ConfigurationMethodEnum,
   CustomConfigurationStatusEnum,
@@ -27,16 +28,27 @@ vi.mock('../../hooks', async () => {
   }
 })
 
-vi.mock('../../model-badge', () => ({
-  default: ({ children }: { children: ReactNode }) => <span>{children}</span>,
-}))
-
 vi.mock('../../model-icon', () => ({
   default: ({ modelName }: { modelName: string }) => <span>{modelName}</span>,
 }))
 
 vi.mock('../../model-name', () => ({
-  default: ({ modelItem, nameClassName }: { modelItem: ModelItem, nameClassName?: string }) => <span className={nameClassName}>{modelItem.label.en_US}</span>,
+  default: ({
+    modelItem,
+    className,
+    nameClassName,
+    children,
+  }: {
+    modelItem: ModelItem
+    className?: string
+    nameClassName?: string
+    children?: ReactNode
+  }) => (
+    <span className={[className, nameClassName].filter(Boolean).join(' ')}>
+      {modelItem.label.en_US}
+      {children}
+    </span>
+  ),
 }))
 
 vi.mock('../feature-icon', () => ({
@@ -227,6 +239,52 @@ describe('PopupItem', () => {
     expect(screen.getByText('GPT-4')).toHaveClass('line-through')
   })
 
+  it('should render incompatible model names with tertiary text color', () => {
+    renderWithCombobox(
+      <PopupItem
+        {...previewCardProps()}
+        model={makeModel({
+          models: [
+            makeModelItem({ model: 'gpt-4o', label: { en_US: 'GPT-4o', zh_Hans: 'GPT-4o' } }),
+            makeModelItem({ model: 'gpt-4', label: { en_US: 'GPT-4', zh_Hans: 'GPT-4' } }),
+          ],
+        })}
+        modelPredicate={(_provider, modelItem) => modelItem.model !== 'gpt-4'}
+        onHide={vi.fn()}
+      />,
+    )
+
+    expect(screen.getByText('GPT-4o')).toHaveClass('text-text-secondary')
+    expect(screen.getByText('GPT-4o')).not.toHaveClass('text-text-quaternary')
+    expect(screen.getByText('GPT-4')).toHaveClass('text-text-quaternary')
+  })
+
+  it('should render suggestion icon with tooltip for suggested models', async () => {
+    renderWithCombobox(
+      <PopupItem
+        {...previewCardProps()}
+        model={makeModel({
+          models: [
+            makeModelItem({ model: 'gpt-5.5', label: { en_US: 'GPT-5.5', zh_Hans: 'GPT-5.5' } }),
+            makeModelItem({ model: 'gpt-5', label: { en_US: 'GPT-5', zh_Hans: 'GPT-5' } }),
+          ],
+        })}
+        modelSuggestionPredicate={(_provider, modelItem) => modelItem.model === 'gpt-5.5'}
+        onHide={vi.fn()}
+      />,
+    )
+
+    expect(screen.getByText('GPT-5.5')).toBeInTheDocument()
+    expect(screen.getByText('GPT-5')).toBeInTheDocument()
+    const suggestionIcon = screen.getByLabelText('common.modelProvider.selector.suggestionTip')
+
+    expect(suggestionIcon).toHaveClass('i-ri-shield-star-line')
+
+    await userEvent.hover(suggestionIcon)
+
+    expect(await screen.findByText('common.modelProvider.selector.suggestionTip')).toBeInTheDocument()
+  })
+
   it('should open model modal when clicking add on unconfigured model', () => {
     const onValueChange = vi.fn()
     const { rerender } = renderWithCombobox(
diff --git a/web/app/components/header/account-setting/model-provider-page/model-selector/index.tsx b/web/app/components/header/account-setting/model-provider-page/model-selector/index.tsx
index fc86e82927c..978eb7e22dc 100644
--- a/web/app/components/header/account-setting/model-provider-page/model-selector/index.tsx
+++ b/web/app/components/header/account-setting/model-provider-page/model-selector/index.tsx
@@ -37,6 +37,7 @@ type ModelSelectorProps = {
   providerSettingsSource?: 'agent'
   showModelMeta?: boolean
   modelPredicate?: ModelSelectorModelPredicate
+  modelSuggestionPredicate?: ModelSelectorModelPredicate
 }
 function ModelSelector({
   defaultModel,
@@ -55,6 +56,7 @@ function ModelSelector({
   providerSettingsSource,
   showModelMeta,
   modelPredicate,
+  modelSuggestionPredicate,
 }: ModelSelectorProps) {
   const { t } = useTranslation()
   const [open, setOpen] = useState(false)
@@ -173,6 +175,7 @@ function ModelSelector({
           hideProviderSettingsFooter={hideProviderSettingsFooter}
           providerSettingsSource={providerSettingsSource}
           modelPredicate={modelPredicate}
+          modelSuggestionPredicate={modelSuggestionPredicate}
           onConfigureEmptyState={onConfigureEmptyState ? handleConfigureEmptyState : undefined}
           onOpenMarketplace={onOpenMarketplace}
           onInputValueChange={setInputValue}
diff --git a/web/app/components/header/account-setting/model-provider-page/model-selector/popup-item.tsx b/web/app/components/header/account-setting/model-provider-page/model-selector/popup-item.tsx
index f0d99838c55..0eb8182b718 100644
--- a/web/app/components/header/account-setting/model-provider-page/model-selector/popup-item.tsx
+++ b/web/app/components/header/account-setting/model-provider-page/model-selector/popup-item.tsx
@@ -1,10 +1,12 @@
 import type { ComponentProps } from 'react'
 import type { DefaultModel, Model, ModelItem } from '../declarations'
+import type { ModelSelectorModelPredicate } from './types'
 import { cn } from '@langgenius/dify-ui/cn'
 import { ComboboxGroup, ComboboxItem, ComboboxItemIndicator } from '@langgenius/dify-ui/combobox'
 import { Popover, PopoverContent, PopoverTrigger } from '@langgenius/dify-ui/popover'
 import { PreviewCardTrigger } from '@langgenius/dify-ui/preview-card'
 import { StatusDot } from '@langgenius/dify-ui/status-dot'
+import { Tooltip, TooltipContent, TooltipTrigger } from '@langgenius/dify-ui/tooltip'
 import { useCallback, useMemo, useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { CreditsCoin } from '@/app/components/base/icons/src/vender/line/financeAndECommerce'
@@ -29,6 +31,8 @@ type PreviewCardHandle = NonNullable<ComponentProps<typeof PreviewCardTrigger>['
 type PopupItemProps = {
   defaultModel?: DefaultModel
   model: Model
+  modelPredicate?: ModelSelectorModelPredicate
+  modelSuggestionPredicate?: ModelSelectorModelPredicate
   previewCardHandle: PreviewCardHandle
   onPreviewCardClose: () => void
   onHide: () => void
@@ -36,6 +40,8 @@ type PopupItemProps = {
 function PopupItem({
   defaultModel,
   model,
+  modelPredicate,
+  modelSuggestionPredicate,
   previewCardHandle,
   onPreviewCardClose,
   onHide,
@@ -44,6 +50,7 @@ function PopupItem({
   const [dropdownOpen, setDropdownOpen] = useState(false)
   const { t } = useTranslation()
   const language = useLanguage()
+  const suggestionTip = t('modelProvider.selector.suggestionTip', { ns: 'common' })
   const { setShowModelModal } = useModalContext()
   const { modelProviders } = useProviderContext()
   const updateModelList = useUpdateModelList()
@@ -156,6 +163,8 @@ function PopupItem({
         </Popover>
       </div>
       {!collapsed && model.models.map((modelItem) => {
+        const isModelCompatible = modelPredicate?.(model, modelItem) ?? true
+        const isModelSuggested = modelSuggestionPredicate?.(model, modelItem) ?? false
         const rowClassName = cn(
           'group relative mx-1 flex h-8 min-w-0 items-center gap-1 rounded-lg px-3 py-1.5 text-left',
           modelItem.status === ModelStatusEnum.active ? 'cursor-pointer hover:bg-state-base-hover' : 'cursor-not-allowed hover:bg-state-base-hover-alt',
@@ -169,10 +178,30 @@ function PopupItem({
                 modelName={modelItem.model}
               />
               <ModelName
-                className={cn('system-sm-medium text-text-secondary', modelItem.status !== ModelStatusEnum.active && 'opacity-60')}
+                className={cn(
+                  'system-sm-medium text-text-secondary',
+                  !isModelCompatible && 'text-text-quaternary',
+                  modelItem.status !== ModelStatusEnum.active && 'opacity-60',
+                )}
                 modelItem={modelItem}
                 nameClassName={modelItem.deprecated ? 'line-through' : undefined}
-              />
+              >
+                {isModelSuggested && (
+                  <Tooltip>
+                    <TooltipTrigger
+                      render={(
+                        <span
+                          aria-label={suggestionTip}
+                          className="i-ri-shield-star-line size-3.5 shrink-0 text-text-accent-secondary"
+                        />
+                      )}
+                    />
+                    <TooltipContent placement="top">
+                      {suggestionTip}
+                    </TooltipContent>
+                  </Tooltip>
+                )}
+              </ModelName>
             </div>
             {
               defaultModel?.model === modelItem.model && defaultModel.provider === currentProvider.provider && (
diff --git a/web/app/components/header/account-setting/model-provider-page/model-selector/popup.tsx b/web/app/components/header/account-setting/model-provider-page/model-selector/popup.tsx
index 7b8c0bf4857..372317ff3c4 100644
--- a/web/app/components/header/account-setting/model-provider-page/model-selector/popup.tsx
+++ b/web/app/components/header/account-setting/model-provider-page/model-selector/popup.tsx
@@ -42,6 +42,7 @@ export type PopupProps = {
   hideProviderSettingsFooter?: boolean
   providerSettingsSource?: 'agent'
   modelPredicate?: ModelSelectorModelPredicate
+  modelSuggestionPredicate?: ModelSelectorModelPredicate
   onConfigureEmptyState?: () => void
   onInputValueChange: (value: string) => void
   onOpenMarketplace?: () => void
@@ -55,6 +56,7 @@ function Popup({
   hideProviderSettingsFooter,
   providerSettingsSource,
   modelPredicate,
+  modelSuggestionPredicate,
   onConfigureEmptyState,
   onInputValueChange,
   onOpenMarketplace,
@@ -212,6 +214,8 @@ function Popup({
                   key={model.provider}
                   defaultModel={defaultModel}
                   model={model}
+                  modelPredicate={modelPredicate}
+                  modelSuggestionPredicate={modelSuggestionPredicate}
                   previewCardHandle={previewCardHandle}
                   onPreviewCardClose={handleClosePreviewCard}
                   onHide={onHide}
diff --git a/web/app/components/plugins/plugin-auth/authorize/add-oauth-button.tsx b/web/app/components/plugins/plugin-auth/authorize/add-oauth-button.tsx
index 1882b1eb89d..144c164ec02 100644
--- a/web/app/components/plugins/plugin-auth/authorize/add-oauth-button.tsx
+++ b/web/app/components/plugins/plugin-auth/authorize/add-oauth-button.tsx
@@ -31,6 +31,11 @@ export type AddOAuthButtonProps = {
   dividerClassName?: string
   disabled?: boolean
   onUpdate?: () => void
+  renderTrigger?: (props: {
+    disabled?: boolean
+    isConfigured: boolean
+    onClick: () => void
+  }) => React.ReactNode
   oAuthData?: {
     schema?: FormSchema[]
     is_oauth_custom_client_enabled?: boolean
@@ -51,6 +56,7 @@ const AddOAuthButton = ({
   dividerClassName,
   disabled,
   onUpdate,
+  renderTrigger,
   oAuthData,
 }: AddOAuthButtonProps) => {
   const { t } = useTranslation()
@@ -67,8 +73,8 @@ const AddOAuthButton = ({
   }, [oAuthData, data])
   const {
     schema = [],
-    is_oauth_custom_client_enabled,
-    is_system_oauth_params_exists,
+    is_oauth_custom_client_enabled = false,
+    is_system_oauth_params_exists = false,
     client_params = {},
     redirect_uri,
   } = mergedOAuthData
@@ -186,7 +192,14 @@ const AddOAuthButton = ({
   return (
     <>
       {
-        isConfigured && (
+        renderTrigger?.({
+          disabled,
+          isConfigured,
+          onClick: isConfigured ? handleOAuth : openOAuthSettings,
+        })
+      }
+      {
+        !renderTrigger && isConfigured && (
           <div className={cn('flex w-full', className)}>
             <Button
               variant={buttonVariant}
@@ -239,7 +252,7 @@ const AddOAuthButton = ({
         )
       }
       {
-        !isConfigured && (
+        !renderTrigger && !isConfigured && (
           <Button
             variant={buttonVariant}
             onClick={openOAuthSettings}
diff --git a/web/app/components/workflow/block-selector/__tests__/blocks.spec.tsx b/web/app/components/workflow/block-selector/__tests__/blocks.spec.tsx
index 47d385e41b1..75c1b30d434 100644
--- a/web/app/components/workflow/block-selector/__tests__/blocks.spec.tsx
+++ b/web/app/components/workflow/block-selector/__tests__/blocks.spec.tsx
@@ -160,7 +160,10 @@ describe('Blocks', () => {
       </QueryClientProvider>,
     )
 
-    await user.hover(screen.getByRole('button', { name: 'Agent' }))
+    const agentBlock = screen.getByRole('button', { name: /Agent/ })
+    expect(agentBlock).toHaveTextContent('common.menus.status')
+
+    await user.hover(agentBlock)
 
     expect(await screen.findByRole('dialog', { name: 'agentV2.roster.nodeSelector.dialogLabel' })).toBeInTheDocument()
   })
@@ -226,7 +229,7 @@ describe('Blocks', () => {
       screen.getByText('Agent').compareDocumentPosition(screen.getByText('LLM')) & Node.DOCUMENT_POSITION_FOLLOWING,
     ).toBeTruthy()
 
-    await user.click(screen.getByRole('button', { name: 'Agent' }))
+    await user.click(screen.getByRole('button', { name: /Agent/ }))
 
     expect(await screen.findByRole('dialog', { name: 'agentV2.roster.nodeSelector.dialogLabel' })).toBeInTheDocument()
     expect(screen.getByRole('combobox', { name: 'agentV2.roster.searchLabel' })).toBeInTheDocument()
@@ -308,7 +311,7 @@ describe('Blocks', () => {
       </QueryClientProvider>,
     )
 
-    await user.click(screen.getByRole('button', { name: 'Agent' }))
+    await user.click(screen.getByRole('button', { name: /Agent/ }))
     expect(await screen.findByText('Nadia')).toBeInTheDocument()
 
     await user.click(screen.getByRole('option', { name: 'Nadia Researcher' }))
@@ -355,7 +358,7 @@ describe('Blocks', () => {
       </QueryClientProvider>,
     )
 
-    await user.click(screen.getByRole('button', { name: 'Agent' }))
+    await user.click(screen.getByRole('button', { name: /Agent/ }))
     const consoleLink = await screen.findByRole('option', { name: 'agentV2.roster.nodeSelector.manageInAgentConsole' })
     expect(consoleLink).toHaveAttribute('href', '/roster')
     expect(consoleLink).toHaveAttribute('target', '_blank')
@@ -408,7 +411,7 @@ describe('Blocks', () => {
       </QueryClientProvider>,
     )
 
-    await user.click(screen.getByRole('button', { name: 'Agent' }))
+    await user.click(screen.getByRole('button', { name: /Agent/ }))
 
     expect(await screen.findByRole('dialog', { name: 'agentV2.roster.nodeSelector.dialogLabel' })).toBeInTheDocument()
 
diff --git a/web/app/components/workflow/block-selector/__tests__/tool-picker.spec.tsx b/web/app/components/workflow/block-selector/__tests__/tool-picker.spec.tsx
index b5949265cad..85d3adccc68 100644
--- a/web/app/components/workflow/block-selector/__tests__/tool-picker.spec.tsx
+++ b/web/app/components/workflow/block-selector/__tests__/tool-picker.spec.tsx
@@ -55,6 +55,14 @@ vi.mock('@/context/i18n', () => ({
   useGetLanguage: vi.fn(),
 }))
 
+vi.mock('@/config', async (importOriginal) => {
+  const actual = await importOriginal<typeof import('@/config')>()
+  return {
+    ...actual,
+    MARKETPLACE_URL_PREFIX: 'https://marketplace.test',
+  }
+})
+
 vi.mock('@/context/app-context', () => ({
   useSelector: <T,>(selector: (state: { workspacePermissionKeys: string[] }) => T): T => selector({
     workspacePermissionKeys: mockWorkspacePermissionKeys,
@@ -191,7 +199,7 @@ vi.mock('@/utils/var', async (importOriginal) => {
   const actual = await importOriginal<typeof import('@/utils/var')>()
   return {
     ...actual,
-    getMarketplaceUrl: () => 'https://marketplace.test/tools',
+    getMarketplaceUrl: (path = '') => `https://marketplace.test${path}`,
   }
 })
 
@@ -404,6 +412,18 @@ describe('ToolPicker', () => {
     expect(disabledOnShowChange).not.toHaveBeenCalled()
   })
 
+  it('should link the find-more footer to the marketplace tool category', () => {
+    renderToolPicker({
+      isShow: true,
+      selectedTools: [],
+    })
+
+    expect(screen.getByRole('link', { name: /plugin\.findMoreInMarketplace/i })).toHaveAttribute(
+      'href',
+      'https://marketplace.test/plugins/tool',
+    )
+  })
+
   it('should render real search and tool lists, then forward tool selections', async () => {
     const user = userEvent.setup()
     const onSelect = vi.fn()
diff --git a/web/app/components/workflow/block-selector/agent-selector.tsx b/web/app/components/workflow/block-selector/agent-selector.tsx
index b6725f47ccc..f6d7c6db5d0 100644
--- a/web/app/components/workflow/block-selector/agent-selector.tsx
+++ b/web/app/components/workflow/block-selector/agent-selector.tsx
@@ -24,6 +24,7 @@ import { useDebounce } from 'ahooks'
 import { useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import AppIcon from '@/app/components/base/app-icon'
+import Badge from '@/app/components/base/badge'
 import { useHooksStore } from '@/app/components/workflow/hooks-store'
 import Link from '@/next/link'
 import { consoleQuery } from '@/service/client'
@@ -308,7 +309,7 @@ export function AgentBlockItem({
   onSelect: (agent: AgentRosterNodeData) => void
   onStartFromScratch: () => void
 }) {
-  const { t } = useTranslation('agentV2')
+  const { t } = useTranslation(['agentV2', 'common'])
   const [open, setOpen] = useState(false)
   const handleSelect = (agent: AgentRosterNodeData) => {
     setOpen(false)
@@ -331,6 +332,12 @@ export function AgentBlockItem({
             <span className="min-w-0 grow truncate system-sm-medium text-text-secondary">
               {block.metaData.title}
             </span>
+            <Badge
+              size="xs"
+              variant="dimm"
+              text={t('menus.status', { ns: 'common' })}
+              className="ml-2 shrink-0"
+            />
             <span aria-hidden className="i-custom-vender-solid-general-arrow-down-round-fill size-4 shrink-0 -rotate-90 text-text-tertiary" />
           </button>
         )}
@@ -341,7 +348,7 @@ export function AgentBlockItem({
         popupClassName="border-none bg-transparent p-0 shadow-none backdrop-blur-none"
       >
         <PopoverTitle className="sr-only">
-          {t('roster.nodeSelector.dialogLabel')}
+          {t('roster.nodeSelector.dialogLabel', { ns: 'agentV2' })}
         </PopoverTitle>
         <AgentSelectorContent
           open={open}
diff --git a/web/app/components/workflow/hooks/__tests__/use-nodes-interactions.spec.ts b/web/app/components/workflow/hooks/__tests__/use-nodes-interactions.spec.ts
index f0ce89a7018..95591c7a844 100644
--- a/web/app/components/workflow/hooks/__tests__/use-nodes-interactions.spec.ts
+++ b/web/app/components/workflow/hooks/__tests__/use-nodes-interactions.spec.ts
@@ -548,6 +548,73 @@ describe('useNodesInteractions', () => {
     expect(mockHandleSyncWorkflowDraft).toHaveBeenCalledWith(true, true)
   })
 
+  it('creates an inline agent binding when Agent v2 default data is pending inline', () => {
+    currentNodes = [
+      createNode({
+        id: 'node-1',
+        width: 100,
+        data: {
+          type: BlockEnum.Code,
+          title: 'Code',
+          desc: '',
+        },
+      }),
+    ]
+    rfState.nodes = currentNodes as unknown as typeof rfState.nodes
+    rfState.edges = []
+    rfState.setNodes.mockImplementation((nextNodes) => {
+      rfState.nodes = nextNodes
+    })
+    runtimeNodesMetaDataMap.value = {
+      [BlockEnum.AgentV2]: {
+        defaultValue: {
+          type: BlockEnum.AgentV2,
+          title: 'Agent',
+          desc: '',
+          agent_binding: {
+            binding_type: 'inline_agent',
+          },
+          agent_node_kind: 'dify_agent',
+          version: '2',
+        },
+        metaData: {
+          isSingleton: false,
+        },
+      },
+    }
+
+    const { result } = renderWorkflowHook(() => useNodesInteractions(), {
+      historyStore: {
+        nodes: currentNodes,
+        edges: [],
+      },
+    })
+
+    act(() => {
+      result.current.handleNodeAdd(
+        {
+          nodeType: BlockEnum.AgentV2,
+        },
+        { prevNodeId: 'node-1' },
+      )
+    })
+
+    const agentNode = rfState.nodes.find(node => node.data.type === BlockEnum.AgentV2)
+    const firstSetNodesPayload = rfState.setNodes.mock.calls[0]?.[0]
+    const pendingAgentNode = firstSetNodesPayload.find((node: Node) => node.data.type === BlockEnum.AgentV2)
+
+    expect(pendingAgentNode?.data._isTempNode).toBe(true)
+    expect(agentNode?.data.agent_binding).toEqual({
+      binding_type: 'inline_agent',
+      agent_id: 'inline-agent-1',
+      current_snapshot_id: 'inline-snapshot-1',
+    })
+    expect(mockCreateInlineAgentBinding).toHaveBeenCalledWith(agentNode?.id, expect.objectContaining({
+      onSuccess: expect.any(Function),
+    }))
+    expect(mockHandleSyncWorkflowDraft).toHaveBeenCalledWith(true, true)
+  })
+
   it('cancels selection state with collaborative nodes snapshot', () => {
     currentNodes = [
       createNode({
diff --git a/web/app/components/workflow/hooks/use-nodes-interactions.ts b/web/app/components/workflow/hooks/use-nodes-interactions.ts
index f06a0288edf..a9eea0c2995 100644
--- a/web/app/components/workflow/hooks/use-nodes-interactions.ts
+++ b/web/app/components/workflow/hooks/use-nodes-interactions.ts
@@ -88,16 +88,29 @@ const ENTRY_NODE_WRAPPER_OFFSET = {
   y: 21, // Adjusted based on visual testing feedback
 } as const
 
-function needsPendingInlineAgentBinding(defaultValue?: BlockDefaultValue) {
-  if (!defaultValue || !('agent_binding' in defaultValue))
+function needsPendingInlineAgentBinding(defaultValue?: unknown) {
+  if (!defaultValue || typeof defaultValue !== 'object' || !('agent_binding' in defaultValue))
     return false
 
-  const binding = defaultValue.agent_binding
+  const binding = (defaultValue as {
+    agent_binding?: {
+      agent_id?: string
+      binding_type?: string
+      current_snapshot_id?: string
+    }
+  }).agent_binding
 
-  return binding.binding_type === 'inline_agent'
+  return binding?.binding_type === 'inline_agent'
     && (!binding.agent_id || !binding.current_snapshot_id)
 }
 
+function agentV2NodeDefaultsNeedInlineBinding(defaultValue?: unknown, pluginDefaultValue?: unknown) {
+  return needsPendingInlineAgentBinding({
+    ...(defaultValue && typeof defaultValue === 'object' ? defaultValue : {}),
+    ...(pluginDefaultValue && typeof pluginDefaultValue === 'object' ? pluginDefaultValue : {}),
+  })
+}
+
 const pruneClipboardNodesWithFilteredAncestors = (
   sourceNodes: Node[],
   candidateNodes: Node[],
@@ -926,7 +939,8 @@ export const useNodesInteractions = () => {
         return
       const { defaultValue } = nodeMetaData
       const nodesWithSameType = getNodesWithSameDefaultDataType(nodes, nodeType, defaultValue)
-      const shouldCreateInlineAgentBinding = nodeType === BlockEnum.AgentV2 && needsPendingInlineAgentBinding(pluginDefaultValue)
+      const shouldCreateInlineAgentBinding = nodeType === BlockEnum.AgentV2
+        && agentV2NodeDefaultsNeedInlineBinding(defaultValue, pluginDefaultValue)
       const { newNode, newIterationStartNode, newLoopStartNode }
         = generateNewNode({
           type: getNodeCustomTypeByNodeDataType(nodeType),
@@ -1506,7 +1520,8 @@ export const useNodesInteractions = () => {
         return
       const { defaultValue } = nodeMetaData
       const nodesWithSameType = getNodesWithSameDefaultDataType(nodes, nodeType, defaultValue)
-      const shouldCreateInlineAgentBinding = nodeType === BlockEnum.AgentV2 && needsPendingInlineAgentBinding(pluginDefaultValue)
+      const shouldCreateInlineAgentBinding = nodeType === BlockEnum.AgentV2
+        && agentV2NodeDefaultsNeedInlineBinding(defaultValue, pluginDefaultValue)
       const {
         newNode: newCurrentNode,
         newIterationStartNode,
diff --git a/web/app/components/workflow/nodes/agent-v2/__tests__/default.spec.ts b/web/app/components/workflow/nodes/agent-v2/__tests__/default.spec.ts
index ed7d9b23395..7460739778f 100644
--- a/web/app/components/workflow/nodes/agent-v2/__tests__/default.spec.ts
+++ b/web/app/components/workflow/nodes/agent-v2/__tests__/default.spec.ts
@@ -93,6 +93,9 @@ describe('agent/default', () => {
 
   it('creates Agent v2 graph data by default', () => {
     expect(nodeDefault.defaultValue).toMatchObject({
+      agent_binding: {
+        binding_type: 'inline_agent',
+      },
       agent_node_kind: 'dify_agent',
       version: '2',
     })
diff --git a/web/app/components/workflow/nodes/agent-v2/components/__tests__/agent-orchestrate-panel-content.spec.tsx b/web/app/components/workflow/nodes/agent-v2/components/__tests__/agent-orchestrate-panel-content.spec.tsx
index ba6dd76ee08..ecfdf41932e 100644
--- a/web/app/components/workflow/nodes/agent-v2/components/__tests__/agent-orchestrate-panel-content.spec.tsx
+++ b/web/app/components/workflow/nodes/agent-v2/components/__tests__/agent-orchestrate-panel-content.spec.tsx
@@ -194,6 +194,14 @@ vi.mock('@/service/client', () => ({
           },
         },
         sandbox: {
+          get: {
+            queryOptions: () => ({
+              queryKey: ['sandbox-info'],
+              queryFn: () => Promise.resolve({
+                workspace_cwd: '.',
+              }),
+            }),
+          },
           files: {
             get: {
               queryOptions: () => ({
@@ -406,9 +414,24 @@ describe('WorkflowInlineAgentConfigureWorkspace', () => {
         }),
       })
 
+      fireEvent.click(await screen.findByRole('button', {
+        name: 'send build message',
+      }))
+      await waitFor(() => expect(screen.getByRole('region', { name: 'build-chat' })).toHaveTextContent('build:build-conversation-new'))
+      expect(screen.queryByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.open',
+      })).not.toBeInTheDocument()
+
       fireEvent.click(await screen.findByRole('button', {
         name: 'complete build conversation',
       }))
+      fireEvent.click(await screen.findByRole('button', {
+        name: 'send build message',
+      }))
+      expect(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.open',
+      })).toBeInTheDocument()
+
       fireEvent.click(await screen.findByRole('button', {
         name: 'agentV2.agentDetail.configure.workingDirectory.open',
       }))
diff --git a/web/app/components/workflow/nodes/agent-v2/components/__tests__/save-inline-agent-to-roster-dialog.spec.tsx b/web/app/components/workflow/nodes/agent-v2/components/__tests__/save-inline-agent-to-roster-dialog.spec.tsx
index 0d48cd93064..131378d292b 100644
--- a/web/app/components/workflow/nodes/agent-v2/components/__tests__/save-inline-agent-to-roster-dialog.spec.tsx
+++ b/web/app/components/workflow/nodes/agent-v2/components/__tests__/save-inline-agent-to-roster-dialog.spec.tsx
@@ -117,7 +117,7 @@ describe('SaveInlineAgentToRosterDialog', () => {
     const dialog = screen.getByRole('dialog', { name: 'agentV2.roster.saveToRosterDialog.title' })
     const nameInput = within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.nameLabel' })
     expect(nameInput).toHaveValue('')
-    expect(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' })).toHaveValue('Tender Analyst')
+    expect(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel common.label.optional' })).toHaveValue('Tender Analyst')
     expect(within(dialog).getByPlaceholderText('agentV2.roster.createForm.descriptionPlaceholder')).toHaveValue('Drafts tender clarifications.')
 
     await user.type(nameInput, 'Roster Tender Agent')
diff --git a/web/app/components/workflow/nodes/agent-v2/components/agent-orchestrate-panel-content.tsx b/web/app/components/workflow/nodes/agent-v2/components/agent-orchestrate-panel-content.tsx
index 7b7e6fe0bb5..231ed9c74ba 100644
--- a/web/app/components/workflow/nodes/agent-v2/components/agent-orchestrate-panel-content.tsx
+++ b/web/app/components/workflow/nodes/agent-v2/components/agent-orchestrate-panel-content.tsx
@@ -257,6 +257,7 @@ function WorkflowInlineAgentConfigureWorkspaceContent({
   const composerState = inlineComposerState
   const [buildDraftActionsDisabled, setBuildDraftActionsDisabled] = useState(false)
   const [clearPreviewChat, setClearPreviewChat] = useState(false)
+  const [completedBuildConversationId, setCompletedBuildConversationId] = useState<string | null>(null)
   const [workflowRunId, setWorkflowRunId] = useState<string | null>(null)
   const conversationIds = useAtomValue(agentConfigureConversationIdsAtom)
   const rightPanelChatMode = useAtomValue(agentConfigureRightPanelChatModeAtom)
@@ -347,6 +348,7 @@ function WorkflowInlineAgentConfigureWorkspaceContent({
   }, [refreshDebugConversationInput, refreshDebugConversationRequestAsync])
   const resetBuildChatSession = useCallback(async () => {
     await refreshDebugConversationAsync().catch(() => undefined)
+    setCompletedBuildConversationId(null)
     setConversationId({ mode: 'build', conversationId: null })
     setWorkflowRunId(null)
     setClearPreviewChat(true)
@@ -467,6 +469,14 @@ function WorkflowInlineAgentConfigureWorkspaceContent({
     || isApplyingInlineBuildDraft
     || discardBuildDraftMutation.isPending
     || isRefreshingDebugConversation
+  const buildConversationHasAgentResponse = !!conversationIds.build && (
+    conversationIds.build === completedBuildConversationId
+    || (
+      conversationIds.build === inlineComposerState?.debug_conversation_id
+      && (inlineComposerState?.debug_conversation_has_messages ?? false)
+    )
+  )
+  const showWorkingDirectoryAction = rightPanelChatMode === 'build' && buildConversationHasAgentResponse
   const restartCurrentChat = () => {
     if (isRestartCurrentChatDisabled)
       return
@@ -539,6 +549,7 @@ function WorkflowInlineAgentConfigureWorkspaceContent({
               onOpenWorkingDirectory={workingDirectoryPanel.openWorkingDirectory}
               onRefresh={restartCurrentChat}
               refreshDisabled={isRestartCurrentChatDisabled}
+              showWorkingDirectoryAction={showWorkingDirectoryAction}
               showChatFeaturesAction={false}
               trailingAction={(
                 <button
@@ -567,6 +578,7 @@ function WorkflowInlineAgentConfigureWorkspaceContent({
               onClearChatListChange={setClearPreviewChat}
               onConversationComplete={(mode, completedConversationId, completedWorkflowRunId) => {
                 if (mode === 'build') {
+                  setCompletedBuildConversationId(completedConversationId)
                   setWorkflowRunId(completedWorkflowRunId ?? completedConversationId)
                   invalidateAgentWorkingDirectoryFiles({
                     agentId,
diff --git a/web/app/components/workflow/nodes/agent-v2/default.ts b/web/app/components/workflow/nodes/agent-v2/default.ts
index d1575b32f07..b1273030112 100644
--- a/web/app/components/workflow/nodes/agent-v2/default.ts
+++ b/web/app/components/workflow/nodes/agent-v2/default.ts
@@ -12,6 +12,9 @@ const metaData = genNodeMetaData({
 const nodeDefault: NodeDefault<AgentV2NodeType> = {
   metaData,
   defaultValue: {
+    agent_binding: {
+      binding_type: 'inline_agent',
+    },
     agent_node_kind: 'dify_agent',
     version: '2',
   },
diff --git a/web/features/agent-v2/agent-composer/__tests__/store.spec.ts b/web/features/agent-v2/agent-composer/__tests__/store.spec.ts
index de441d8fcba..480c0d115ea 100644
--- a/web/features/agent-v2/agent-composer/__tests__/store.spec.ts
+++ b/web/features/agent-v2/agent-composer/__tests__/store.spec.ts
@@ -423,6 +423,48 @@ describe('agent composer store conversions', () => {
     ])
   })
 
+  it('should hydrate oauth tool authorization state from credential refs', () => {
+    const formState = agentSoulConfigToFormState({
+      tools: {
+        dify_tools: [
+          {
+            provider: 'google',
+            provider_id: 'google',
+            provider_type: 'builtin',
+            tool_name: 'search',
+            credential_type: 'oauth2',
+          },
+          {
+            provider: 'slack',
+            provider_id: 'slack',
+            provider_type: 'builtin',
+            tool_name: 'post_message',
+            credential_type: 'oauth2',
+            credential_ref: {
+              id: 'slack-oauth',
+              provider: 'slack',
+              type: 'provider',
+            },
+          },
+        ],
+      },
+    })
+
+    expect(formState.tools).toEqual([
+      expect.objectContaining({
+        id: 'google',
+        credentialType: 'oauth2',
+        credentialVariant: 'unauthorized',
+      }),
+      expect.objectContaining({
+        id: 'slack',
+        credentialId: 'slack-oauth',
+        credentialType: 'oauth2',
+        credentialVariant: 'authorized',
+      }),
+    ])
+  })
+
   it('should not save credentialed tool config without a credential reference', () => {
     const baseConfig = {
       tools: {
diff --git a/web/features/agent-v2/agent-composer/conversions.ts b/web/features/agent-v2/agent-composer/conversions.ts
index 9e9ec8c359c..046f4962a2b 100644
--- a/web/features/agent-v2/agent-composer/conversions.ts
+++ b/web/features/agent-v2/agent-composer/conversions.ts
@@ -174,6 +174,12 @@ const getDifyToolActionId = (tool: AgentSoulDifyToolConfig) => `${tool.provider_
 const toCredentialVariant = (tool: AgentSoulDifyToolConfig) => {
   const credentialType = tool.credential_type
 
+  if (credentialType === 'api-key')
+    return 'authorized' as const
+
+  if (credentialType === 'oauth2')
+    return tool.credential_ref?.id ? 'authorized' as const : 'unauthorized' as const
+
   if (credentialType === 'unauthorized')
     return 'unauthorized' as const
 
@@ -221,9 +227,9 @@ const toProviderToolFormState = (config?: AgentSoulConfig): {
       kind: 'provider',
       iconClassName: 'i-custom-public-other-default-tool-icon text-text-tertiary',
       providerType: tool.provider_type,
-      allowDelete: tool.credential_type === 'api-key' || tool.credential_type === 'unauthorized',
+      allowDelete: tool.credential_type === 'api-key' || tool.credential_type === 'oauth2' || tool.credential_type === 'unauthorized',
       credentialId: tool.credential_ref?.id ?? undefined,
-      credentialKey: tool.credential_type === 'api-key'
+      credentialKey: tool.credential_type === 'api-key' || tool.credential_type === 'oauth2'
         ? 'agentDetail.configure.tools.credential.authOne'
         : undefined,
       credentialType: tool.credential_type,
@@ -544,7 +550,7 @@ export const formStateToAgentSoulConfig = ({
     env: toEnvConfig(formState.envVariables),
     config_skills: toConfigSkillConfigs(formState.skills, baseConfig),
     config_files: toConfigFileConfigs(formState.files, baseConfig),
-    config_note: baseConfig?.config_note ?? '',
+    config_note: formState.configNote,
   }
 }
 
@@ -556,6 +562,7 @@ export const agentSoulConfigToFormState = (
 
   return {
     ...baseDraft,
+    configNote: config?.config_note ?? '',
     prompt: config?.prompt?.system_prompt ?? '',
     model: toDraftModel(config),
     appFeatures: config?.app_features,
diff --git a/web/features/agent-v2/agent-composer/form-state.ts b/web/features/agent-v2/agent-composer/form-state.ts
index fc61e968672..257842e4406 100644
--- a/web/features/agent-v2/agent-composer/form-state.ts
+++ b/web/features/agent-v2/agent-composer/form-state.ts
@@ -108,6 +108,7 @@ export type AgentCliTool = AgentToolBase & {
 export type AgentTool = AgentProviderTool | AgentCliTool
 
 export type AgentSoulConfigFormState = {
+  configNote: string
   prompt: string
   model?: AgentComposerModel
   appFeatures?: AgentSoulAppFeaturesConfig
@@ -120,6 +121,7 @@ export type AgentSoulConfigFormState = {
 }
 
 export const defaultAgentSoulConfigFormState: AgentSoulConfigFormState = {
+  configNote: '',
   prompt: '',
   skills: [],
   files: [],
diff --git a/web/features/agent-v2/agent-composer/knowledge-validation.ts b/web/features/agent-v2/agent-composer/knowledge-validation.ts
index 3d2c89c4318..76498606ba0 100644
--- a/web/features/agent-v2/agent-composer/knowledge-validation.ts
+++ b/web/features/agent-v2/agent-composer/knowledge-validation.ts
@@ -56,10 +56,11 @@ export const useKnowledgeValidationMessage = () => {
           field: t('agentDetail.configure.knowledgeRetrieval.dialog.query.customInputLabel'),
         })
       case 'single_model_required':
-      case 'metadata_model_required':
         return tCommon('errorMsg.fieldRequired', {
           field: tCommon('modelProvider.systemReasoningModel.key'),
         })
+      case 'metadata_model_required':
+        return t('agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired')
       case 'metadata_conditions_required':
         return tCommon('errorMsg.fieldRequired', {
           field: tWorkflow('nodes.knowledgeRetrieval.metadata.panel.conditions'),
diff --git a/web/features/agent-v2/agent-composer/store-modules/tools.ts b/web/features/agent-v2/agent-composer/store-modules/tools.ts
index c1d8c64082c..804f758b4a7 100644
--- a/web/features/agent-v2/agent-composer/store-modules/tools.ts
+++ b/web/features/agent-v2/agent-composer/store-modules/tools.ts
@@ -1,4 +1,4 @@
-import type { AgentSoulConfigFormState, AgentTool } from '../form-state'
+import type { AgentProviderTool, AgentSoulConfigFormState, AgentTool } from '../form-state'
 import type { DraftFieldUpdate } from './utils'
 import { atom, useSetAtom } from 'jotai'
 import { useCallback } from 'react'
@@ -91,15 +91,19 @@ export function useRemoveProviderToolAction() {
 export function useSetProviderToolCredential() {
   const setTools = useSetAtom(agentComposerToolsAtom)
 
-  return useCallback((toolId: string, credentialId?: string) => {
+  return useCallback((toolId: string, credentialId?: string, credentialType?: AgentProviderTool['credentialType']) => {
     setTools(tools => tools.map((tool) => {
       if (tool.kind !== 'provider' || tool.id !== toolId)
         return tool
 
+      const nextCredentialType = credentialType === 'oauth2' || tool.credentialType === 'oauth2'
+        ? 'oauth2'
+        : 'api-key'
+
       return {
         ...tool,
         credentialId,
-        credentialType: 'api-key',
+        credentialType: nextCredentialType,
         credentialVariant: 'authorized',
       }
     }))
diff --git a/web/features/agent-v2/agent-detail/access/components/__tests__/access-surface-card.spec.tsx b/web/features/agent-v2/agent-detail/access/components/__tests__/access-surface-card.spec.tsx
index d7b95a5aa84..c6512469795 100644
--- a/web/features/agent-v2/agent-detail/access/components/__tests__/access-surface-card.spec.tsx
+++ b/web/features/agent-v2/agent-detail/access/components/__tests__/access-surface-card.spec.tsx
@@ -25,6 +25,25 @@ describe('AccessSurfaceCard', () => {
   })
 
   describe('Copy feedback', () => {
+    it('should expose the surface card by its title', () => {
+      render(
+        <AccessSurfaceCard
+          title="Web app"
+          icon="i-ri-window-line"
+          iconClassName="bg-state-accent-solid"
+          endpointLabel="Access URL"
+          endpoint="https://chat.example.test/agent/token"
+          enabled
+          onEnabledChange={vi.fn()}
+          copyLabel="Copy access URL"
+        >
+          <button type="button">Action</button>
+        </AccessSurfaceCard>,
+      )
+
+      expect(screen.getByRole('article', { name: 'Web app' })).toBeInTheDocument()
+    })
+
     it('should copy the endpoint and render copied state from the clipboard hook', async () => {
       const user = userEvent.setup()
       const { rerender } = render(
diff --git a/web/features/agent-v2/agent-detail/access/components/access-surface-card.tsx b/web/features/agent-v2/agent-detail/access/components/access-surface-card.tsx
index d546c318313..4a7a3707f17 100644
--- a/web/features/agent-v2/agent-detail/access/components/access-surface-card.tsx
+++ b/web/features/agent-v2/agent-detail/access/components/access-surface-card.tsx
@@ -7,6 +7,7 @@ import { StatusDot } from '@langgenius/dify-ui/status-dot'
 import { Switch } from '@langgenius/dify-ui/switch'
 import { toast } from '@langgenius/dify-ui/toast'
 import { useClipboard } from 'foxact/use-clipboard'
+import { useId } from 'react'
 import { useTranslation } from 'react-i18next'
 
 export type AccessSurfaceCardProps = {
@@ -44,6 +45,7 @@ export function AccessSurfaceCard({
 }: AccessSurfaceCardProps) {
   const { t } = useTranslation('agentV2')
   const { t: tCommon } = useTranslation('common')
+  const titleId = useId()
   const { copied, copy } = useClipboard({
     timeout: 2000,
     onCopyError: () => {
@@ -61,14 +63,14 @@ export function AccessSurfaceCard({
   }
 
   return (
-    <article className="rounded-xl border border-components-panel-border bg-components-panel-bg shadow-xs">
+    <article aria-labelledby={titleId} className="rounded-xl border border-components-panel-border bg-components-panel-bg shadow-xs">
       <div className="px-4 pt-4 pb-4">
         <div className="flex min-w-0 flex-wrap items-center gap-x-3 gap-y-2">
           <div className="flex min-w-48 flex-1 items-center gap-2">
             <span className={cn('flex size-6 shrink-0 items-center justify-center rounded-lg', iconClassName)}>
               <span aria-hidden className={cn(icon, 'size-4')} />
             </span>
-            <h3 className="truncate system-md-semibold text-text-secondary">
+            <h3 id={titleId} className="truncate system-md-semibold text-text-secondary">
               {title}
             </h3>
             {badge}
diff --git a/web/features/agent-v2/agent-detail/access/page.tsx b/web/features/agent-v2/agent-detail/access/page.tsx
index 7e1d653fb70..edfef44976d 100644
--- a/web/features/agent-v2/agent-detail/access/page.tsx
+++ b/web/features/agent-v2/agent-detail/access/page.tsx
@@ -5,6 +5,7 @@ import { useQuery } from '@tanstack/react-query'
 import { useTranslation } from 'react-i18next'
 import { useDocLink } from '@/context/i18n'
 import { consoleQuery } from '@/service/client'
+import { AgentDetailSectionSurface } from '../section-surface'
 import { ServiceApiAccessCard } from './components/service-api-access-card'
 import { WebAppAccessCard } from './components/web-app-access-card'
 import { WorkflowReferencesTable } from './components/workflow-references-table'
@@ -27,10 +28,7 @@ export function AgentAccessPage({
   }))
 
   return (
-    <section
-      aria-label={t('agentDetail.sections.access')}
-      className="flex h-full min-w-0 flex-1 flex-col overflow-hidden bg-components-panel-bg-blur"
-    >
+    <AgentDetailSectionSurface label={t('agentDetail.sections.access')}>
       <header className="h-15.5 shrink-0 px-6 pt-3 pb-2">
         <div className="min-w-0">
           <h2 className="system-xl-semibold text-text-primary">
@@ -77,6 +75,6 @@ export function AgentAccessPage({
           </section>
         </div>
       </ScrollArea>
-    </section>
+    </AgentDetailSectionSurface>
   )
 }
diff --git a/web/features/agent-v2/agent-detail/configure/__tests__/model-compatibility.spec.ts b/web/features/agent-v2/agent-detail/configure/__tests__/model-compatibility.spec.ts
index 4a8bbd0f5b5..75f9e827220 100644
--- a/web/features/agent-v2/agent-detail/configure/__tests__/model-compatibility.spec.ts
+++ b/web/features/agent-v2/agent-detail/configure/__tests__/model-compatibility.spec.ts
@@ -4,7 +4,7 @@ import {
   ModelStatusEnum,
   ModelTypeEnum,
 } from '@/app/components/header/account-setting/model-provider-page/declarations'
-import { isAgentCompatibleModel } from '../model-compatibility'
+import { isAgentCompatibleModel, isAgentSuggestedModel } from '../model-compatibility'
 
 const createModel = (provider: string): Model => ({
   provider,
@@ -25,47 +25,143 @@ const createModelItem = (model: string, overrides: Partial<ModelItem> = {}): Mod
   ...overrides,
 })
 
+const createModelItemWithLabel = (model: string, label: string, overrides: Partial<ModelItem> = {}): ModelItem => createModelItem(model, {
+  label: { en_US: label, zh_Hans: label },
+  ...overrides,
+})
+
 describe('isAgentCompatibleModel', () => {
-  it('should reject configured OpenAI models below the Agent-compatible baseline', () => {
-    const provider = createModel('langgenius/openai/openai')
+  it('should reject configured GPT models below the Agent-compatible baseline', () => {
+    const provider = createModel('any-provider')
 
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-4o-mini'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('chatgpt-4o-latest'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-4.1-mini'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-4.1-nano'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5-pro'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5.1'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5.2'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5-mini'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5-nano'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-5-chat'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-4o'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gpt-4.1'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-4'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-4-turbo'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-4-vision-preview'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('gpt-3.5-turbo-16k'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('o1'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('o1-preview'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('o3-mini'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('o3'))).toBe(false)
     expect(isAgentCompatibleModel(provider, createModelItem('o4-mini'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('o4'))).toBe(false)
   })
 
-  it('should allow newer OpenAI models and other providers', () => {
-    expect(isAgentCompatibleModel(createModel('openai'), createModelItem('gpt-4o'))).toBe(true)
-    expect(isAgentCompatibleModel(createModel('openai'), createModelItem('gpt-4.1'))).toBe(true)
-    expect(isAgentCompatibleModel(createModel('anthropic'), createModelItem('other-model'))).toBe(true)
+  it('should allow models that are not in the blacklist', () => {
+    expect(isAgentCompatibleModel(createModel('any-provider'), createModelItem('gpt-5.5'))).toBe(true)
+    expect(isAgentCompatibleModel(createModel('any-provider'), createModelItem('other-model'))).toBe(true)
   })
 
   it('should reject specifically configured models that do not meet the Agent baseline', () => {
-    expect(isAgentCompatibleModel(createModel('anthropic'), createModelItem('claude-3-haiku-20240307'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('anthropic'), createModelItem('claude-3.5-sonnet-20241022'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('langgenius/gemini/google'), createModelItem('gemini-2.5-flash-lite'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('langgenius/gemini/google'), createModelItem('gemini-1.5-flash-8b'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('deepseek'), createModelItem('deepseek-r1-distill-qwen-32b'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('minimax'), createModelItem('minimax-text-01'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('minimax'), createModelItem('minimax-m1'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('tongyi'), createModelItem('qwen2.5-72b-instruct'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('tongyi'), createModelItem('qwen2.5-coder-32b-instruct'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('tongyi'), createModelItem('qwen3-30b'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('langgenius/zhipuai/zhipuai'), createModelItem('glm-4-airx'))).toBe(false)
-    expect(isAgentCompatibleModel(createModel('langgenius/zhipuai/zhipuai'), createModelItem('glm-z1-flash'))).toBe(false)
+    const provider = createModel('any-provider')
+
+    expect(isAgentCompatibleModel(provider, createModelItem('claude-3-haiku-20240307'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('claude-3.5-sonnet-20241022'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gemini-2.5-flash-lite'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('Gemini 2.5 Flash'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('Gemini 2.0 Flash'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gemini-2.5-pro-preview'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('Gemini 3.1 Flash-Lite Preview'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('gemini-1.5-flash-8b'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('Nano Banana Pro'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('grok-code-fast'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('grok-code-fast-1'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('grok-4-beta'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('grok-2'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('grok-3-mini'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('deepseek-chat'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('deepseek-coder'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('deepseek-reasoner'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('deepseek-chat-v3'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('deepseek-R1'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('DeepSeek-R1-Distill-Qwen-32B'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('deepseek-v3.1'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('abab6.5-chat'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('minimax-text-01'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('minimax-m1'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('minimax-m2'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('minimax-m2-chat'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('kimi-k2-thinking'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('moonshot-v1-128k'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qvq-max'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwq-plus'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen-max'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen2.5-72b-instruct'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen2.5-coder-32b-instruct'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen3-coder-plus'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen3.5-max'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen3.7-flash'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen-flash'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('farui-plus'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('chatglm-3-turbo'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('glm-3-turbo'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('glm-4-airx'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('glm-4.7'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('glm-z1-flash'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItem('glm-z2-plus'))).toBe(false)
   })
 
-  it('should allow unconfigured models from the same providers', () => {
-    expect(isAgentCompatibleModel(createModel('anthropic'), createModelItem('claude-sonnet-4'))).toBe(true)
-    expect(isAgentCompatibleModel(createModel('langgenius/gemini/google'), createModelItem('gemini-2.5-pro'))).toBe(true)
-    expect(isAgentCompatibleModel(createModel('deepseek'), createModelItem('deepseek-coder'))).toBe(true)
-    expect(isAgentCompatibleModel(createModel('tongyi'), createModelItem('qwen3-coder-plus'))).toBe(true)
-    expect(isAgentCompatibleModel(createModel('langgenius/zhipuai/zhipuai'), createModelItem('glm-4.7'))).toBe(true)
+  it('should ignore provider when evaluating blacklist patterns', () => {
+    expect(isAgentCompatibleModel(createModel('custom-provider'), createModelItem('gpt-4o'))).toBe(false)
+    expect(isAgentCompatibleModel(createModel('custom-provider'), createModelItem('claude-3-haiku-20240307'))).toBe(false)
+    expect(isAgentCompatibleModel(createModel('openai'), createModelItem('claude-sonnet-4'))).toBe(true)
+  })
+
+  it('should evaluate blacklist patterns against the English model label', () => {
+    const provider = createModel('any-provider')
+
+    expect(isAgentCompatibleModel(provider, createModelItemWithLabel('model-id', 'gpt-4o'))).toBe(false)
+    expect(isAgentCompatibleModel(provider, createModelItemWithLabel('gpt-4o', 'safe-model-label'))).toBe(true)
+  })
+
+  it('should allow unconfigured models from providers with blacklisted model families', () => {
+    const provider = createModel('any-provider')
+
+    expect(isAgentCompatibleModel(provider, createModelItem('claude-sonnet-4'))).toBe(true)
+    expect(isAgentCompatibleModel(provider, createModelItem('gemini-2.5-pro'))).toBe(true)
+    expect(isAgentCompatibleModel(provider, createModelItem('Gemini 2.5 Pro'))).toBe(true)
+    expect(isAgentCompatibleModel(provider, createModelItem('grok-4'))).toBe(true)
+    expect(isAgentCompatibleModel(provider, createModelItem('qwen3.7-max'))).toBe(true)
+  })
+})
+
+describe('isAgentSuggestedModel', () => {
+  it('should suggest configured Agent baseline models by English label', () => {
+    const provider = createModel('any-provider')
+
+    expect(isAgentSuggestedModel(provider, createModelItem('gpt-5.5'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('gpt-5.5-pro'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('Claude Opus 4.8'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('opus-4.7'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('Claude Sonnet 4.6'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('Gemini 3.1 Pro Preview'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('Gemini 3.1 Pro Preview 001'))).toBe(false)
+    expect(isAgentSuggestedModel(provider, createModelItem('grok-4.3'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('deepseek-v4-pro'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('kimi-k2.6'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('minimax-m3'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('qwen3.7-max'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('qwen3-coder-plus'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('glm-5.1'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItem('other-model'))).toBe(false)
+  })
+
+  it('should evaluate suggestions against the English model label', () => {
+    const provider = createModel('any-provider')
+
+    expect(isAgentSuggestedModel(provider, createModelItemWithLabel('model-id', 'GPT 5.5 Pro'))).toBe(true)
+    expect(isAgentSuggestedModel(provider, createModelItemWithLabel('gpt-5.5', 'safe-model-label'))).toBe(false)
   })
 })
diff --git a/web/features/agent-v2/agent-detail/configure/__tests__/page.spec.tsx b/web/features/agent-v2/agent-detail/configure/__tests__/page.spec.tsx
index fdd0e299929..cdce3ee690a 100644
--- a/web/features/agent-v2/agent-detail/configure/__tests__/page.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/__tests__/page.spec.tsx
@@ -230,6 +230,7 @@ vi.mock('../components/orchestrate', async () => {
 
 vi.mock('../components/orchestrate/build-draft-bar', () => ({
   AgentBuildDraftBar: (props: {
+    changeSummary?: unknown
     changesCount: number
     disabled?: boolean
     onApply: () => void
@@ -324,6 +325,7 @@ vi.mock('../components/preview/header', () => ({
     onOpenWorkingDirectory: () => void
     onRefresh: () => void
     refreshDisabled?: boolean
+    showWorkingDirectoryAction?: boolean
   }) => (
     <div>
       <div>{props.mode}</div>
@@ -336,9 +338,11 @@ vi.mock('../components/preview/header', () => ({
       <button type="button" onClick={props.onToggleChatFeatures}>
         chat features
       </button>
-      <button type="button" onClick={props.onOpenWorkingDirectory}>
-        open working directory
-      </button>
+      {props.showWorkingDirectoryAction && (
+        <button type="button" onClick={props.onOpenWorkingDirectory}>
+          open working directory
+        </button>
+      )}
       <button type="button" disabled={props.refreshDisabled} onClick={props.onRefresh}>
         restart preview
       </button>
@@ -805,6 +809,120 @@ describe('AgentConfigurePage', () => {
       expect(mocks.checkoutBuildDraft).not.toHaveBeenCalled()
     })
 
+    it('should show the working directory action after the first build reply completes', async () => {
+      const queryClient = new QueryClient()
+      mocks.queryState.composer = {
+        data: {
+          agent_soul: {
+            prompt: {
+              system_prompt: 'draft prompt',
+            },
+          },
+        },
+        isFetching: false,
+        isError: false,
+        isPending: false,
+        isSuccess: true,
+        refetch: vi.fn(),
+      }
+      mocks.queryState.buildDraft = {
+        data: {
+          agent_soul: {
+            prompt: {
+              system_prompt: 'build prompt',
+            },
+          },
+          draft: {},
+          variant: 'agent_app',
+        },
+        dataUpdatedAt: 1,
+        error: null,
+        isFetching: false,
+        isError: false,
+        isPending: false,
+        isSuccess: true,
+        refetch: vi.fn(),
+      }
+
+      render(
+        <QueryClientProvider client={queryClient}>
+          <AgentConfigurePage agentId="agent-1" />
+        </QueryClientProvider>,
+      )
+
+      expect(screen.getByRole('button', { name: 'apply build draft' })).toBeEnabled()
+      expect(screen.getByRole('button', { name: 'open working directory' })).toBeInTheDocument()
+
+      fireEvent.click(screen.getByRole('button', { name: 'send build message' }))
+
+      await waitFor(() => {
+        expect(screen.getByRole('region', { name: 'build-chat' })).toHaveTextContent('sent:yes')
+      })
+      expect(screen.getByRole('button', { name: 'apply build draft' })).toBeDisabled()
+      expect(screen.queryByRole('button', { name: 'open working directory' })).not.toBeInTheDocument()
+
+      fireEvent.click(screen.getByRole('button', { name: 'complete build conversation' }))
+
+      expect(screen.getByRole('button', { name: 'open working directory' })).toBeInTheDocument()
+
+      fireEvent.click(screen.getByRole('button', { name: 'send build message' }))
+
+      expect(screen.getByRole('button', { name: 'open working directory' })).toBeInTheDocument()
+    })
+
+    it('should hide the working directory action when the build chat has no conversation', () => {
+      const queryClient = new QueryClient()
+      mocks.queryState.agent = {
+        ...mocks.queryState.agent,
+        data: {
+          ...mocks.queryState.agent.data,
+          debug_conversation_has_messages: false,
+          debug_conversation_id: '',
+          debug_conversation_message_count: 0,
+        },
+      }
+      mocks.queryState.composer = {
+        data: {
+          agent_soul: {
+            prompt: {
+              system_prompt: 'draft prompt',
+            },
+          },
+        },
+        isFetching: false,
+        isError: false,
+        isPending: false,
+        isSuccess: true,
+        refetch: vi.fn(),
+      }
+      mocks.queryState.buildDraft = {
+        data: {
+          agent_soul: {
+            prompt: {
+              system_prompt: 'build prompt',
+            },
+          },
+          draft: {},
+          variant: 'agent_app',
+        },
+        dataUpdatedAt: 1,
+        error: null,
+        isFetching: false,
+        isError: false,
+        isPending: false,
+        isSuccess: true,
+        refetch: vi.fn(),
+      }
+
+      render(
+        <QueryClientProvider client={queryClient}>
+          <AgentConfigurePage agentId="agent-1" />
+        </QueryClientProvider>,
+      )
+
+      expect(screen.queryByRole('button', { name: 'open working directory' })).not.toBeInTheDocument()
+    })
+
     it('should show chat features from the active build draft source as read-only', async () => {
       const user = userEvent.setup()
       const queryClient = new QueryClient()
diff --git a/web/features/agent-v2/agent-detail/configure/__tests__/use-agent-configure-sync.spec.tsx b/web/features/agent-v2/agent-detail/configure/__tests__/use-agent-configure-sync.spec.tsx
index 0bc0be92734..ad402fb2e38 100644
--- a/web/features/agent-v2/agent-detail/configure/__tests__/use-agent-configure-sync.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/__tests__/use-agent-configure-sync.spec.tsx
@@ -2,6 +2,7 @@ import type { PropsWithChildren } from 'react'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
 import { act, renderHook } from '@testing-library/react'
 import { createStore, Provider as JotaiProvider } from 'jotai'
+import { MetadataFilteringModeEnum } from '@/app/components/workflow/nodes/knowledge-retrieval/types'
 import { defaultAgentSoulConfigFormState } from '@/features/agent-v2/agent-composer/form-state'
 import { agentComposerDraftAtom, agentComposerPublishedDraftAtom } from '@/features/agent-v2/agent-composer/store'
 import { agentComposerFilesAtom } from '@/features/agent-v2/agent-composer/store-modules/files'
@@ -437,7 +438,7 @@ describe('useAgentConfigureSync', () => {
     }))
   })
 
-  it('should skip autosave when knowledge retrieval validation fails', async () => {
+  it('should autosave when knowledge retrieval validation fails', async () => {
     const { result, store } = renderUseAgentConfigureSync()
 
     act(() => {
@@ -457,8 +458,8 @@ describe('useAgentConfigureSync', () => {
       await vi.advanceTimersByTimeAsync(5000)
     })
 
-    expect(composerPutMutationFn).not.toHaveBeenCalled()
-    expect(result.current.draftSavedAt).toBeUndefined()
+    expect(composerPutMutationFn).toHaveBeenCalledTimes(1)
+    expect(result.current.draftSavedAt).toBeDefined()
   })
 
   it('should keep autosave failures silent and leave the local draft dirty', async () => {
@@ -582,7 +583,7 @@ describe('useAgentConfigureSync', () => {
     }))
   })
 
-  it('should reject manual save when knowledge retrieval validation fails', async () => {
+  it('should save draft manually when knowledge retrieval validation fails', async () => {
     const { result, store } = renderUseAgentConfigureSync()
 
     act(() => {
@@ -598,8 +599,11 @@ describe('useAgentConfigureSync', () => {
       })
     })
 
-    await expect(result.current.saveDraft()).rejects.toThrow('Agent knowledge retrieval configuration is invalid.')
-    expect(composerPutMutationFn).not.toHaveBeenCalled()
+    await act(async () => {
+      await result.current.saveDraft()
+    })
+
+    expect(composerPutMutationFn).toHaveBeenCalledTimes(1)
   })
 
   it('should publish only when publishDraft is called explicitly', async () => {
@@ -755,7 +759,7 @@ describe('useAgentConfigureSync', () => {
     expect(toastMock.error).toHaveBeenCalledWith('common.api.actionFailed')
   })
 
-  it('should reject publish when knowledge retrieval validation fails', async () => {
+  it('should toast and skip publish when knowledge retrieval validation fails', async () => {
     const { result, store } = renderUseAgentConfigureSync()
 
     act(() => {
@@ -771,9 +775,39 @@ describe('useAgentConfigureSync', () => {
       })
     })
 
-    await expect(result.current.publishDraft()).rejects.toThrow('Agent knowledge retrieval configuration is invalid.')
+    await act(async () => {
+      await result.current.publishDraft()
+    })
+
     expect(composerPutMutationFn).not.toHaveBeenCalled()
     expect(publishAgentMutationFn).not.toHaveBeenCalled()
+    expect(toastMock.error).toHaveBeenCalledWith('common.errorMsg.fieldRequired:{"field":"agentV2.agentDetail.configure.knowledgeRetrieval.dialog.knowledge.label"}')
+  })
+
+  it('should toast metadata filtering model error when publishing with automatic metadata filtering and no model', async () => {
+    const { result, store } = renderUseAgentConfigureSync()
+
+    act(() => {
+      store.set(agentComposerDraftAtom, {
+        ...defaultAgentSoulConfigFormState,
+        knowledgeRetrievals: [
+          {
+            id: 'retrieval-1',
+            name: 'Docs Search',
+            datasetRefs: [{ id: 'dataset-1', name: 'Docs' }],
+            metadataFilterMode: MetadataFilteringModeEnum.automatic,
+          },
+        ],
+      })
+    })
+
+    await act(async () => {
+      await result.current.publishDraft()
+    })
+
+    expect(composerPutMutationFn).not.toHaveBeenCalled()
+    expect(publishAgentMutationFn).not.toHaveBeenCalled()
+    expect(toastMock.error).toHaveBeenCalledWith('agentV2.agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired')
   })
 
   it('should expose publishing status from the publish mutation while publish is pending', async () => {
diff --git a/web/features/agent-v2/agent-detail/configure/components/composer-session.tsx b/web/features/agent-v2/agent-detail/configure/components/composer-session.tsx
index c9a53698b30..ac5e8b886ea 100644
--- a/web/features/agent-v2/agent-detail/configure/components/composer-session.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/composer-session.tsx
@@ -216,6 +216,7 @@ function AgentConfigurePageComposerContent({
   } = configureData
   const [buildDraftActionsDisabled, setBuildDraftActionsDisabled] = useState(false)
   const [clearPreviewChat, setClearPreviewChat] = useState(false)
+  const [completedBuildConversationId, setCompletedBuildConversationId] = useState<string | null>(null)
   const conversationIds = useAtomValue(agentConfigureConversationIdsAtom)
   const rightPanelChatMode = useAtomValue(agentConfigureRightPanelChatModeAtom)
   const workingDirectoryPanel = useAgentWorkingDirectoryPanel({
@@ -237,6 +238,7 @@ function AgentConfigurePageComposerContent({
       await onRefreshDebugConversationAsync()
     }
     finally {
+      setCompletedBuildConversationId(null)
       setConversationId({ mode: 'build', conversationId: null })
       setClearPreviewChat(true)
     }
@@ -288,6 +290,14 @@ function AgentConfigurePageComposerContent({
     || buildDraftActions.isApplyingBuildDraft
     || buildDraftActions.isDiscardingBuildDraft
   const isChatFeaturesReadOnly = (isViewingVersion && versionQuery.isPending) || buildDraft.isActive
+  const buildConversationHasAgentResponse = !!conversationIds.build && (
+    conversationIds.build === completedBuildConversationId
+    || (
+      conversationIds.build === agentQuery.data?.debug_conversation_id
+      && (agentQuery.data?.debug_conversation_has_messages ?? false)
+    )
+  )
+  const showWorkingDirectoryAction = rightPanelChatMode === 'build' && buildConversationHasAgentResponse
   const restartCurrentChat = () => {
     if (isRestartCurrentChatDisabled)
       return
@@ -326,6 +336,7 @@ function AgentConfigurePageComposerContent({
           bottomAction={showBuildDraftBar
             ? (
                 <AgentBuildDraftBar
+                  changeSummary={buildDraft.changeSummary}
                   changesCount={buildDraft.changesCount}
                   disabled={buildDraftActionsDisabled}
                   isApplying={buildDraftActions.isApplyingBuildDraft}
@@ -364,6 +375,7 @@ function AgentConfigurePageComposerContent({
               }}
               onRefresh={restartCurrentChat}
               refreshDisabled={isRestartCurrentChatDisabled}
+              showWorkingDirectoryAction={showWorkingDirectoryAction}
             />
           )}
           chat={(
@@ -381,6 +393,7 @@ function AgentConfigurePageComposerContent({
               onClearChatListChange={setClearPreviewChat}
               onConversationComplete={(mode, completedConversationId) => {
                 if (mode === 'build') {
+                  setCompletedBuildConversationId(completedConversationId)
                   invalidateAgentWorkingDirectoryFiles({
                     agentId,
                     conversationId: completedConversationId,
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/bottom-actions.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/bottom-actions.spec.tsx
new file mode 100644
index 00000000000..084599d650c
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/bottom-actions.spec.tsx
@@ -0,0 +1,22 @@
+import { render, screen } from '@testing-library/react'
+import { AgentOrchestrateBottomActions } from '../bottom-actions'
+
+describe('AgentOrchestrateBottomActions', () => {
+  it('should allow callers to keep the bottom action width when nested content opens', () => {
+    const { rerender } = render(
+      <AgentOrchestrateBottomActions>
+        <div data-testid="bottom-action" />
+      </AgentOrchestrateBottomActions>,
+    )
+
+    expect(screen.getByTestId('bottom-action').parentElement).toHaveClass('has-[[data-open]]:max-w-96')
+
+    rerender(
+      <AgentOrchestrateBottomActions shrinkOnOpen={false}>
+        <div data-testid="bottom-action" />
+      </AgentOrchestrateBottomActions>,
+    )
+
+    expect(screen.getByTestId('bottom-action').parentElement).not.toHaveClass('has-[[data-open]]:max-w-96')
+  })
+})
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/build-draft-bar.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/build-draft-bar.spec.tsx
index 91d34550e2f..2b16ac7456d 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/build-draft-bar.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/build-draft-bar.spec.tsx
@@ -1,7 +1,30 @@
+import type { AgentBuildDraftChangeSummary } from '../build-draft-changes-context'
 import { render, screen } from '@testing-library/react'
 import userEvent from '@testing-library/user-event'
 import { AgentBuildDraftBar } from '../build-draft-bar'
 
+const changeSummary: AgentBuildDraftChangeSummary = {
+  changedKeys: ['skills', 'files', 'envVariables'],
+  changesCount: 5,
+  skills: [
+    { id: 'skill-1', name: 'tender-analyzer', operation: 'added' },
+    { id: 'skill-2', name: 'figma-code-connect', operation: 'removed' },
+  ],
+  files: [
+    {
+      id: '__agent_config_build_note__',
+      name: 'build_note.md',
+      operation: 'updated',
+      icon: 'markdown',
+      descriptionKey: 'agentDetail.configure.buildDraft.buildNoteDescription',
+    },
+    { id: 'file-1', name: 'index.json', operation: 'added', icon: 'json' },
+  ],
+  envVariables: [
+    { id: 'env-1', name: 'API_KEY', operation: 'updated' },
+  ],
+}
+
 describe('AgentBuildDraftBar', () => {
   it('should disable both build draft actions when the bar is disabled', async () => {
     const user = userEvent.setup()
@@ -72,20 +95,25 @@ describe('AgentBuildDraftBar', () => {
     expect(screen.getByRole('button', { name: 'agentV2.agentDetail.configure.buildDraft.discard' })).toBeDisabled()
   })
 
-  it('should not show build draft change metadata', () => {
-    const { rerender } = render(
+  it('should show build draft change metadata', () => {
+    const { container, rerender } = render(
       <AgentBuildDraftBar
+        changeSummary={changeSummary}
         changesCount={0}
         onApply={vi.fn()}
         onDiscard={vi.fn()}
       />,
     )
 
+    expect(container.firstElementChild).toHaveClass('w-full')
+    expect(container.firstElementChild).not.toHaveClass('w-fit')
     expect(screen.getByText('agentV2.agentDetail.configure.buildDraft.title')).toBeInTheDocument()
-    expect(screen.getAllByText(/^agentV2\.agentDetail\.configure\.buildDraft\./)).toHaveLength(2)
+    expect(screen.getByRole('button', { name: 'agentV2.agentDetail.configure.buildDraft.changesToApply:{"count":0}' })).toBeInTheDocument()
+    expect(document.querySelector('.i-ri-arrow-right-s-line')).toBeInTheDocument()
 
     rerender(
       <AgentBuildDraftBar
+        changeSummary={changeSummary}
         changesCount={2}
         onApply={vi.fn()}
         onDiscard={vi.fn()}
@@ -93,7 +121,60 @@ describe('AgentBuildDraftBar', () => {
     )
 
     expect(screen.getByText('agentV2.agentDetail.configure.buildDraft.title')).toBeInTheDocument()
-    expect(screen.getAllByText(/^agentV2\.agentDetail\.configure\.buildDraft\./)).toHaveLength(2)
+    expect(screen.getByRole('button', { name: 'agentV2.agentDetail.configure.buildDraft.changesToApply:{"count":2}' })).toBeInTheDocument()
+  })
+
+  it('should open build draft change details from the metadata trigger', async () => {
+    const user = userEvent.setup()
+    const onApply = vi.fn()
+    const onDiscard = vi.fn()
+    const getBoundingClientRectSpy = vi
+      .spyOn(HTMLElement.prototype, 'getBoundingClientRect')
+      .mockReturnValue({
+        bottom: 50,
+        height: 50,
+        left: 0,
+        right: 432,
+        top: 0,
+        width: 432,
+        x: 0,
+        y: 0,
+        toJSON: () => ({}),
+      })
+
+    render(
+      <AgentBuildDraftBar
+        changeSummary={changeSummary}
+        changesCount={5}
+        onApply={onApply}
+        onDiscard={onDiscard}
+      />,
+    )
+
+    const changesTrigger = screen.getByRole('button', { name: 'agentV2.agentDetail.configure.buildDraft.changesToApply:{"count":5}' })
+
+    await user.click(changesTrigger)
+
+    const changesPanelId = changesTrigger.getAttribute('aria-controls')
+    expect(changesPanelId).toBeTruthy()
+    expect(document.getElementById(changesPanelId!)).toHaveStyle({ width: '432px' })
+    expect(screen.getByText('agentV2.agentDetail.configure.skills.label')).toBeInTheDocument()
+    expect(screen.getByText('tender-analyzer')).toBeInTheDocument()
+    expect(screen.getByText('figma-code-connect')).toBeInTheDocument()
+    expect(screen.getByText('build_note.md')).toBeInTheDocument()
+    expect(screen.getByText('agentV2.agentDetail.configure.buildDraft.buildNoteDescription')).toBeInTheDocument()
+    expect(screen.getByText('index.json')).toBeInTheDocument()
+    expect(screen.getByText('agentV2.agentDetail.configure.advancedSettings.envEditor.shortLabel')).toBeInTheDocument()
+    expect(screen.getByText('API_KEY')).toBeInTheDocument()
+    expect(document.querySelector('.text-text-accent')).toHaveClass('i-ri-add-circle-fill')
+
+    await user.click(screen.getByRole('button', { name: 'agentV2.agentDetail.configure.buildDraft.discard' }))
+    await user.click(screen.getByRole('button', { name: 'custom.apply' }))
+
+    expect(onDiscard).toHaveBeenCalledTimes(1)
+    expect(onApply).toHaveBeenCalledTimes(1)
+
+    getBoundingClientRectSpy.mockRestore()
   })
 
   it('should keep both actions enabled when there are no build draft changes', async () => {
@@ -111,10 +192,7 @@ describe('AgentBuildDraftBar', () => {
 
     const discardButton = screen.getByRole('button', { name: 'agentV2.agentDetail.configure.buildDraft.discard' })
     const applyButton = screen.getByRole('button', { name: 'custom.apply' })
-    const buttons = screen.getAllByRole('button')
 
-    expect(buttons[0]).toBe(discardButton)
-    expect(buttons[1]).toBe(applyButton)
     expect(discardButton).toBeEnabled()
     expect(applyButton).toBeEnabled()
 
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/publish-bar.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/publish-bar.spec.tsx
index 7ba52ad0841..a2400bc8e72 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/publish-bar.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/__tests__/publish-bar.spec.tsx
@@ -250,7 +250,7 @@ describe('AgentConfigurePublishBar', () => {
     )
   })
 
-  it('should block publish when knowledge retrieval validation fails', () => {
+  it('should allow publish request when knowledge retrieval validation fails', async () => {
     const { onPublish } = renderPublishBar({
       setupStore: (store) => {
         store.set(agentComposerDraftAtom, {
@@ -266,15 +266,17 @@ describe('AgentConfigurePublishBar', () => {
       },
     })
 
-    expect(screen.getByRole('button', { name: /agentV2\.agentDetail\.configure\.publishBar\.publishUpdate/ })).toBeDisabled()
-    expect(screen.getByText('common.errorMsg.fieldRequired:{"field":"agentV2.agentDetail.configure.knowledgeRetrieval.dialog.knowledge.label"}')).toBeInTheDocument()
+    expect(screen.getByRole('button', { name: /agentV2\.agentDetail\.configure\.publishBar\.publishUpdate/ })).toBeEnabled()
+    expect(screen.queryByText('common.errorMsg.fieldRequired:{"field":"agentV2.agentDetail.configure.knowledgeRetrieval.dialog.knowledge.label"}')).not.toBeInTheDocument()
     expect(hotkeyRegistrations.get('Mod+Shift+P')?.options).toEqual(
-      expect.objectContaining({ enabled: false, ignoreInputs: false }),
+      expect.objectContaining({ enabled: true, ignoreInputs: false }),
     )
 
     fireEvent.click(screen.getByRole('button', { name: /agentV2\.agentDetail\.configure\.publishBar\.publishUpdate/ }))
 
-    expect(onPublish).not.toHaveBeenCalled()
+    await waitFor(() => {
+      expect(onPublish).toHaveBeenCalledTimes(1)
+    })
   })
 
   it('should restore the selected version from view-only mode', async () => {
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/bottom-actions.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/bottom-actions.tsx
index cd61ce1b2f1..bf798ba9b8f 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/bottom-actions.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/bottom-actions.tsx
@@ -1,9 +1,12 @@
 import type { ReactNode } from 'react'
+import { cn } from '@langgenius/dify-ui/cn'
 
 export function AgentOrchestrateBottomActions({
   children,
+  shrinkOnOpen = true,
 }: {
   children: ReactNode
+  shrinkOnOpen?: boolean
 }) {
   return (
     <div
@@ -14,7 +17,10 @@ export function AgentOrchestrateBottomActions({
         className="pointer-events-none absolute inset-0 z-0 bg-gradient-to-t from-components-panel-bg to-components-panel-bg-transparent [mask-image:linear-gradient(to_top,black,transparent)] backdrop-blur-[2px] [-webkit-mask-image:linear-gradient(to_top,black,transparent)]"
       />
       <div
-        className="relative z-10 flex w-full max-w-[506px] flex-col items-center justify-end transition-[max-width] duration-150 ease-out has-[[data-open]]:max-w-96 motion-reduce:transition-none"
+        className={cn(
+          'relative z-10 flex w-full max-w-[506px] flex-col items-center justify-end transition-[max-width] duration-150 ease-out motion-reduce:transition-none',
+          shrinkOnOpen && 'has-[[data-open]]:max-w-96',
+        )}
       >
         {children}
       </div>
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-bar.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-bar.tsx
index 924f3e44f15..f053d47287d 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-bar.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-bar.tsx
@@ -1,10 +1,15 @@
 'use client'
 
+import type { AgentBuildDraftChangeSummary } from './build-draft-changes-context'
 import { Button } from '@langgenius/dify-ui/button'
+import { CollapsiblePanel, CollapsibleRoot } from '@langgenius/dify-ui/collapsible'
+import { useId, useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { AgentBuildGridTexture } from '../build-grid-texture'
+import { AgentBuildDraftChangesPanel } from './build-draft-changes-panel'
 
 type AgentBuildDraftBarProps = {
+  changeSummary?: AgentBuildDraftChangeSummary
   changesCount: number
   disabled?: boolean
   isApplying?: boolean
@@ -14,6 +19,8 @@ type AgentBuildDraftBarProps = {
 }
 
 export function AgentBuildDraftBar({
+  changeSummary,
+  changesCount,
   disabled = false,
   isApplying = false,
   isDiscarding = false,
@@ -22,42 +29,85 @@ export function AgentBuildDraftBar({
 }: AgentBuildDraftBarProps) {
   const { t } = useTranslation('agentV2')
   const { t: tCustom } = useTranslation('custom')
+  const [open, setOpen] = useState(false)
+  const [panelWidth, setPanelWidth] = useState<number>()
+  const collapsedBarRef = useRef<HTMLDivElement>(null)
+  const changesPanelId = useId()
   const isActionPending = isApplying || isDiscarding
   const applyDisabled = disabled || isActionPending
   const discardDisabled = disabled || isActionPending
+  const changesLabel = t('agentDetail.configure.buildDraft.changesToApply', { count: changesCount })
+  const handleOpenChange = (nextOpen: boolean) => {
+    if (nextOpen) {
+      const width = collapsedBarRef.current?.getBoundingClientRect().width
+
+      if (width)
+        setPanelWidth(width)
+    }
+    else {
+      setPanelWidth(undefined)
+    }
+
+    setOpen(nextOpen)
+  }
 
   return (
-    <div className="pointer-events-auto relative flex h-[50px] w-fit max-w-full min-w-0 items-center gap-2 overflow-hidden rounded-xl border-[1.5px] border-[#A0BDFF] bg-components-panel-bg-blur p-2 shadow-lg shadow-shadow-shadow-5 backdrop-blur-[10px]">
+    <CollapsibleRoot
+      open={open}
+      onOpenChange={handleOpenChange}
+      className="group/build-draft pointer-events-auto relative w-full max-w-full min-w-0 overflow-hidden rounded-xl border-[1.5px] border-[#A0BDFF] bg-components-panel-bg-blur shadow-lg shadow-shadow-shadow-5 backdrop-blur-[10px]"
+    >
       <AgentBuildGridTexture
         aria-hidden
         cellOpacityMultiplier={3}
         className="pointer-events-none absolute top-[-104px] left-[-1171px] z-0 opacity-70"
         dotClassName="bg-[#5C90FF]"
       />
-      <div className="relative z-1 flex min-w-0 flex-1 flex-col justify-center gap-0.5 pr-8 pl-2">
-        <p className="min-w-0 truncate system-sm-semibold text-text-primary">
-          {t('agentDetail.configure.buildDraft.title')}
-        </p>
+      <CollapsiblePanel id={changesPanelId} className="relative z-1" style={panelWidth ? { width: panelWidth } : undefined}>
+        <AgentBuildDraftChangesPanel
+          changeSummary={changeSummary}
+          changesLabel={changesLabel}
+          onToggle={() => handleOpenChange(!open)}
+        />
+      </CollapsiblePanel>
+      <div ref={collapsedBarRef} className="relative z-1 flex h-[50px] max-w-full min-w-0 items-center gap-2 p-2 group-data-open/build-draft:justify-end">
+        <div className="flex min-w-0 flex-1 items-center gap-3 pr-8 pl-2 group-data-open/build-draft:hidden">
+          <p className="min-w-0 truncate system-sm-semibold text-text-primary">
+            {t('agentDetail.configure.buildDraft.title')}
+          </p>
+          <button
+            type="button"
+            aria-controls={changesPanelId}
+            aria-expanded={open}
+            className="flex min-w-0 cursor-pointer items-center gap-0.5 rounded-sm text-text-tertiary hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden"
+            onClick={() => handleOpenChange(true)}
+          >
+            <span className="min-w-0 truncate system-xs-regular">
+              {changesLabel}
+            </span>
+            <span aria-hidden className="i-ri-arrow-right-s-line size-4 shrink-0" />
+          </button>
+        </div>
+        <Button
+          type="button"
+          variant="secondary"
+          disabled={discardDisabled}
+          className="relative z-1 h-8 shrink-0 rounded-lg px-3"
+          onClick={onDiscard}
+        >
+          {t('agentDetail.configure.buildDraft.discard')}
+        </Button>
+        <Button
+          type="button"
+          variant="primary"
+          loading={isApplying}
+          disabled={applyDisabled}
+          className="relative z-1 h-8 min-w-20 shrink-0 rounded-lg px-3"
+          onClick={onApply}
+        >
+          {tCustom('apply')}
+        </Button>
       </div>
-      <Button
-        type="button"
-        variant="secondary"
-        disabled={discardDisabled}
-        className="relative z-1 h-8 shrink-0 rounded-lg px-3"
-        onClick={onDiscard}
-      >
-        {t('agentDetail.configure.buildDraft.discard')}
-      </Button>
-      <Button
-        type="button"
-        variant="primary"
-        loading={isApplying}
-        disabled={applyDisabled}
-        className="relative z-1 h-8 min-w-20 shrink-0 rounded-lg px-3"
-        onClick={onApply}
-      >
-        {tCustom('apply')}
-      </Button>
-    </div>
+    </CollapsibleRoot>
   )
 }
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-context.ts b/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-context.ts
index 111952903cd..7454049d52d 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-context.ts
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-context.ts
@@ -1,11 +1,31 @@
 'use client'
 
+import type { FileTreeIconType } from '@langgenius/dify-ui/file-tree'
 import type { ReactNode } from 'react'
 import type { AgentSoulConfigFormState } from '@/features/agent-v2/agent-composer/form-state'
+import type { I18nKeysWithPrefix } from '@/types/i18n'
 import { createContext, createElement, useContext, useMemo } from 'react'
 
 export type AgentBuildDraftChangedKey = keyof AgentSoulConfigFormState
 
+type AgentBuildDraftChangeOperation = 'added' | 'removed' | 'updated'
+
+export type AgentBuildDraftChangeItem = {
+  id: string
+  name: string
+  operation: AgentBuildDraftChangeOperation
+  icon?: FileTreeIconType
+  descriptionKey?: I18nKeysWithPrefix<'agentV2', 'agentDetail.configure.buildDraft.'>
+}
+
+export type AgentBuildDraftChangeSummary = {
+  changedKeys: readonly AgentBuildDraftChangedKey[]
+  changesCount: number
+  skills: readonly AgentBuildDraftChangeItem[]
+  files: readonly AgentBuildDraftChangeItem[]
+  envVariables: readonly AgentBuildDraftChangeItem[]
+}
+
 export type AgentBuildDraftChangeSection
   = | 'skills'
     | 'files'
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-panel.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-panel.tsx
new file mode 100644
index 00000000000..01cc465bad0
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/build-draft-changes-panel.tsx
@@ -0,0 +1,147 @@
+'use client'
+
+import type { TFunction } from 'i18next'
+import type { AgentBuildDraftChangeItem, AgentBuildDraftChangeSummary } from './build-draft-changes-context'
+import { cn } from '@langgenius/dify-ui/cn'
+import { FileTreeIcon } from '@langgenius/dify-ui/file-tree'
+import { useTranslation } from 'react-i18next'
+
+type AgentBuildDraftChangeSection = {
+  key: string
+  label: string
+  items?: readonly AgentBuildDraftChangeItem[]
+}
+
+export function AgentBuildDraftChangesPanel({
+  changeSummary,
+  changesLabel,
+  onToggle,
+}: {
+  changeSummary?: AgentBuildDraftChangeSummary
+  changesLabel: string
+  onToggle: () => void
+}) {
+  const { t } = useTranslation('agentV2')
+  const sections = getChangeSections({ changeSummary, t })
+
+  return (
+    <section className="flex w-full max-w-full flex-col px-2 pt-2">
+      <div className="flex h-6 min-w-0 items-center gap-3 pr-8 pl-2">
+        <p className="min-w-0 truncate system-sm-semibold text-text-primary">
+          {t('agentDetail.configure.buildDraft.title')}
+        </p>
+        <button
+          type="button"
+          className="flex min-w-0 cursor-pointer items-center gap-0.5 rounded-sm text-text-tertiary hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden"
+          onClick={onToggle}
+        >
+          <span className="min-w-0 truncate system-xs-regular">
+            {changesLabel}
+          </span>
+          <span aria-hidden className="i-ri-arrow-right-s-line size-4 shrink-0 rotate-90" />
+        </button>
+      </div>
+      <div className="flex max-h-[232px] flex-col gap-1 overflow-y-auto pt-2 pb-1">
+        {sections.map(section => (
+          <AgentBuildDraftChangeSectionRow key={section.key} section={section} />
+        ))}
+      </div>
+    </section>
+  )
+}
+
+function AgentBuildDraftChangeSectionRow({
+  section,
+}: {
+  section: AgentBuildDraftChangeSection
+}) {
+  return (
+    <div className="flex w-full items-start p-2">
+      <div className="flex min-w-20 shrink-0 items-center gap-1.5">
+        <span aria-hidden className="size-[5px] rounded-full bg-text-warning" />
+        <p className="min-w-0 flex-1 truncate system-xs-medium-uppercase text-text-tertiary">
+          {section.label}
+        </p>
+      </div>
+      {section.items?.length
+        ? (
+            <div className="flex min-w-0 flex-1 flex-col gap-2">
+              {section.items.map(item => (
+                <AgentBuildDraftChangeItemRow key={`${item.operation}-${item.id}`} item={item} />
+              ))}
+            </div>
+          )
+        : (
+            null
+          )}
+    </div>
+  )
+}
+
+function AgentBuildDraftChangeItemRow({
+  item,
+}: {
+  item: AgentBuildDraftChangeItem
+}) {
+  const { t } = useTranslation('agentV2')
+
+  return (
+    <div className="flex min-w-0 flex-col gap-0.5">
+      <div className="flex min-w-0 items-center gap-1">
+        <span
+          aria-hidden
+          className={cn(
+            'size-3 shrink-0',
+            item.operation === 'removed'
+              ? 'i-ri-indeterminate-circle-fill text-text-destructive'
+              : item.operation === 'added'
+                ? 'i-ri-add-circle-fill text-text-success'
+                : 'i-ri-add-circle-fill text-text-accent',
+          )}
+        />
+        <div className="flex min-w-0 flex-1 items-center gap-1">
+          {item.icon
+            ? <FileTreeIcon type={item.icon} className="size-4" />
+            : <span aria-hidden className="i-custom-public-agent-building-blocks size-4 shrink-0 text-text-tertiary" />}
+          <p className="min-w-0 truncate system-sm-regular text-text-secondary">
+            {item.name}
+          </p>
+        </div>
+      </div>
+      {item.descriptionKey && (
+        <p className="ms-4 system-xs-regular text-text-tertiary">
+          {t(item.descriptionKey)}
+        </p>
+      )}
+    </div>
+  )
+}
+
+function getChangeSections({
+  changeSummary,
+  t,
+}: {
+  changeSummary?: AgentBuildDraftChangeSummary
+  t: TFunction<'agentV2'>
+}): AgentBuildDraftChangeSection[] {
+  if (!changeSummary)
+    return []
+
+  const sections: AgentBuildDraftChangeSection[] = []
+  const pushItemSection = (key: string, label: string, items: readonly AgentBuildDraftChangeItem[]) => {
+    if (items.length === 0)
+      return
+
+    sections.push({
+      key,
+      label,
+      items,
+    })
+  }
+
+  pushItemSection('skills', t('agentDetail.configure.skills.label'), changeSummary.skills)
+  pushItemSection('files', t('agentDetail.configure.files.label'), changeSummary.files)
+  pushItemSection('envVariables', t('agentDetail.configure.advancedSettings.envEditor.shortLabel'), changeSummary.envVariables)
+
+  return sections
+}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/docs-link.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/docs-link.tsx
new file mode 100644
index 00000000000..bc530dc4ba5
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/docs-link.tsx
@@ -0,0 +1,20 @@
+import type { ReactNode } from 'react'
+
+export function DocsLink({
+  children,
+  href,
+}: {
+  children?: ReactNode
+  href: string
+}) {
+  return (
+    <a
+      href={href}
+      target="_blank"
+      rel="noopener noreferrer"
+      className="text-text-accent hover:underline"
+    >
+      {children}
+    </a>
+  )
+}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/tip-content.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/tip-content.tsx
index 7edbcbe6a27..2b393b1ed35 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/tip-content.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/common/tip-content.tsx
@@ -1,30 +1,11 @@
 'use client'
 
-import type { ReactNode } from 'react'
 import { Trans, useTranslation } from 'react-i18next'
 import { useDocLink } from '@/context/i18n'
+import { DocsLink } from './docs-link'
 
 type AgentConfigureTipContentProps = {
-  type: 'prompt' | 'skills' | 'files' | 'tools' | 'env'
-}
-
-function DocsLink({
-  children,
-  href,
-}: {
-  children?: ReactNode
-  href: string
-}) {
-  return (
-    <a
-      href={href}
-      target="_blank"
-      rel="noopener noreferrer"
-      className="text-text-accent hover:underline"
-    >
-      {children}
-    </a>
-  )
+  type: 'prompt' | 'skills' | 'files' | 'tools' | 'knowledge' | 'env'
 }
 
 export function AgentConfigureTipContent({ type }: AgentConfigureTipContentProps) {
@@ -47,13 +28,15 @@ export function AgentConfigureTipContent({ type }: AgentConfigureTipContentProps
 
   if (type === 'skills') {
     return (
-      <Trans
-        i18nKey="agentDetail.configure.skills.richTip"
-        ns="agentV2"
-        components={{
-          docLink: <DocsLink href={docLink()} />,
-        }}
-      />
+      <span className="whitespace-pre-line">
+        <Trans
+          i18nKey="agentDetail.configure.skills.richTip"
+          ns="agentV2"
+          components={{
+            docLink: <DocsLink href={docLink('/use-dify/build/new-agent/build#skills')} />,
+          }}
+        />
+      </span>
     )
   }
 
@@ -64,13 +47,27 @@ export function AgentConfigureTipContent({ type }: AgentConfigureTipContentProps
           i18nKey="agentDetail.configure.tools.richTip"
           ns="agentV2"
           components={{
-            pluginDocLink: <DocsLink href={docLink('/develop-plugin/getting-started/getting-started-dify-plugin')} />,
-            buildDocLink: <DocsLink href={docLink('/use-dify/build/agent')} />,
+            docLink: <DocsLink href={docLink('/use-dify/build/new-agent/build#tools')} />,
           }}
         />
       </span>
     )
   }
 
+  if (type === 'knowledge') {
+    return (
+      <Trans
+        i18nKey="agentDetail.configure.knowledgeRetrieval.richTip"
+        ns="agentV2"
+        components={{
+          docLink: <DocsLink href={docLink('/use-dify/build/new-agent/build#knowledge-retrieval')} />,
+        }}
+      />
+    )
+  }
+
+  if (type === 'files')
+    return <span className="whitespace-pre-line">{t('agentDetail.configure.files.tip')}</span>
+
   return <>{t(`agentDetail.configure.${type}.tip`)}</>
 }
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/__tests__/index.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/__tests__/index.spec.tsx
index d6921776e3e..028bd17f304 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/__tests__/index.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/__tests__/index.spec.tsx
@@ -125,8 +125,8 @@ function ConfigSnapshotProbe() {
   )
 }
 
-function renderAgentFiles({
-  initialDraft = {
+function createInitialDraft(overrides: Partial<AgentSoulConfigFormState> = {}): AgentSoulConfigFormState {
+  return {
     ...defaultAgentSoulConfigFormState,
     files: [
       {
@@ -144,7 +144,12 @@ function renderAgentFiles({
         configName: 'brief.md',
       },
     ],
-  } satisfies AgentSoulConfigFormState,
+    ...overrides,
+  }
+}
+
+function renderAgentFiles({
+  initialDraft = createInitialDraft(),
   initialOriginalConfig,
   apiContext = { agentId: 'agent-1', draftType: 'draft' } satisfies AgentConfigApiContext,
   readOnly = false,
@@ -366,9 +371,7 @@ describe('AgentFiles', () => {
   it('should show config note as a virtual build note file and preview its content locally', async () => {
     const user = userEvent.setup()
     renderAgentFiles({
-      initialOriginalConfig: {
-        config_note: 'Build context from the latest build chat.',
-      },
+      initialDraft: createInitialDraft({ configNote: 'Build context from the latest build chat.' }),
     })
 
     expect(screen.getByText('build_note.md')).toBeInTheDocument()
@@ -396,6 +399,36 @@ describe('AgentFiles', () => {
     }))
   })
 
+  it('should show generated build note metadata with an explanatory infotip', async () => {
+    const user = userEvent.setup()
+    renderAgentFiles({
+      initialDraft: createInitialDraft({ configNote: 'Build context from the latest build chat.' }),
+    })
+
+    expect(screen.getByText('agentV2.agentDetail.configure.files.buildNote.generated')).toBeInTheDocument()
+
+    await user.click(screen.getByRole('button', { name: 'agentV2.agentDetail.configure.files.buildNote.tooltip' }))
+
+    expect(await screen.findByText('agentDetail.configure.files.buildNote.richTooltip')).toBeInTheDocument()
+  })
+
+  it('should clear config note when deleting the virtual build note file', async () => {
+    const user = userEvent.setup()
+    renderAgentFiles({
+      initialDraft: createInitialDraft({ configNote: 'Build context from the latest build chat.' }),
+    })
+
+    await user.click(screen.getByRole('button', {
+      name: /agentV2\.agentDetail\.configure\.files\.remove.*build_note\.md/,
+    }))
+
+    expect(screen.queryByText('build_note.md')).not.toBeInTheDocument()
+    expect(mocks.deleteFileMutationFn).not.toHaveBeenCalled()
+
+    const snapshot = JSON.parse(screen.getByTestId('config-snapshot-probe').textContent ?? '{}')
+    expect(snapshot.config_note).toBe('')
+  })
+
   it('should keep flat config files visible without drive-prefix filtering and disable add in read-only mode', () => {
     renderAgentFiles({ readOnly: true })
 
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/index.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/index.tsx
index 54099b7fad5..da9fa0bbc35 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/index.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/index.tsx
@@ -4,22 +4,29 @@ import type { ReactNode } from 'react'
 import type { AgentOrchestrateAddActionOptions } from '../add-actions-context'
 import type { AgentConfigApiContext } from '../config-context'
 import type { AgentFileNode } from '@/features/agent-v2/agent-composer/form-state'
+import { cn } from '@langgenius/dify-ui/cn'
 import {
   Dialog,
   DialogTrigger,
 } from '@langgenius/dify-ui/dialog'
 import {
+  FileTreeBadge,
   FileTreeGuide,
+  FileTreeIcon,
+  FileTreeLabel,
 } from '@langgenius/dify-ui/file-tree'
 import { useMutation, useQuery } from '@tanstack/react-query'
 import { useAtomValue, useSetAtom } from 'jotai'
 import { useCallback, useRef, useState } from 'react'
-import { useTranslation } from 'react-i18next'
-import { agentComposerOriginalConfigAtom } from '@/features/agent-v2/agent-composer/store'
+import { Trans, useTranslation } from 'react-i18next'
+import { Infotip } from '@/app/components/base/infotip'
+import { useDocLink } from '@/context/i18n'
+import { agentComposerDraftAtom } from '@/features/agent-v2/agent-composer/store'
 import { agentComposerFilesAtom } from '@/features/agent-v2/agent-composer/store-modules/files'
 import { consoleQuery } from '@/service/client'
 import { useRegisterAgentOrchestrateAddAction } from '../add-actions-context'
 import { ConfigureSectionAddButton } from '../common/add-button'
+import { DocsLink } from '../common/docs-link'
 import { ConfigureSectionEmpty } from '../common/empty'
 import { ConfigureSection } from '../common/section'
 import { AgentConfigureTipContent } from '../common/tip-content'
@@ -91,6 +98,7 @@ function AgentFileItem({
   const selectedFile = selectedFileId ? findAgentFileNode(files, selectedFileId) : undefined
   const selectedPreviewFile = selectedFile ?? file
   const isVirtualPreviewFile = selectedPreviewFile.virtualContent !== undefined
+  const isBuildNoteFile = file.id === BUILD_NOTE_FILE_ID
   const previewFileId = isVirtualPreviewFile ? undefined : getAgentFilePreviewKey(selectedPreviewFile)
   const agentPreviewQuery = useQuery({
     ...consoleQuery.agent.byAgentId.config.files.byName.preview.get.queryOptions({
@@ -176,14 +184,17 @@ function AgentFileItem({
               type="button"
               data-selected={selected || undefined}
               aria-current={selected ? 'true' : undefined}
-              className="group/file-tree-row relative flex h-6 w-full min-w-0 cursor-pointer items-center rounded-md pr-7 pl-2 text-left outline-hidden select-none hover:bg-state-base-hover focus-visible:bg-state-base-hover focus-visible:inset-ring-2 focus-visible:inset-ring-state-accent-solid data-[selected]:bg-state-base-active"
+              className={cn(
+                'group/file-tree-row relative flex h-6 w-full min-w-0 cursor-pointer items-center rounded-md pl-2 text-left outline-hidden select-none group-hover/file-row:bg-state-base-hover hover:bg-state-base-hover focus-visible:bg-state-base-hover focus-visible:inset-ring-2 focus-visible:inset-ring-state-accent-solid data-[selected]:bg-state-base-active',
+                'pr-7',
+              )}
             />
           )}
         >
           {Array.from({ length: Math.max(depth - 1, 0) }, (_, index) => (
             <FileTreeGuide key={index} />
           ))}
-          <div className="flex min-w-0 flex-[1_0_0] items-center py-0.5">
+          <div className="flex min-w-0 flex-1 items-center overflow-hidden py-0.5">
             {children}
           </div>
         </DialogTrigger>
@@ -209,7 +220,12 @@ function AgentFileItem({
           }}
         />
       </Dialog>
-      {!readOnly && !file.virtualContent && (
+      {isBuildNoteFile && (
+        <AgentBuildNoteInfotip
+          className={cn(!readOnly && 'group-focus-within/file-row:opacity-0 group-hover/file-row:opacity-0')}
+        />
+      )}
+      {!readOnly && (!file.virtualContent || isBuildNoteFile) && (
         <button
           type="button"
           data-agent-file-remove-button
@@ -224,6 +240,51 @@ function AgentFileItem({
   )
 }
 
+function AgentBuildNoteFileRow() {
+  const { t } = useTranslation('agentV2')
+
+  return (
+    <>
+      <FileTreeIcon type="markdown" />
+      <FileTreeLabel title={BUILD_NOTE_FILE_NAME}>
+        {BUILD_NOTE_FILE_NAME}
+      </FileTreeLabel>
+      <FileTreeBadge className="ml-0.5 gap-0.5 px-1 py-0.5">
+        <span aria-hidden className="i-ri-sparkling-line size-3 shrink-0" />
+        <span>{t('agentDetail.configure.files.buildNote.generated')}</span>
+      </FileTreeBadge>
+    </>
+  )
+}
+
+function AgentBuildNoteInfotip({
+  className,
+}: {
+  className?: string
+}) {
+  const { t } = useTranslation('agentV2')
+  const docLink = useDocLink()
+
+  return (
+    <Infotip
+      aria-label={t('agentDetail.configure.files.buildNote.tooltip')}
+      className={cn('absolute top-1/2 right-1 z-10 size-5 -translate-y-1/2', className)}
+      iconClassName="size-4"
+      popupClassName="w-[230px] rounded-xl bg-components-tooltip-bg px-4 py-3.5 text-text-secondary shadow-lg backdrop-blur-[5px]"
+    >
+      <p className="body-xs-regular text-text-secondary">
+        <Trans
+          i18nKey="agentDetail.configure.files.buildNote.richTooltip"
+          ns="agentV2"
+          components={{
+            docLink: <DocsLink href={docLink('/use-dify/build/new-agent/build#the-build-note')} />,
+          }}
+        />
+      </p>
+    </Infotip>
+  )
+}
+
 export function AgentFiles() {
   const { t } = useTranslation('agentV2')
   const filesTip = t('agentDetail.configure.files.tip')
@@ -231,14 +292,23 @@ export function AgentFiles() {
   const [isUploadOpen, setIsUploadOpen] = useState(false)
   const promptAddCallbackRef = useRef<AgentOrchestrateAddActionOptions['onAdded']>(undefined)
   const apiContext = useAgentConfigApiContext()
-  const originalConfig = useAtomValue(agentComposerOriginalConfigAtom)
+  const draft = useAtomValue(agentComposerDraftAtom)
+  const setDraft = useSetAtom(agentComposerDraftAtom)
   const files = useAtomValue(agentComposerFilesAtom)
   const setFiles = useSetAtom(agentComposerFilesAtom)
-  const buildNoteFile = getBuildNoteFile(originalConfig?.config_note)
+  const buildNoteFile = getBuildNoteFile(draft.configNote)
   const visibleFiles = buildNoteFile ? [buildNoteFile, ...files] : files
   const { mutate: deleteAgentFile } = useMutation(consoleQuery.agent.byAgentId.config.files.byName.delete.mutationOptions())
   const { mutate: deleteWorkflowAgentFile } = useMutation(consoleQuery.apps.byAppId.agent.config.files.byName.delete.mutationOptions())
   const removeFile = useCallback((fileId: string) => {
+    if (fileId === BUILD_NOTE_FILE_ID) {
+      setDraft(draft => ({
+        ...draft,
+        configNote: '',
+      }))
+      return
+    }
+
     const file = findAgentFileNode(files, fileId)
     const configName = file?.configName ?? file?.name
 
@@ -273,7 +343,7 @@ export function AgentFiles() {
         version_id: apiContext.versionId,
       },
     }, { onSuccess })
-  }, [apiContext, deleteAgentFile, deleteWorkflowAgentFile, files, setFiles])
+  }, [apiContext, deleteAgentFile, deleteWorkflowAgentFile, files, setDraft, setFiles])
   const handleOpenUpload = useCallback((options?: AgentOrchestrateAddActionOptions) => {
     promptAddCallbackRef.current = options?.onAdded
     setIsUploadOpen(true)
@@ -324,18 +394,22 @@ export function AgentFiles() {
                 treeLabel={t('agentDetail.configure.files.treeLabel')}
                 className="rounded-lg border-[0.5px] border-components-panel-border bg-components-panel-on-panel-item-bg p-1 shadow-xs shadow-shadow-shadow-3"
                 scrollAreaClassName="max-h-[250px] flex-none"
-                renderFile={({ depth, file, selected, children }) => (
-                  <AgentFileItem
-                    depth={depth}
-                    file={file}
-                    files={visibleFiles}
-                    apiContext={apiContext}
-                    selected={selected}
-                    onRemove={removeFile}
-                  >
-                    {children}
-                  </AgentFileItem>
-                )}
+                renderFile={({ depth, file, selected, children }) => {
+                  const isBuildNoteFile = file.id === BUILD_NOTE_FILE_ID
+
+                  return (
+                    <AgentFileItem
+                      depth={depth}
+                      file={file}
+                      files={visibleFiles}
+                      apiContext={apiContext}
+                      selected={selected}
+                      onRemove={removeFile}
+                    >
+                      {isBuildNoteFile ? <AgentBuildNoteFileRow /> : children}
+                    </AgentFileItem>
+                  )
+                }}
               />
             )}
       </ConfigureSection>
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/tree.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/tree.tsx
index 6950191e457..c1fb78524ea 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/tree.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/tree.tsx
@@ -52,6 +52,7 @@ function AgentFileTreeRows({
   folderOpenStrategy,
   folderOpenState,
   onFolderOpenChange,
+  onFolderDoubleClick,
   onFolderOpen,
   renderFile,
   renderFolderSuffix,
@@ -63,6 +64,7 @@ function AgentFileTreeRows({
   folderOpenStrategy: AgentFileTreeFolderOpenStrategy
   folderOpenState?: AgentFileTreeFolderOpenState
   onFolderOpenChange?: (context: { file: AgentFileNode, depth: number, open: boolean }) => void
+  onFolderDoubleClick?: (context: { file: AgentFileNode, depth: number }) => void
   onFolderOpen?: (file: AgentFileNode) => void
   renderFile: AgentFileTreeRenderFile
   renderFolderSuffix?: AgentFileTreeRenderFolderSuffix
@@ -84,7 +86,10 @@ function AgentFileTreeRows({
           open={folderOpenState?.({ file, depth })}
           onOpenChange={open => onFolderOpenChange?.({ file, depth, open })}
         >
-          <FileTreeFolderTrigger onClick={() => onFolderOpen?.(file)}>
+          <FileTreeFolderTrigger
+            onClick={() => onFolderOpen?.(file)}
+            onDoubleClick={() => onFolderDoubleClick?.({ file, depth })}
+          >
             <FileTreeIcon type="folder" />
             <FileTreeLabel className="max-w-full" title={file.name}>{file.name}</FileTreeLabel>
             {renderFolderSuffix?.({ depth, file })}
@@ -98,6 +103,7 @@ function AgentFileTreeRows({
               folderOpenStrategy={folderOpenStrategy}
               folderOpenState={folderOpenState}
               onFolderOpenChange={onFolderOpenChange}
+              onFolderDoubleClick={onFolderDoubleClick}
               onFolderOpen={onFolderOpen}
               renderFile={renderFile}
               renderFolderSuffix={renderFolderSuffix}
@@ -121,8 +127,8 @@ function AgentFileTreeRows({
   })
 }
 
-const defaultRenderFile: AgentFileTreeRenderFile = ({ selected, children }) => (
-  <FileTreeFile selected={selected}>
+const defaultRenderFile: AgentFileTreeRenderFile = ({ depth, selected, children }) => (
+  <FileTreeFile level={depth} selected={selected}>
     {children}
   </FileTreeFile>
 )
@@ -142,6 +148,7 @@ export function AgentFileTree({
   folderOpenStrategy = firstLevelFolderOpenStrategy,
   folderOpenState,
   onFolderOpenChange,
+  onFolderDoubleClick,
   onFolderOpen,
   renderFile = defaultRenderFile,
   renderFolderSuffix,
@@ -161,6 +168,7 @@ export function AgentFileTree({
   folderOpenStrategy?: AgentFileTreeFolderOpenStrategy
   folderOpenState?: AgentFileTreeFolderOpenState
   onFolderOpenChange?: (context: { file: AgentFileNode, depth: number, open: boolean }) => void
+  onFolderDoubleClick?: (context: { file: AgentFileNode, depth: number }) => void
   onFolderOpen?: (file: AgentFileNode) => void
   renderFile?: AgentFileTreeRenderFile
   renderFolderSuffix?: AgentFileTreeRenderFolderSuffix
@@ -192,6 +200,7 @@ export function AgentFileTree({
               folderOpenStrategy={folderOpenStrategy}
               folderOpenState={folderOpenState}
               onFolderOpenChange={onFolderOpenChange}
+              onFolderDoubleClick={onFolderDoubleClick}
               onFolderOpen={onFolderOpen}
               renderFile={renderFile}
               renderFolderSuffix={renderFolderSuffix}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/upload-dialog.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/upload-dialog.tsx
index f499ead02de..1db662f58f5 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/upload-dialog.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/files/upload-dialog.tsx
@@ -148,8 +148,8 @@ function AgentFileUploader({
           <div className="flex items-center justify-center p-3">
             <FileTreeIcon type={getFileIconType(file.name, file.type)} />
           </div>
-          <div className="flex grow flex-col items-start gap-0.5 py-1 pr-2">
-            <span className="max-w-[calc(100%-30px)] overflow-hidden text-[12px] leading-4 font-medium text-ellipsis whitespace-nowrap text-text-secondary">{file.name}</span>
+          <div className="flex min-w-0 grow flex-col items-start gap-0.5 py-1 pr-2">
+            <span className="max-w-full min-w-0 truncate text-[12px] leading-4 font-medium text-text-secondary">{file.name}</span>
             <div className="flex h-3 items-center gap-1 self-stretch text-[10px] leading-3 font-medium text-text-tertiary uppercase">
               <span>{t('agentDetail.configure.files.upload.fileType')}</span>
               <span className="text-text-quaternary">·</span>
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/index.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/index.tsx
index ac33c14e773..894265d0ad6 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/index.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/index.tsx
@@ -157,7 +157,7 @@ export function AgentOrchestratePanel({
 
       {orchestrateBottomAction
         ? (
-            <AgentOrchestrateBottomActions>
+            <AgentOrchestrateBottomActions shrinkOnOpen={!bottomAction}>
               {orchestrateBottomAction}
             </AgentOrchestrateBottomActions>
           )
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/__tests__/index.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/__tests__/index.spec.tsx
index 58892224a8a..25fba5b077f 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/__tests__/index.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/__tests__/index.spec.tsx
@@ -1,13 +1,15 @@
 import type { AgentSoulConfigFormState } from '@/features/agent-v2/agent-composer/form-state'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
-import { fireEvent, render, screen, within } from '@testing-library/react'
+import { fireEvent, render, screen, waitFor, within } from '@testing-library/react'
 import userEvent from '@testing-library/user-event'
 import { useAtomValue } from 'jotai'
 import { beforeEach, describe, expect, it, vi } from 'vitest'
+import { MetadataFilteringModeEnum } from '@/app/components/workflow/nodes/knowledge-retrieval/types'
 import { formStateToAgentSoulConfig } from '@/features/agent-v2/agent-composer/conversions'
 import { defaultAgentSoulConfigFormState } from '@/features/agent-v2/agent-composer/form-state'
 import { AgentComposerProvider } from '@/features/agent-v2/agent-composer/provider'
 import { agentComposerDraftAtom } from '@/features/agent-v2/agent-composer/store'
+import { RerankingModeEnum } from '@/models/datasets'
 import { AgentOrchestrateReadOnlyContext } from '../../read-only-context'
 import { AgentKnowledgeRetrieval } from '../index'
 
@@ -15,13 +17,35 @@ vi.mock('@/app/components/workflow/nodes/knowledge-retrieval/components/add-data
   default: function MockAddKnowledge({
     onChange,
   }: {
-    onChange: (datasets: Array<{ id: string, name: string }>) => void
+    onChange: (datasets: Array<{
+      id: string
+      name: string
+      indexing_technique: string
+      provider: string
+      embedding_model_provider: string
+      embedding_model: string
+      retrieval_model_dict: {
+        search_method: string
+      }
+      is_multimodal: boolean
+    }>) => void
   }) {
     return (
       <button
         type="button"
         aria-label="common.operation.add workflow.nodes.knowledgeRetrieval.knowledge"
-        onClick={() => onChange([{ id: 'dataset-2', name: 'Release Docs' }])}
+        onClick={() => onChange([{
+          id: 'dataset-2',
+          name: 'Release Docs',
+          indexing_technique: 'high_quality',
+          provider: 'internal',
+          embedding_model_provider: 'openai',
+          embedding_model: 'text-embedding-3',
+          retrieval_model_dict: {
+            search_method: 'semantic',
+          },
+          is_multimodal: false,
+        }])}
       >
         Add mock knowledge
       </button>
@@ -29,6 +53,25 @@ vi.mock('@/app/components/workflow/nodes/knowledge-retrieval/components/add-data
   },
 }))
 
+vi.mock('@/app/components/header/account-setting/model-provider-page/hooks', () => ({
+  useModelListAndDefaultModelAndCurrentProviderAndModel: vi.fn(() => ({
+    modelList: [{
+      provider: 'rerank-provider',
+      models: [{ model: 'rerank-model' }],
+    }],
+    defaultModel: {
+      provider: {
+        provider: 'rerank-provider',
+      },
+      model: 'rerank-model',
+    },
+  })),
+  useCurrentProviderAndModel: vi.fn(() => ({
+    currentProvider: { provider: 'rerank-provider' },
+    currentModel: { model: 'rerank-model' },
+  })),
+}))
+
 const agentKnowledgeDraft = {
   ...defaultAgentSoulConfigFormState,
   knowledgeRetrievals: [
@@ -147,7 +190,7 @@ describe('AgentKnowledgeRetrieval', () => {
       expect(within(dialog).getByRole('textbox', {
         name: 'agentV2.agentDetail.configure.knowledgeRetrieval.dialog.nameLabel',
       })).toHaveValue('agentV2.agentDetail.configure.knowledgeRetrieval.retrievalTwo')
-      expect(within(dialog).getByText('appDebug.datasetConfig.knowledgeTip')).toBeInTheDocument()
+      expect(within(dialog).queryByText('appDebug.datasetConfig.knowledgeTip')).not.toBeInTheDocument()
       expect(within(dialog).getByRole('button', {
         name: 'common.operation.add workflow.nodes.knowledgeRetrieval.knowledge',
       })).toBeInTheDocument()
@@ -220,6 +263,33 @@ describe('AgentKnowledgeRetrieval', () => {
       expect(within(dialog).getByText('common.errorMsg.fieldRequired:{"field":"agentV2.agentDetail.configure.knowledgeRetrieval.dialog.query.customInputLabel"}')).toBeInTheDocument()
     })
 
+    it('should not show inline validation for automatic metadata filtering without a model', async () => {
+      const user = userEvent.setup()
+      renderKnowledgeRetrieval({
+        initialDraft: {
+          ...defaultAgentSoulConfigFormState,
+          knowledgeRetrievals: [
+            {
+              id: 'retrieval-1',
+              name: 'Docs Search',
+              datasetRefs: [{ id: 'dataset-1', name: 'Docs' }],
+              metadataFilterMode: MetadataFilteringModeEnum.automatic,
+            },
+          ],
+        },
+      })
+
+      await user.click(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.knowledgeRetrieval.edit:{"name":"Docs Search"}',
+      }))
+
+      const dialog = screen.getByRole('dialog', {
+        name: 'agentV2.agentDetail.configure.knowledgeRetrieval.dialog.title',
+      })
+
+      expect(within(dialog).queryByText('agentV2.agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired')).not.toBeInTheDocument()
+    })
+
     it('should show duplicate-name validation in the dialog', async () => {
       const user = userEvent.setup()
       renderKnowledgeRetrieval({
@@ -308,6 +378,12 @@ describe('AgentKnowledgeRetrieval', () => {
           },
           retrieval: expect.objectContaining({
             mode: 'multiple',
+            reranking_enable: true,
+            reranking_mode: RerankingModeEnum.RerankingModel,
+            reranking_model: {
+              provider: 'rerank-provider',
+              model: 'rerank-model',
+            },
             top_k: 4,
           }),
         }),
@@ -367,6 +443,63 @@ describe('AgentKnowledgeRetrieval', () => {
       expect(within(dialog).queryByText('appDebug.datasetConfig.knowledgeTip')).not.toBeInTheDocument()
     })
 
+    it('should save the default rerank model when editing retrieval with existing knowledge', async () => {
+      const user = userEvent.setup()
+      renderKnowledgeRetrieval({
+        showConfigSnapshot: true,
+        initialDraft: {
+          ...defaultAgentSoulConfigFormState,
+          knowledgeRetrievals: [
+            {
+              id: 'retrieval-1',
+              name: 'Search Docs',
+              datasetRefs: [
+                {
+                  id: 'dataset-1',
+                  name: 'Product Docs',
+                  description: 'Docs corpus',
+                },
+              ],
+              multipleRetrievalConfig: {
+                top_k: 4,
+                score_threshold: null,
+                reranking_enable: false,
+              },
+            },
+          ],
+        },
+      })
+
+      await user.click(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.knowledgeRetrieval.edit:{"name":"Search Docs"}',
+      }))
+
+      const dialog = screen.getByRole('dialog', {
+        name: 'agentV2.agentDetail.configure.knowledgeRetrieval.dialog.title',
+      })
+      await user.click(within(dialog).getByRole('button', {
+        name: 'Search Docs',
+      }))
+      const nameInput = within(dialog).getByRole('textbox', {
+        name: 'agentV2.agentDetail.configure.knowledgeRetrieval.dialog.nameLabel',
+      })
+      await user.clear(nameInput)
+      await user.type(nameInput, 'Default Rerank Docs')
+
+      await waitFor(() => {
+        const knowledgeConfig = JSON.parse(screen.getByLabelText('config snapshot').textContent ?? '{}')
+        expect(knowledgeConfig.sets[0].retrieval).toEqual(expect.objectContaining({
+          mode: 'multiple',
+          reranking_enable: true,
+          reranking_mode: RerankingModeEnum.RerankingModel,
+          reranking_model: {
+            provider: 'rerank-provider',
+            model: 'rerank-model',
+          },
+        }))
+      })
+    })
+
     it('should save edited retrieval data into the config snapshot', async () => {
       const user = userEvent.setup()
       renderKnowledgeRetrieval({ showConfigSnapshot: true })
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/dialog.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/dialog.tsx
index d809521336b..497040367c2 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/dialog.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/dialog.tsx
@@ -19,9 +19,11 @@ import { RadioGroup } from '@langgenius/dify-ui/radio-group'
 import { Textarea } from '@langgenius/dify-ui/textarea'
 import { intersectionBy } from 'es-toolkit/compat'
 import { useAtomValue } from 'jotai'
-import { useEffect, useMemo, useRef, useState } from 'react'
+import { useCallback, useEffect, useMemo, useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { IndexingType } from '@/app/components/datasets/create/step-two/hooks/use-indexing-config'
+import { ModelTypeEnum } from '@/app/components/header/account-setting/model-provider-page/declarations'
+import { useCurrentProviderAndModel, useModelListAndDefaultModelAndCurrentProviderAndModel } from '@/app/components/header/account-setting/model-provider-page/hooks'
 import Field from '@/app/components/workflow/nodes/_base/components/field'
 import AddKnowledge from '@/app/components/workflow/nodes/knowledge-retrieval/components/add-dataset'
 import DatasetList from '@/app/components/workflow/nodes/knowledge-retrieval/components/dataset-list'
@@ -33,6 +35,7 @@ import {
   MetadataFilteringVariableType,
   MetadataFilteringModeEnum as WorkflowMetadataFilteringModeEnum,
 } from '@/app/components/workflow/nodes/knowledge-retrieval/types'
+import { getMultipleRetrievalConfig } from '@/app/components/workflow/nodes/knowledge-retrieval/utils'
 import { DATASET_DEFAULT } from '@/config'
 import { useDocLink } from '@/context/i18n'
 import { useKnowledgeValidationMessage, validateKnowledgeRetrievals } from '@/features/agent-v2/agent-composer/knowledge-validation'
@@ -254,6 +257,42 @@ export function AgentKnowledgeRetrievalDialog({
     selectedDatasets,
     singleRetrievalConfig,
   } = dialogState
+  const {
+    modelList: rerankModelList,
+    defaultModel: rerankDefaultModel,
+  } = useModelListAndDefaultModelAndCurrentProviderAndModel(ModelTypeEnum.rerank)
+  const {
+    currentModel: currentRerankModel,
+    currentProvider: currentRerankProvider,
+  } = useCurrentProviderAndModel(
+    rerankModelList,
+    rerankDefaultModel
+      ? {
+          ...rerankDefaultModel,
+          provider: rerankDefaultModel.provider.provider,
+        }
+      : undefined,
+  )
+  const fallbackRerankModel = useMemo(() => ({
+    provider: currentRerankProvider?.provider,
+    model: currentRerankModel?.model,
+  }), [currentRerankModel?.model, currentRerankProvider?.provider])
+  const resolveMultipleRetrievalConfig = useCallback((
+    config: MultipleRetrievalConfig,
+    nextSelectedDatasets = selectedDatasets,
+    originalDatasets = selectedDatasets,
+  ) => getMultipleRetrievalConfig(
+    config,
+    nextSelectedDatasets,
+    originalDatasets,
+    fallbackRerankModel,
+  ), [fallbackRerankModel, selectedDatasets])
+  const effectiveMultipleRetrievalConfig = useMemo(() => {
+    if (retrievalMode !== RETRIEVE_TYPE.multiWay || selectedDatasets.length === 0)
+      return multipleRetrievalConfig
+
+    return resolveMultipleRetrievalConfig(multipleRetrievalConfig)
+  }, [multipleRetrievalConfig, resolveMultipleRetrievalConfig, retrievalMode, selectedDatasets.length])
   const patchDialogState = (patch: Partial<KnowledgeRetrievalDialogState>) => {
     setDialogState(current => ({
       ...current,
@@ -303,7 +342,7 @@ export function AgentKnowledgeRetrievalDialog({
     customQuery,
     selectedDatasets,
     retrievalMode,
-    multipleRetrievalConfig,
+    multipleRetrievalConfig: effectiveMultipleRetrievalConfig,
     singleRetrievalConfig,
     metadataFilterMode,
     metadataFilteringConditions,
@@ -321,7 +360,7 @@ export function AgentKnowledgeRetrievalDialog({
       customQuery,
       selectedDatasets,
       retrievalMode,
-      multipleRetrievalConfig,
+      multipleRetrievalConfig: effectiveMultipleRetrievalConfig,
       singleRetrievalConfig,
       metadataFilterMode,
       metadataFilteringConditions,
@@ -330,15 +369,29 @@ export function AgentKnowledgeRetrievalDialog({
     })
   }
   const handleSelectedDatasetsChange = (nextDatasets: DataSet[]) => {
-    patchDialogState({ selectedDatasets: nextDatasets })
+    const nextMultipleRetrievalConfig = retrievalMode === RETRIEVE_TYPE.multiWay && nextDatasets.length > 0
+      ? resolveMultipleRetrievalConfig(multipleRetrievalConfig, nextDatasets)
+      : multipleRetrievalConfig
+
+    patchDialogState({
+      multipleRetrievalConfig: nextMultipleRetrievalConfig,
+      selectedDatasets: nextDatasets,
+    })
 
     if (item) {
-      updateItem({ selectedDatasets: nextDatasets })
+      updateItem({
+        multipleRetrievalConfig: nextMultipleRetrievalConfig,
+        selectedDatasets: nextDatasets,
+      })
       return
     }
 
-    if (nextDatasets.length > 0)
-      onItemCreate?.(createItemFromDialogState({ selectedDatasets: nextDatasets }))
+    if (nextDatasets.length > 0) {
+      onItemCreate?.(createItemFromDialogState({
+        multipleRetrievalConfig: nextMultipleRetrievalConfig,
+        selectedDatasets: nextDatasets,
+      }))
+    }
   }
   const metadataList = useMemo(() => {
     const datasetsWithMetadata = selectedDatasets.filter(dataset => !!dataset.doc_metadata)
@@ -354,7 +407,9 @@ export function AgentKnowledgeRetrievalDialog({
   const datasetsError = getValidationMessage(itemValidation?.datasets)
   const queryError = getValidationMessage(itemValidation?.query)
   const retrievalError = getValidationMessage(itemValidation?.retrieval)
-  const metadataError = getValidationMessage(itemValidation?.metadata)
+  const metadataError = itemValidation?.metadata === 'metadata_model_required'
+    ? undefined
+    : getValidationMessage(itemValidation?.metadata)
 
   useEffect(() => {
     if (hydratedKey !== hydrationKey) {
@@ -500,16 +555,28 @@ export function AgentKnowledgeRetrievalDialog({
                   <RetrievalConfig
                     payload={{
                       retrieval_mode: retrievalMode,
-                      multiple_retrieval_config: multipleRetrievalConfig,
+                      multiple_retrieval_config: effectiveMultipleRetrievalConfig,
                       single_retrieval_config: singleRetrievalConfig,
                     }}
                     onRetrievalModeChange={(nextRetrievalMode) => {
-                      patchDialogState({ retrievalMode: nextRetrievalMode })
-                      updateItem({ retrievalMode: nextRetrievalMode })
+                      const nextMultipleRetrievalConfig = nextRetrievalMode === RETRIEVE_TYPE.multiWay
+                        ? resolveMultipleRetrievalConfig(multipleRetrievalConfig)
+                        : multipleRetrievalConfig
+
+                      patchDialogState({
+                        multipleRetrievalConfig: nextMultipleRetrievalConfig,
+                        retrievalMode: nextRetrievalMode,
+                      })
+                      updateItem({
+                        multipleRetrievalConfig: nextMultipleRetrievalConfig,
+                        retrievalMode: nextRetrievalMode,
+                      })
                     }}
                     onMultipleRetrievalConfigChange={(nextMultipleRetrievalConfig) => {
-                      patchDialogState({ multipleRetrievalConfig: nextMultipleRetrievalConfig })
-                      updateItem({ multipleRetrievalConfig: nextMultipleRetrievalConfig })
+                      const normalizedMultipleRetrievalConfig = resolveMultipleRetrievalConfig(nextMultipleRetrievalConfig)
+
+                      patchDialogState({ multipleRetrievalConfig: normalizedMultipleRetrievalConfig })
+                      updateItem({ multipleRetrievalConfig: normalizedMultipleRetrievalConfig })
                     }}
                     singleRetrievalModelConfig={singleRetrievalConfig?.model}
                     onSingleRetrievalModelChange={(model) => {
@@ -556,14 +623,16 @@ export function AgentKnowledgeRetrievalDialog({
               )}
             >
               <>
-                <DatasetList
-                  list={selectedDatasets}
-                  onChange={handleSelectedDatasetsChange}
-                  settingsDrawerBackdropClassName="bg-background-overlay"
-                  settingsDrawerBackdropForceRender
-                  settingsDrawerPopupClassName="data-[swipe-direction=right]:top-6 data-[swipe-direction=right]:bottom-6"
-                  settingsModalHeight="100%"
-                />
+                {selectedDatasets.length > 0 && (
+                  <DatasetList
+                    list={selectedDatasets}
+                    onChange={handleSelectedDatasetsChange}
+                    settingsDrawerBackdropClassName="bg-background-overlay"
+                    settingsDrawerBackdropForceRender
+                    settingsDrawerPopupClassName="data-[swipe-direction=right]:top-6 data-[swipe-direction=right]:bottom-6"
+                    settingsModalHeight="100%"
+                  />
+                )}
                 {(datasetsError || retrievalError) && (
                   <div role="alert" className="pt-2 system-xs-regular text-text-destructive">
                     {datasetsError ?? retrievalError}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/index.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/index.tsx
index 1fd3b4d5508..c0feb0ed44f 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/index.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/knowledge/index.tsx
@@ -11,6 +11,7 @@ import { ConfigureSectionAddButton } from '../common/add-button'
 import { ConfigureSectionConfigurableItem } from '../common/configurable-item'
 import { ConfigureSectionEmpty } from '../common/empty'
 import { ConfigureSection } from '../common/section'
+import { AgentConfigureTipContent } from '../common/tip-content'
 import { AgentKnowledgeRetrievalDialog } from './dialog'
 
 function KnowledgeRetrievalIcon() {
@@ -87,7 +88,7 @@ export function AgentKnowledgeRetrieval() {
         label={t('agentDetail.configure.knowledgeRetrieval.label')}
         labelId="agent-configure-knowledge-retrieval-label"
         panelId={retrievalListId}
-        tip={knowledgeRetrievalTip}
+        tip={<AgentConfigureTipContent type="knowledge" />}
         tipAriaLabel={knowledgeRetrievalTip}
         rootClassName="border-b border-divider-subtle pt-4"
         panelContentClassName="flex flex-col gap-1 pb-4"
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/model-config/field.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/model-config/field.tsx
index 3a937a9bc66..88005d68c5f 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/model-config/field.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/model-config/field.tsx
@@ -7,7 +7,7 @@ import { Tooltip, TooltipContent, TooltipTrigger } from '@langgenius/dify-ui/too
 import { useTranslation } from 'react-i18next'
 import ModelParameterModal from '@/app/components/header/account-setting/model-provider-page/model-parameter-modal'
 import ModelSelector from '@/app/components/header/account-setting/model-provider-page/model-selector'
-import { isAgentCompatibleModel } from '../../../model-compatibility'
+import { isAgentCompatibleModel, isAgentSuggestedModel } from '../../../model-compatibility'
 import { useAgentOrchestrateReadOnly } from '../read-only-context'
 
 type AgentModelFieldProps = {
@@ -48,6 +48,7 @@ export function AgentModelField({
                   providerSettingsSource="agent"
                   showModelMeta={false}
                   modelPredicate={isAgentCompatibleModel}
+                  modelSuggestionPredicate={isAgentSuggestedModel}
                   onSelect={onSelect}
                 />
                 <div className="w-8 shrink-0">
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/publish-bar/index.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/publish-bar/index.tsx
index e526dc8e015..9ce15bed43c 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/publish-bar/index.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/publish-bar/index.tsx
@@ -12,9 +12,7 @@ import { useMutation, useQuery, useQueryClient } from '@tanstack/react-query'
 import { useAtomValue } from 'jotai'
 import { useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
-import { useKnowledgeValidationMessage, validateKnowledgeRetrievals } from '@/features/agent-v2/agent-composer/knowledge-validation'
 import { hasAgentComposerUnpublishedChangesAtom, isAgentComposerDirtyAtom } from '@/features/agent-v2/agent-composer/store'
-import { agentComposerKnowledgeRetrievalsAtom } from '@/features/agent-v2/agent-composer/store-modules/knowledge'
 import { useFormatTimeFromNow } from '@/hooks/use-format-time-from-now'
 import useTimestamp from '@/hooks/use-timestamp'
 import { consoleQuery } from '@/service/client'
@@ -109,9 +107,6 @@ export function AgentConfigurePublishBar({
   const stableActiveConfigIsPublished = activeConfigIsPublished ?? (lastKnownPublishedRef.current ? true : undefined)
   const hasUnpublishedChanges = useAtomValue(hasAgentComposerUnpublishedChangesAtom)
   const hasLocalChanges = useAtomValue(isAgentComposerDirtyAtom)
-  const knowledgeRetrievals = useAtomValue(agentComposerKnowledgeRetrievalsAtom)
-  const knowledgeValidation = validateKnowledgeRetrievals(knowledgeRetrievals)
-  const getValidationMessage = useKnowledgeValidationMessage()
   const publishableState = getPublishState({
     activeConfigIsPublished: stableActiveConfigIsPublished,
     activeConfigSnapshot,
@@ -127,7 +122,6 @@ export function AgentConfigurePublishBar({
     isPublishing,
   })
   const publishIsAvailable = !isPublishing && (publishableState === 'draft' || publishableState === 'unpublished')
-  const publishValidationMessage = getValidationMessage(knowledgeValidation.firstIssue?.code)
   const workflowReferencesQueryOptions = consoleQuery.agent.byAgentId.referencingWorkflows.get.queryOptions({
     input: {
       params: {
@@ -140,7 +134,7 @@ export function AgentConfigurePublishBar({
     enabled: publishIsAvailable && !selectedVersionSnapshot,
   })
   const restoreVersionMutation = useMutation(consoleQuery.agent.byAgentId.versions.byVersionId.restore.post.mutationOptions())
-  const canPublish = publishIsAvailable && knowledgeValidation.isValid
+  const canPublish = publishIsAvailable
 
   const handleRestoreVersion = (versionId: string) => {
     if (restoreVersionMutation.isPending)
@@ -284,9 +278,6 @@ export function AgentConfigurePublishBar({
   const currentStateMeta = stateMeta[publishState]
   const isConfirmingImpact = publishBarMode.status === 'confirmingImpact' && (canPublish || isPublishing)
   const impactReferences = publishBarMode.status === 'confirmingImpact' ? publishBarMode.references : []
-  const effectiveMetaLabel = publishValidationMessage && publishIsAvailable
-    ? publishValidationMessage
-    : currentStateMeta.metaLabel
 
   return (
     <CollapsibleRoot
@@ -305,7 +296,7 @@ export function AgentConfigurePublishBar({
         actionLabel={currentStateMeta.actionLabel}
         dotStatus={currentStateMeta.dotStatus}
         isPublishing={isPublishing}
-        metaLabel={effectiveMetaLabel}
+        metaLabel={currentStateMeta.metaLabel}
         showShortcut={currentStateMeta.showShortcut}
         statusLabel={currentStateMeta.statusLabel}
         canPublish={canPublish}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/__tests__/index.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/__tests__/index.spec.tsx
index fe8e7252430..3c3815e625d 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/__tests__/index.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/__tests__/index.spec.tsx
@@ -308,6 +308,29 @@ describe('AgentSkills', () => {
     expect(toast.success).toHaveBeenCalled()
   })
 
+  it('should not show the frontend fallback error when skill upload fails', async () => {
+    const user = userEvent.setup()
+    mocks.uploadSkillMutationFn.mockRejectedValueOnce(new Error('Backend upload error'))
+    renderAgentSkills({ initialDraft: defaultAgentSoulConfigFormState })
+
+    await user.click(screen.getByRole('button', { name: /agentV2\.agentDetail\.configure\.skills\.add/i }))
+
+    const input = await waitFor(() => {
+      const element = document.querySelector('input[type="file"]')
+      expect(element).not.toBeNull()
+      return element as HTMLInputElement
+    })
+    const file = new File(['skill'], 'invoice-helper.skill', { type: 'application/zip' })
+    await user.upload(input, file)
+    await user.click(screen.getByRole('button', { name: /agentDetail\.configure\.skills\.upload\.action/i }))
+
+    await waitFor(() => {
+      expect(mocks.uploadSkillMutationFn).toHaveBeenCalled()
+    })
+
+    expect(toast.error).not.toHaveBeenCalledWith('agentV2.agentDetail.configure.skills.upload.failed')
+  })
+
   it('should use workflow config skill endpoints with node_id for uploads and skill member queries', async () => {
     const user = userEvent.setup()
     renderAgentSkills({
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/detail-dialog.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/detail-dialog.tsx
index bd64cca23b8..36be64434e9 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/detail-dialog.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/detail-dialog.tsx
@@ -2,6 +2,7 @@
 
 import type { ReactNode } from 'react'
 import type { AgentFileNode } from '@/features/agent-v2/agent-composer/form-state'
+import { cn } from '@langgenius/dify-ui/cn'
 import {
   DialogCloseButton,
   DialogContent,
@@ -26,6 +27,11 @@ type AgentSkillDetailSection = {
 export type AgentSkillDetail = {
   description: string
   fileCount?: number
+  fileListHeader?: ReactNode
+  fileListLoading?: boolean
+  fileListPanelClassName?: string
+  fileListTreeClassName?: string
+  fileListTreeListClassName?: string
   fileListTitle?: string
   files: AgentSkillFileNode[]
   folderOpenState?: (context: { file: AgentSkillFileNode, depth: number }) => boolean
@@ -41,6 +47,7 @@ export type AgentSkillDetail = {
     isLoading?: boolean
   }
   onFolderOpenChange?: (context: { file: AgentSkillFileNode, depth: number, open: boolean }) => void
+  onFolderDoubleClick?: (context: { file: AgentSkillFileNode, depth: number }) => void
   onSelectFile?: (file: AgentSkillFileNode) => void
   renderFolderSuffix?: (context: { file: AgentSkillFileNode, depth: number }) => ReactNode
   selectedFileId?: string
@@ -50,48 +57,75 @@ export type AgentSkillDetail = {
 const keepSkillFoldersClosed = () => false
 
 function AgentSkillFileList({
+  fileListHeader,
+  fileListLoading,
+  fileListTreeClassName,
+  fileListTreeListClassName,
   fileListTitle,
   files,
   folderOpenState,
   onFolderOpenChange,
+  onFolderDoubleClick,
   onSelectFile,
   renderFolderSuffix,
   selectedFileId,
 }: {
+  fileListHeader?: ReactNode
+  fileListLoading?: boolean
+  fileListTreeClassName?: string
+  fileListTreeListClassName?: string
   fileListTitle?: string
   files: AgentSkillFileNode[]
   folderOpenState?: AgentSkillDetail['folderOpenState']
   onFolderOpenChange?: AgentSkillDetail['onFolderOpenChange']
+  onFolderDoubleClick?: AgentSkillDetail['onFolderDoubleClick']
   onSelectFile?: (file: AgentSkillFileNode) => void
   renderFolderSuffix?: AgentSkillDetail['renderFolderSuffix']
   selectedFileId?: string
 }) {
   const { t } = useTranslation('agentV2')
 
+  if (fileListLoading) {
+    return (
+      <div className={cn('flex h-full flex-col bg-background-section', fileListTreeClassName)}>
+        {fileListHeader ?? (
+          <h3 id="agent-skill-detail-files-heading" className="px-4 pt-3.5 pb-3 system-xl-semibold text-text-primary">
+            {fileListTitle ?? t('agentDetail.configure.skills.detail.files')}
+          </h3>
+        )}
+        <div className="flex min-h-0 flex-1 items-center justify-center">
+          <Loading type="area" />
+        </div>
+      </div>
+    )
+  }
+
   return (
     <AgentFileTree
       files={files}
       selectedFileId={selectedFileId}
       labelledBy="agent-skill-detail-files-heading"
-      className="h-full bg-background-section p-1"
+      className={cn('h-full bg-background-section p-1', fileListTreeClassName)}
+      listClassName={fileListTreeListClassName}
       scrollAreaClassName="flex-1"
       folderOpenStrategy={keepSkillFoldersClosed}
       folderOpenState={folderOpenState}
       onFolderOpenChange={onFolderOpenChange}
+      onFolderDoubleClick={onFolderDoubleClick}
       renderFile={onSelectFile
-        ? ({ file, selected, children }) => (
-            <FileTreeFile selected={selected} onClick={() => onSelectFile(file)}>
+        ? ({ depth, file, selected, children }) => (
+            <FileTreeFile level={depth} selected={selected} onClick={() => onSelectFile(file)}>
               {children}
             </FileTreeFile>
           )
         : undefined}
       renderFolderSuffix={renderFolderSuffix}
       header={(
-        <>
+        fileListHeader ?? (
           <h3 id="agent-skill-detail-files-heading" className="px-4 pt-3.5 pb-3 system-xl-semibold text-text-primary">
             {fileListTitle ?? t('agentDetail.configure.skills.detail.files')}
           </h3>
-        </>
+        )
       )}
     />
   )
@@ -239,7 +273,7 @@ export function AgentSkillDetailDialog({
 
   return (
     <DialogContent backdropProps={{ forceRender: true }} backdropClassName="fixed" className="flex h-[min(720px,calc(100dvh-2rem))] max-h-none w-[min(960px,calc(100vw-2rem))] flex-row overflow-hidden rounded-2xl p-0">
-      <div className="flex w-56 min-w-0 shrink-0 border-r-[0.5px] border-divider-subtle bg-background-section">
+      <div className={cn('flex w-56 min-w-0 shrink-0 border-r-[0.5px] border-divider-subtle bg-background-section', detail.fileListPanelClassName)}>
         <DialogDescription className="sr-only">
           {detail.description}
         </DialogDescription>
@@ -248,10 +282,15 @@ export function AgentSkillDetailDialog({
         </DialogTitle>
         <div className="min-h-0 w-full">
           <AgentSkillFileList
+            fileListHeader={detail.fileListHeader}
+            fileListLoading={detail.fileListLoading}
+            fileListTreeClassName={detail.fileListTreeClassName}
+            fileListTreeListClassName={detail.fileListTreeListClassName}
             fileListTitle={detail.fileListTitle}
             files={detail.files}
             folderOpenState={detail.folderOpenState}
             onFolderOpenChange={detail.onFolderOpenChange}
+            onFolderDoubleClick={detail.onFolderDoubleClick}
             selectedFileId={detail.selectedFileId}
             onSelectFile={detail.onSelectFile}
             renderFolderSuffix={detail.renderFolderSuffix}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/item.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/item.tsx
index daaa8e89a4c..f81e9d5d111 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/item.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/item.tsx
@@ -47,7 +47,7 @@ export function AgentSkillItem({
           onClick={handleOpenPreview}
         >
           <span aria-hidden className="i-custom-public-agent-building-blocks size-4 shrink-0" />
-          <span className="min-w-0 flex-1 truncate system-sm-medium text-text-secondary">
+          <span className="w-0 min-w-0 flex-1 truncate system-sm-medium text-text-secondary">
             {skill.name}
           </span>
           <span
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/upload-dialog.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/upload-dialog.tsx
index 86a53e15fdd..d67da232f14 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/upload-dialog.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/skills/upload-dialog.tsx
@@ -162,8 +162,8 @@ function AgentSkillPackageUploader({
           <div className="flex items-center justify-center p-3">
             <span aria-hidden className="i-custom-public-files-yaml size-6 shrink-0" />
           </div>
-          <div className="flex grow flex-col items-start gap-0.5 py-1 pr-2">
-            <span className="max-w-[calc(100%-30px)] overflow-hidden text-[12px] leading-4 font-medium text-ellipsis whitespace-nowrap text-text-secondary">{file.name}</span>
+          <div className="flex min-w-0 grow flex-col items-start gap-0.5 py-1 pr-2">
+            <span className="max-w-full min-w-0 truncate text-[12px] leading-4 font-medium text-text-secondary">{file.name}</span>
             <div className="flex h-3 items-center gap-1 self-stretch text-[10px] leading-3 font-medium text-text-tertiary uppercase">
               <span>{t('agentDetail.configure.skills.upload.fileType')}</span>
               <span className="text-text-quaternary">·</span>
@@ -212,9 +212,6 @@ export function AgentSkillUploadDialog({
         setFile(undefined)
         onOpenChange(false)
       },
-      onError: () => {
-        toast.error(t('agentDetail.configure.skills.upload.failed'))
-      },
     }
 
     if (apiContext.workflow) {
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/hooks.spec.ts b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/hooks.spec.ts
index 5afbb2b2771..e46a028ff27 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/hooks.spec.ts
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/hooks.spec.ts
@@ -29,6 +29,14 @@ const unauthorizedCredentialTool = {
   credentialRequired: true,
 } satisfies AgentProviderToolDefaultValue
 
+const unauthorizedOAuthTool = {
+  ...unauthorizedCredentialTool,
+  provider_id: 'slack',
+  provider_name: 'slack',
+  provider_show_name: 'Slack',
+  credentialType: 'oauth2',
+} satisfies AgentProviderToolDefaultValue
+
 describe('addProviderTools', () => {
   it('should not mark tools that do not need credentials as unauthorized', () => {
     const nextTools = addProviderTools([], [noCredentialTool])
@@ -53,4 +61,16 @@ describe('addProviderTools', () => {
       }),
     ])
   })
+
+  it('should preserve oauth credential type for credential-required OAuth tools', () => {
+    const nextTools = addProviderTools([], [unauthorizedOAuthTool])
+
+    expect(nextTools).toEqual([
+      expect.objectContaining({
+        credentialId: undefined,
+        credentialType: 'oauth2',
+        credentialVariant: 'unauthorized',
+      }),
+    ])
+  })
 })
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/index.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/index.spec.tsx
index 74ff5f7aab4..99bb919b7e7 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/index.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/__tests__/index.spec.tsx
@@ -1,3 +1,4 @@
+import type { AddOAuthButtonProps } from '@/app/components/plugins/plugin-auth/types'
 import type { ToolWithProvider } from '@/app/components/workflow/types'
 import type { AgentSoulConfigFormState } from '@/features/agent-v2/agent-composer/form-state'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
@@ -37,6 +38,23 @@ vi.mock('@/app/components/workflow/block-icon', () => ({
   ),
 }))
 
+vi.mock('@/app/components/plugins/plugin-auth/authorize/add-oauth-button', () => ({
+  default: ({ buttonText, onUpdate, renderTrigger }: AddOAuthButtonProps) => {
+    if (renderTrigger) {
+      return renderTrigger({
+        isConfigured: false,
+        onClick: () => onUpdate?.(),
+      })
+    }
+
+    return (
+      <button type="button" onClick={onUpdate}>
+        {buttonText}
+      </button>
+    )
+  },
+}))
+
 vi.mock('@/app/components/header/account-setting/model-provider-page/model-modal/Form', () => ({
   default: ({ formSchemas }: { formSchemas: Array<{ label?: Record<string, string>, variable?: string }> }) => (
     <div data-testid="tool-setting-form">
@@ -52,6 +70,7 @@ vi.mock('@/service/use-tools', () => ({
   useAllCustomTools: () => ({ data: [] }),
   useAllWorkflowTools: () => ({ data: [] }),
   useAllMCPTools: () => ({ data: [] }),
+  useInvalidToolsByType: () => vi.fn(),
 }))
 
 const agentToolsDraft = {
@@ -130,6 +149,28 @@ const reflectedUnauthorizedNoCredentialDraft = {
   ],
 } satisfies AgentSoulConfigFormState
 
+const reflectedUnauthorizedOAuthCredentialTypeDraft = {
+  ...defaultAgentSoulConfigFormState,
+  tools: [
+    {
+      id: 'google',
+      kind: 'provider',
+      name: 'google',
+      iconClassName: 'i-custom-public-other-default-tool-icon',
+      credentialType: 'unauthorized',
+      credentialVariant: 'none',
+      actions: [
+        {
+          id: 'google-search',
+          name: 'search',
+          toolName: 'search',
+          description: '',
+        },
+      ],
+    },
+  ],
+} satisfies AgentSoulConfigFormState
+
 const googleProvider = {
   id: 'google',
   name: 'google',
@@ -393,7 +434,10 @@ describe('AgentTools', () => {
   describe('Display Metadata', () => {
     it('should enrich reflected provider tools with provider icon and localized names', async () => {
       const user = userEvent.setup()
-      toolProviderState.builtInTools = [googleProvider]
+      toolProviderState.builtInTools = [{
+        ...googleProvider,
+        allow_delete: false,
+      }]
       renderAgentTools(reflectedAgentToolsDraft)
 
       expect(screen.getByRole('button', {
@@ -433,6 +477,21 @@ describe('AgentTools', () => {
       expect(store.get(isAgentComposerDirtyAtom)).toBe(false)
     })
 
+    it('should show authorization action for reflected OAuth provider tools with unauthorized credential type', () => {
+      toolProviderState.builtInTools = [{
+        ...googleProvider,
+        allow_delete: true,
+        is_team_authorization: false,
+        team_credentials: {},
+      }]
+      renderAgentTools(reflectedUnauthorizedOAuthCredentialTypeDraft)
+
+      expect(screen.getByRole('button', {
+        name: 'tools.notAuthorized',
+      })).toBeInTheDocument()
+      expect(screen.queryByText('plugin.auth.setupOAuth')).not.toBeInTheDocument()
+    })
+
     it('should open provider tool settings with catalog icon and parameters', async () => {
       const user = userEvent.setup()
       toolProviderState.builtInTools = [duckDuckGoProvider]
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/hooks.ts b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/hooks.ts
index 9be5cebfc91..9b11f711d02 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/hooks.ts
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/hooks.ts
@@ -40,6 +40,9 @@ const getCredentialType = (tool: AgentProviderToolDefaultValue) => {
   if (!tool.credentialRequired)
     return undefined
 
+  if (tool.credentialType === 'oauth2')
+    return 'oauth2' as const
+
   if (!tool.allowDelete)
     return tool.credential_id ? 'api-key' as const : 'unauthorized' as const
 
@@ -102,26 +105,10 @@ export function useAgentToolsOperations() {
   const [tools, setTools] = useAtom(agentComposerToolsAtom)
   const removeProviderTool = useRemoveProviderTool()
   const removeProviderToolAction = useRemoveProviderToolAction()
-  const [expandedToolIds, setExpandedToolIds] = useState<Set<string>>(() => new Set())
   const [settingTarget, setSettingTarget] = useState<ToolSettingTarget | null>(null)
   const [isCliToolDialogOpen, setIsCliToolDialogOpen] = useState(false)
   const [editingCliTool, setEditingCliTool] = useState<AgentCliTool | null>(null)
 
-  const setToolOpen = useCallback((tool: AgentTool, open: boolean) => {
-    if (tool.kind === 'cli')
-      return
-
-    setExpandedToolIds((currentIds) => {
-      const nextIds = new Set(currentIds)
-      if (open)
-        nextIds.add(tool.id)
-      else
-        nextIds.delete(tool.id)
-
-      return nextIds
-    })
-  }, [])
-
   const addTools = useCallback((selectedTools: AgentProviderToolDefaultValue[]) => {
     setTools(addProviderTools(tools, selectedTools))
   }, [setTools, tools])
@@ -168,11 +155,6 @@ export function useAgentToolsOperations() {
   }, [])
 
   const deleteProviderTool = useCallback((toolId: string) => {
-    setExpandedToolIds((currentIds) => {
-      const nextIds = new Set(currentIds)
-      nextIds.delete(toolId)
-      return nextIds
-    })
     closeSettingTargetIfRemoved(toolId)
     removeProviderTool(toolId)
   }, [closeSettingTargetIfRemoved, removeProviderTool])
@@ -191,12 +173,10 @@ export function useAgentToolsOperations() {
   return {
     tools,
     selectedTools,
-    expandedToolIds,
     settingTarget,
     isCliToolDialogOpen,
     editingCliTool,
     setTools,
-    setToolOpen,
     setSettingTarget,
     addTools,
     deleteCliTool,
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/index.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/index.tsx
index 20cbb247888..f17a931b2f8 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/index.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/index.tsx
@@ -7,8 +7,9 @@ import type { ToolWithProvider } from '@/app/components/workflow/types'
 import type { AgentCliTool, AgentProviderTool, AgentTool } from '@/features/agent-v2/agent-composer/form-state'
 import { cn } from '@langgenius/dify-ui/cn'
 import { Popover, PopoverContent, PopoverTrigger } from '@langgenius/dify-ui/popover'
-import { useCallback, useMemo, useRef, useState } from 'react'
+import { memo, useCallback, useMemo, useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
+import { CollectionType } from '@/app/components/tools/types'
 import { ToolPickerContent } from '@/app/components/workflow/block-selector/tool-picker'
 import { useGetLanguage } from '@/context/i18n'
 import { useSetProviderToolCredential } from '@/features/agent-v2/agent-composer/store-modules/tools'
@@ -31,10 +32,8 @@ import { useAgentToolsOperations } from './hooks'
 import { ProviderToolSettingsDialog } from './provider-tool/dialog'
 import { AgentProviderToolItem } from './provider-tool/item'
 
-function AgentToolItem({
+const AgentToolItem = memo(({
   tool,
-  isExpanded,
-  onOpenChange,
   onConfigureAction,
   onDeleteCliTool,
   onDeleteProviderTool,
@@ -43,18 +42,14 @@ function AgentToolItem({
   onCredentialChange,
 }: {
   tool: AgentTool
-  isExpanded: boolean
-  onOpenChange: (tool: AgentTool, open: boolean) => void
   onConfigureAction: (target: ToolSettingTarget) => void
   onDeleteCliTool: (toolId: string) => void
   onDeleteProviderTool: (toolId: string) => void
   onDeleteProviderToolAction: (toolId: string, actionId: string) => void
   onEditCliTool: (tool: AgentCliTool) => void
-  onCredentialChange: (toolId: string, credentialId?: string) => void
-}) {
-  const handleOpenChange = useCallback((open: boolean) => {
-    onOpenChange(tool, open)
-  }, [onOpenChange, tool])
+  onCredentialChange: (toolId: string, credentialId?: string, credentialType?: AgentProviderTool['credentialType']) => void
+}) => {
+  const [isExpanded, setIsExpanded] = useState(false)
 
   const handleRemoveProvider = useCallback(() => {
     onDeleteProviderTool(tool.id)
@@ -73,8 +68,8 @@ function AgentToolItem({
       onEditCliTool(tool)
   }, [onEditCliTool, tool])
 
-  const handleCredentialChange = useCallback((credentialId?: string) => {
-    onCredentialChange(tool.id, credentialId)
+  const handleCredentialChange = useCallback((credentialId?: string, credentialType?: AgentProviderTool['credentialType']) => {
+    onCredentialChange(tool.id, credentialId, credentialType)
   }, [onCredentialChange, tool.id])
 
   if (tool.kind === 'provider') {
@@ -82,7 +77,7 @@ function AgentToolItem({
       <AgentProviderToolItem
         tool={tool}
         isExpanded={isExpanded}
-        onOpenChange={handleOpenChange}
+        onOpenChange={setIsExpanded}
         onConfigureAction={onConfigureAction}
         onRemoveAction={handleRemoveProviderAction}
         onRemoveProvider={handleRemoveProvider}
@@ -98,7 +93,7 @@ function AgentToolItem({
       onEdit={handleEditCliTool}
     />
   )
-}
+})
 
 function useAgentToolProviderMap() {
   const { data: buildInTools } = useAllBuiltInTools()
@@ -139,8 +134,44 @@ function getLocalizedText(
   return text?.[language] ?? text?.en_US ?? text?.zh_Hans
 }
 
-function getProviderCredentialRequired(provider?: ToolWithProvider) {
-  return Object.keys(provider?.team_credentials ?? {}).length > 0
+function getProviderCredentialType(provider?: ToolWithProvider): AgentProviderTool['credentialType'] {
+  if (!provider)
+    return undefined
+
+  if (Object.keys(provider.team_credentials ?? {}).length > 0)
+    return 'api-key'
+
+  if (provider.type === CollectionType.builtIn && provider.allow_delete)
+    return 'oauth2'
+
+  return undefined
+}
+
+function getDisplayCredentialType(
+  tool: AgentProviderTool,
+  providerCredentialType: AgentProviderTool['credentialType'],
+) {
+  if (!providerCredentialType)
+    return undefined
+
+  if (providerCredentialType === 'oauth2' && tool.credentialType === 'unauthorized')
+    return 'oauth2' as const
+
+  return tool.credentialType ?? providerCredentialType
+}
+
+function getProviderCredentialVariant(
+  tool: AgentProviderTool,
+  provider: ToolWithProvider,
+  providerCredentialType: AgentProviderTool['credentialType'],
+) {
+  if (!providerCredentialType)
+    return 'none' as const
+
+  if (tool.credentialVariant !== 'none')
+    return tool.credentialVariant
+
+  return tool.credentialId || provider.is_team_authorization ? 'authorized' as const : 'unauthorized' as const
 }
 
 function useDisplayTools(
@@ -161,7 +192,7 @@ function useDisplayTools(
         return tool
 
       const providerToolByName = new Map(provider.tools.map(providerTool => [providerTool.name, providerTool]))
-      const credentialRequired = getProviderCredentialRequired(provider)
+      const providerCredentialType = getProviderCredentialType(provider)
 
       return {
         ...tool,
@@ -170,9 +201,11 @@ function useDisplayTools(
         iconDark: tool.iconDark ?? provider.icon_dark,
         providerType: tool.providerType ?? provider.type,
         allowDelete: tool.allowDelete ?? provider.allow_delete,
-        credentialKey: credentialRequired ? tool.credentialKey : undefined,
-        credentialType: credentialRequired ? tool.credentialType : undefined,
-        credentialVariant: credentialRequired ? tool.credentialVariant : 'none',
+        credentialKey: providerCredentialType
+          ? tool.credentialKey ?? 'agentDetail.configure.tools.credential.authOne'
+          : undefined,
+        credentialType: getDisplayCredentialType(tool, providerCredentialType),
+        credentialVariant: getProviderCredentialVariant(tool, provider, providerCredentialType),
         actions: tool.actions.map((action) => {
           const providerTool = providerToolByName.get(action.toolName)
 
@@ -266,7 +299,10 @@ function AddToolMenu({
     allowDelete: (providerById.get(tool.provider_id)
       ?? providerById.get(tool.provider_name)
       ?? (tool.plugin_id ? providerById.get(tool.plugin_id) : undefined))?.allow_delete,
-    credentialRequired: getProviderCredentialRequired(providerById.get(tool.provider_id)
+    credentialType: getProviderCredentialType(providerById.get(tool.provider_id)
+      ?? providerById.get(tool.provider_name)
+      ?? (tool.plugin_id ? providerById.get(tool.plugin_id) : undefined)),
+    credentialRequired: !!getProviderCredentialType(providerById.get(tool.provider_id)
       ?? providerById.get(tool.provider_name)
       ?? (tool.plugin_id ? providerById.get(tool.plugin_id) : undefined)),
   }), [providerById])
@@ -316,7 +352,7 @@ function AddToolMenu({
           : (
               <ToolPickerContent
                 focusSearchOnMount
-                panelClassName="w-full"
+                panelClassName="w-full overflow-hidden"
                 supportAddCustomTool
                 selectedTools={selectedTools}
                 onSelect={handleSelectTool}
@@ -336,11 +372,9 @@ export function AgentTools() {
   const {
     tools,
     selectedTools,
-    expandedToolIds,
     settingTarget,
     isCliToolDialogOpen,
     editingCliTool,
-    setToolOpen,
     setSettingTarget,
     addTools,
     deleteCliTool,
@@ -462,8 +496,6 @@ export function AgentTools() {
               <AgentToolItem
                 key={tool.id}
                 tool={tool}
-                isExpanded={tool.kind === 'provider' && expandedToolIds.has(tool.id)}
-                onOpenChange={setToolOpen}
                 onConfigureAction={setSettingTarget}
                 onDeleteCliTool={deleteCliTool}
                 onDeleteProviderTool={deleteProviderTool}
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/provider-tool/item.tsx b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/provider-tool/item.tsx
index 8c15189c56b..64a2321a5d4 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/provider-tool/item.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/provider-tool/item.tsx
@@ -16,11 +16,12 @@ import {
   DropdownMenuTrigger,
 } from '@langgenius/dify-ui/dropdown-menu'
 import { StatusDot } from '@langgenius/dify-ui/status-dot'
-import { useCallback, useMemo, useState } from 'react'
+import { memo, useCallback, useMemo, useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import {
   AuthCategory,
 } from '@/app/components/plugins/plugin-auth'
+import AddOAuthButton from '@/app/components/plugins/plugin-auth/authorize/add-oauth-button'
 import ApiKeyModal from '@/app/components/plugins/plugin-auth/authorize/api-key-modal'
 import AuthorizedInNode from '@/app/components/plugins/plugin-auth/authorized-in-node'
 import { useInvalidPluginCredentialInfoHook } from '@/app/components/plugins/plugin-auth/hooks/use-credential'
@@ -60,7 +61,7 @@ function UnauthorizedCredentialStatus({
   onCredentialChange,
 }: {
   tool: AgentProviderTool
-  onCredentialChange: (credentialId?: string) => void
+  onCredentialChange: (credentialId?: string, credentialType?: AgentProviderTool['credentialType']) => void
 }) {
   const { t } = useTranslation()
   const [isApiKeyModalOpen, setIsApiKeyModalOpen] = useState(false)
@@ -78,8 +79,29 @@ function UnauthorizedCredentialStatus({
   }, [])
   const handleCredentialUpdate = useCallback(() => {
     invalidPluginCredentialInfo()
-    onCredentialChange()
-  }, [invalidPluginCredentialInfo, onCredentialChange])
+    onCredentialChange(undefined, tool.credentialType)
+  }, [invalidPluginCredentialInfo, onCredentialChange, tool.credentialType])
+
+  if (tool.credentialType === 'oauth2') {
+    return (
+      <AddOAuthButton
+        pluginPayload={pluginPayload}
+        onUpdate={handleCredentialUpdate}
+        renderTrigger={({ disabled, onClick }) => (
+          <Button
+            variant="secondary"
+            size="small"
+            className="shrink-0"
+            disabled={disabled}
+            onClick={onClick}
+          >
+            {t('notAuthorized', { ns: 'tools' })}
+            <StatusDot className="ml-2" status="warning" />
+          </Button>
+        )}
+      />
+    )
+  }
 
   return (
     <>
@@ -108,16 +130,16 @@ function CredentialStatus({
   onCredentialChange,
 }: {
   tool: AgentProviderTool
-  onCredentialChange: (credentialId?: string) => void
+  onCredentialChange: (credentialId?: string, credentialType?: AgentProviderTool['credentialType']) => void
 }) {
   const canSwitchCredential = (tool.providerType ?? CollectionType.builtIn) === CollectionType.builtIn && tool.allowDelete
   const handleAuthorizationItemClick = useCallback((id: string) => {
-    onCredentialChange(id === '__workspace_default__' ? undefined : id || undefined)
-  }, [onCredentialChange])
+    onCredentialChange(id === '__workspace_default__' ? undefined : id || undefined, tool.credentialType)
+  }, [onCredentialChange, tool.credentialType])
   const handleDefaultCredentialChange = useCallback((id?: string) => {
     if (!tool.credentialId && id)
-      onCredentialChange(id)
-  }, [onCredentialChange, tool.credentialId])
+      onCredentialChange(id, tool.credentialType)
+  }, [onCredentialChange, tool.credentialId, tool.credentialType])
 
   if (tool.credentialVariant === 'none')
     return null
@@ -150,7 +172,7 @@ function CredentialStatus({
   )
 }
 
-function ProviderToolActionItem({
+const ProviderToolActionItem = memo(({
   action,
   tool,
   onConfigureAction,
@@ -160,7 +182,7 @@ function ProviderToolActionItem({
   tool: AgentProviderTool
   onConfigureAction: (target: ToolSettingTarget) => void
   onRemoveAction: (actionId: string) => void
-}) {
+}) => {
   const { t } = useTranslation('agentV2')
   const readOnly = useAgentOrchestrateReadOnly()
   const handleConfigureAction = useCallback(() => {
@@ -200,9 +222,9 @@ function ProviderToolActionItem({
       )}
     </div>
   )
-}
+})
 
-export function AgentProviderToolItem({
+export const AgentProviderToolItem = memo(({
   tool,
   isExpanded,
   onOpenChange,
@@ -217,8 +239,8 @@ export function AgentProviderToolItem({
   onConfigureAction: (target: ToolSettingTarget) => void
   onRemoveAction: (actionId: string) => void
   onRemoveProvider: () => void
-  onCredentialChange: (credentialId?: string) => void
-}) {
+  onCredentialChange: (credentialId?: string, credentialType?: AgentProviderTool['credentialType']) => void
+}) => {
   const { t } = useTranslation('agentV2')
   const readOnly = useAgentOrchestrateReadOnly()
   const { theme } = useTheme()
@@ -275,18 +297,20 @@ export function AgentProviderToolItem({
       </div>
 
       <CollapsiblePanel>
-        <div className="flex flex-col">
-          {tool.actions.map(action => (
-            <ProviderToolActionItem
-              key={action.id}
-              action={action}
-              tool={tool}
-              onConfigureAction={onConfigureAction}
-              onRemoveAction={onRemoveAction}
-            />
-          ))}
-        </div>
+        {isExpanded && (
+          <div className="flex flex-col">
+            {tool.actions.map(action => (
+              <ProviderToolActionItem
+                key={action.id}
+                action={action}
+                tool={tool}
+                onConfigureAction={onConfigureAction}
+                onRemoveAction={onRemoveAction}
+              />
+            ))}
+          </div>
+        )}
       </CollapsiblePanel>
     </CollapsibleRoot>
   )
-}
+})
diff --git a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/types.ts b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/types.ts
index 4401994fa37..e977496bcf1 100644
--- a/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/types.ts
+++ b/web/features/agent-v2/agent-detail/configure/components/orchestrate/tools/types.ts
@@ -6,6 +6,7 @@ import type {
 
 export type AgentProviderToolDefaultValue = ToolDefaultValue & {
   allowDelete?: boolean
+  credentialType?: AgentProviderTool['credentialType']
   credentialRequired?: boolean
 }
 
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/chat.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/chat.spec.tsx
index c626e2fc813..f6fbcb61e31 100644
--- a/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/chat.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/chat.spec.tsx
@@ -1,4 +1,4 @@
-import type { ComponentProps } from 'react'
+import type { ComponentProps, ReactNode } from 'react'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
 import { act, fireEvent, render, screen, waitFor } from '@testing-library/react'
 import { createStore, Provider as JotaiProvider } from 'jotai'
@@ -29,11 +29,23 @@ vi.mock('@/next/dynamic', async () => {
     default: () => function MockChat(props: {
       onSend: (message: string) => unknown
       onStopResponding: () => void
+      sendButtonLabel?: string
+      sendButtonLoading?: boolean
+      showPromptLog?: boolean
+      footerNotice?: string
+      chatNode?: ReactNode
     }) {
       const [sent, setSent] = useState(false)
 
       return (
-        <div>
+        <div
+          data-testid="mock-chat"
+          data-send-button-label={props.sendButtonLabel ?? ''}
+          data-send-button-loading={String(!!props.sendButtonLoading)}
+          data-show-prompt-log={String(!!props.showPromptLog)}
+          data-footer-notice={props.footerNotice ?? ''}
+        >
+          {props.chatNode}
           <span>{`sessionSent:${sent ? 'yes' : 'no'}`}</span>
           <button
             type="button"
@@ -53,6 +65,14 @@ vi.mock('@/next/dynamic', async () => {
   }
 })
 
+vi.mock('@/app/components/base/chat/chat/chat-input-area', () => ({
+  default: ({ footerNotice }: { footerNotice?: ReactNode }) => (
+    <div data-testid="agent-preview-chat-input">
+      {footerNotice}
+    </div>
+  ),
+}))
+
 vi.mock('@/app/components/base/chat/chat/hooks', () => ({
   useChat: useChatMock.mockImplementation((
     _config: unknown,
@@ -387,6 +407,142 @@ describe('AgentPreviewChat', () => {
     })
   })
 
+  it('should map agent SSE error events to chat send errors', async () => {
+    renderPreviewChat()
+
+    fireEvent.click(screen.getByRole('button', { name: 'send' }))
+
+    await waitFor(() => expect(handleSendMock).toHaveBeenCalledTimes(1))
+    const callbacks = handleSendMock.mock.calls.at(0)?.[2]
+
+    expect(callbacks.onUnhandledEvent({
+      event: 'error',
+      conversation_id: 'conversation-1',
+      message_id: 'message-1',
+      code: 'agent_run_failed',
+      message: 'Agent execution failed',
+    })).toEqual({
+      conversationId: 'conversation-1',
+      messageId: 'message-1',
+      errorCode: 'agent_run_failed',
+      errorMessage: 'Agent execution failed',
+    })
+    expect(callbacks.onUnhandledEvent({
+      event: 'unknown',
+      message: 'Ignored',
+    })).toBeUndefined()
+  })
+
+  it('should show the send button loading state while preparing a build run', async () => {
+    let resolveSaveDraftBeforeRun: () => void = () => {}
+    const saveDraftBeforeRun = vi.fn(() => new Promise<void>((resolve) => {
+      resolveSaveDraftBeforeRun = resolve
+    }))
+    renderPreviewChat({
+      sendButtonLabel: 'Start build',
+      renderEmptyState: ({ inputNode }) => <div>{inputNode}</div>,
+      onSaveDraftBeforeRun: saveDraftBeforeRun,
+    })
+
+    fireEvent.click(screen.getByRole('button', { name: 'send' }))
+
+    expect(saveDraftBeforeRun).toHaveBeenCalledTimes(1)
+    await waitFor(() => {
+      expect(screen.getByTestId('mock-chat')).toHaveAttribute('data-send-button-loading', 'true')
+    })
+    expect(handleSendMock).not.toHaveBeenCalled()
+
+    await act(async () => {
+      resolveSaveDraftBeforeRun()
+    })
+    await waitFor(() => expect(handleSendMock).toHaveBeenCalledTimes(1))
+  })
+
+  it('should not show the send button loading state while an icon send button prepares the run', async () => {
+    const saveDraftBeforeRun = vi.fn(() => new Promise<void>(() => {}))
+    renderPreviewChat({
+      onSaveDraftBeforeRun: saveDraftBeforeRun,
+    })
+
+    await waitFor(() => expect(screen.getByTestId('mock-chat')).toBeInTheDocument())
+
+    fireEvent.click(screen.getByRole('button', { name: 'send' }))
+
+    expect(saveDraftBeforeRun).toHaveBeenCalledTimes(1)
+    await waitFor(() => {
+      expect(screen.getByTestId('mock-chat')).toHaveAttribute('data-send-button-loading', 'false')
+    })
+    expect(handleSendMock).not.toHaveBeenCalled()
+  })
+
+  it('should not show the send button loading state while an icon send button is responding', async () => {
+    useChatMock.mockImplementationOnce((
+      _config: unknown,
+      _formSettings: unknown,
+      chatList: unknown[],
+      stopCallback: (taskId: string) => void,
+    ) => {
+      stopCallbackRef.current = stopCallback
+
+      return {
+        chatList,
+        setTargetMessageId: vi.fn(),
+        isResponding: true,
+        handleSend: handleSendMock,
+        suggestedQuestions: [],
+        handleStop: () => stopCallback('task-1'),
+        handleAnnotationAdded: vi.fn(),
+        handleAnnotationEdited: vi.fn(),
+        handleAnnotationRemoved: vi.fn(),
+      }
+    })
+
+    renderPreviewChat()
+
+    await waitFor(() => expect(screen.getByTestId('mock-chat')).toBeInTheDocument())
+
+    expect(screen.getByTestId('mock-chat')).toHaveAttribute('data-send-button-loading', 'false')
+  })
+
+  it('should use the default send button after the first build message', async () => {
+    useChatMock.mockImplementationOnce((
+      _config: unknown,
+      _formSettings: unknown,
+      _chatList: unknown[],
+      stopCallback: (taskId: string) => void,
+    ) => {
+      stopCallbackRef.current = stopCallback
+
+      return {
+        chatList: [
+          {
+            id: 'question-1',
+            content: 'Build an agent',
+            isAnswer: false,
+          },
+          {
+            id: 'answer-1',
+            content: 'Done',
+            isAnswer: true,
+          },
+        ],
+        setTargetMessageId: vi.fn(),
+        isResponding: false,
+        handleSend: handleSendMock,
+        suggestedQuestions: [],
+        handleStop: () => stopCallback('task-1'),
+        handleAnnotationAdded: vi.fn(),
+        handleAnnotationEdited: vi.fn(),
+        handleAnnotationRemoved: vi.fn(),
+      }
+    })
+    renderPreviewChat({
+      sendButtonLabel: 'Start build',
+    })
+
+    expect(screen.getByTestId('mock-chat')).toHaveAttribute('data-send-button-label', '')
+  })
+
   it('should sync the completed conversation history into the query cache', async () => {
     const conversationMessagesResponse = {
       data: [
@@ -514,6 +670,30 @@ describe('AgentPreviewChat', () => {
     )
   })
 
+  it('should keep prompt log action hidden in the build chat runtime', async () => {
+    renderPreviewChat({
+      draftType: 'debug_build',
+    })
+
+    await waitFor(() => expect(screen.getByTestId('mock-chat')).toBeInTheDocument())
+
+    expect(screen.getByTestId('mock-chat')).toHaveAttribute('data-show-prompt-log', 'false')
+  })
+
+  it('should hide the sandbox notice after the first send starts', async () => {
+    renderPreviewChat({
+      renderEmptyState: ({ inputNode }) => <div>{inputNode}</div>,
+    })
+
+    expect(screen.getByText('agentV2.agentDetail.configure.preview.sandboxNotice')).toBeInTheDocument()
+
+    fireEvent.click(screen.getByRole('button', { name: 'send' }))
+
+    await waitFor(() => {
+      expect(screen.queryByText('agentV2.agentDetail.configure.preview.sandboxNotice')).not.toBeInTheDocument()
+    })
+  })
+
   it('should send build chat inputs from the prepared build draft snapshot', async () => {
     const saveDraftBeforeRun = vi.fn().mockResolvedValue({
       app_variables: [
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/header.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/header.spec.tsx
index 4bb7ca8a226..cdecdc3d8ff 100644
--- a/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/header.spec.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/header.spec.tsx
@@ -10,6 +10,7 @@ function renderHeader({
   onOpenWorkingDirectory = vi.fn(),
   onRefresh = vi.fn(),
   refreshDisabled = false,
+  showWorkingDirectoryAction = false,
 }: {
   mode?: 'build' | 'preview'
   previewEnabled?: boolean
@@ -18,6 +19,7 @@ function renderHeader({
   onOpenWorkingDirectory?: () => void
   onRefresh?: () => void
   refreshDisabled?: boolean
+  showWorkingDirectoryAction?: boolean
 } = {}) {
   render(
     <AgentPreviewHeader
@@ -29,6 +31,7 @@ function renderHeader({
       onOpenWorkingDirectory={onOpenWorkingDirectory}
       onRefresh={onRefresh}
       refreshDisabled={refreshDisabled}
+      showWorkingDirectoryAction={showWorkingDirectoryAction}
     />,
   )
 }
@@ -71,13 +74,22 @@ describe('AgentPreviewHeader', () => {
   it('should open the working directory from the build header', async () => {
     const user = userEvent.setup()
     const onOpenWorkingDirectory = vi.fn()
-    renderHeader({ mode: 'build', onOpenWorkingDirectory })
+    renderHeader({ mode: 'build', onOpenWorkingDirectory, showWorkingDirectoryAction: true })
 
-    await user.click(screen.getByRole('button', { name: 'agentV2.agentDetail.configure.workingDirectory.open' }))
+    const fileSystemButton = screen.getByRole('button', { name: 'agentV2.agentDetail.configure.workingDirectory.open' })
+    expect(fileSystemButton).toHaveTextContent('agentV2.agentDetail.configure.workingDirectory.fileSystem')
+
+    await user.click(fileSystemButton)
 
     expect(onOpenWorkingDirectory).toHaveBeenCalledTimes(1)
   })
 
+  it('should hide the working directory action when unavailable', () => {
+    renderHeader({ mode: 'build' })
+
+    expect(screen.queryByRole('button', { name: 'agentV2.agentDetail.configure.workingDirectory.open' })).not.toBeInTheDocument()
+  })
+
   it('should disable preview mode when preview is unavailable', async () => {
     const user = userEvent.setup()
     const onModeChange = vi.fn()
@@ -89,8 +101,20 @@ describe('AgentPreviewHeader', () => {
 
     const modeControl = screen.getByRole('group', { name: 'agentV2.agentDetail.configure.rightPanel.modeLabel' })
 
-    await user.click(within(modeControl).getByRole('button', { name: /agentV2\.agentDetail\.configure\.rightPanel\.preview/ }))
+    await user.click(within(modeControl).getByRole('button', { name: 'agentV2.agentDetail.configure.rightPanel.preview' }))
 
     expect(onModeChange).not.toHaveBeenCalled()
   })
+
+  it('should explain disabled preview mode on hover', async () => {
+    const user = userEvent.setup()
+    renderHeader({
+      mode: 'build',
+      previewEnabled: false,
+    })
+
+    await user.hover(screen.getByLabelText('agentV2.agentDetail.configure.rightPanel.previewDisabledTip'))
+
+    expect(await screen.findByText('agentV2.agentDetail.configure.rightPanel.previewDisabledTip')).toBeInTheDocument()
+  })
 })
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/working-directory-breadcrumb.spec.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/working-directory-breadcrumb.spec.tsx
new file mode 100644
index 00000000000..f7332f3dd3f
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/__tests__/working-directory-breadcrumb.spec.tsx
@@ -0,0 +1,129 @@
+import { render, screen } from '@testing-library/react'
+import userEvent from '@testing-library/user-event'
+import { AgentWorkingDirectoryBreadcrumb } from '../working-directory-breadcrumb'
+
+describe('AgentWorkingDirectoryBreadcrumb', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+  })
+
+  describe('Rendering', () => {
+    it('should render the root path by default', () => {
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="."
+          onPathChange={vi.fn()}
+        />,
+      )
+
+      expect(screen.getByRole('navigation', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.breadcrumbLabel',
+      })).toBeInTheDocument()
+      expect(screen.getByRole('button', { name: '.' })).toHaveAttribute('aria-current', 'page')
+    })
+
+    it('should render home as the current path when path is home', () => {
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="~"
+          onPathChange={vi.fn()}
+        />,
+      )
+
+      expect(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.home',
+      })).toHaveAttribute('aria-current', 'page')
+      expect(screen.queryByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.workingDirectory',
+      })).not.toBeInTheDocument()
+    })
+
+    it('should render the workspace cwd directly and show tilde as home', () => {
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="~/web-game"
+          onPathChange={vi.fn()}
+        />,
+      )
+
+      expect(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.home',
+      })).toBeInTheDocument()
+      expect(screen.getByRole('button', {
+        name: 'web-game',
+      })).toHaveAttribute('aria-current', 'page')
+    })
+
+    it('should collapse middle breadcrumb layers when path is deeper than three layers', () => {
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="~/web-game/src/app"
+          onPathChange={vi.fn()}
+        />,
+      )
+
+      expect(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.home',
+      })).toBeInTheDocument()
+      expect(screen.getByRole('button', { name: '...' })).toBeInTheDocument()
+      expect(screen.queryByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.workingDirectory',
+      })).not.toBeInTheDocument()
+      expect(screen.queryByRole('button', { name: 'web-game' })).not.toBeInTheDocument()
+      expect(screen.getByRole('button', { name: 'src' })).toBeInTheDocument()
+      expect(screen.getByRole('button', { name: 'app' })).toHaveAttribute('aria-current', 'page')
+    })
+  })
+
+  describe('User Interactions', () => {
+    it('should request home path when home is clicked', async () => {
+      const user = userEvent.setup()
+      const handlePathChange = vi.fn()
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="~/web-game"
+          onPathChange={handlePathChange}
+        />,
+      )
+
+      await user.click(screen.getByRole('button', {
+        name: 'agentV2.agentDetail.configure.workingDirectory.home',
+      }))
+
+      expect(handlePathChange).toHaveBeenCalledWith('~')
+    })
+
+    it('should request the selected path segment when a breadcrumb item is clicked', async () => {
+      const user = userEvent.setup()
+      const handlePathChange = vi.fn()
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="~/web-game/src"
+          onPathChange={handlePathChange}
+        />,
+      )
+
+      await user.click(screen.getByRole('button', { name: 'web-game' }))
+
+      expect(handlePathChange).toHaveBeenCalledWith('~/web-game')
+    })
+
+    it('should request a hidden breadcrumb path from the ellipsis menu', async () => {
+      const user = userEvent.setup()
+      const handlePathChange = vi.fn()
+      render(
+        <AgentWorkingDirectoryBreadcrumb
+          path="~/web-game/src/app"
+          onPathChange={handlePathChange}
+        />,
+      )
+
+      await user.click(screen.getByRole('button', { name: '...' }))
+      await user.click(screen.getByRole('menuitem', {
+        name: 'web-game',
+      }))
+
+      expect(handlePathChange).toHaveBeenCalledWith('~/web-game')
+    })
+  })
+})
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/chat-runtime.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/chat-runtime.tsx
index 7fe67be6972..8d919286e4d 100644
--- a/web/features/agent-v2/agent-detail/configure/components/preview/chat-runtime.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/chat-runtime.tsx
@@ -24,6 +24,7 @@ import { cn } from '@langgenius/dify-ui/cn'
 import { skipToken, useQuery, useQueryClient } from '@tanstack/react-query'
 import { useAtomValue } from 'jotai'
 import { useCallback, useEffect, useMemo, useRef, useState } from 'react'
+import { useTranslation } from 'react-i18next'
 import ChatInputArea from '@/app/components/base/chat/chat/chat-input-area'
 import { useChat } from '@/app/components/base/chat/chat/hooks'
 import { buildChatItemTree, getLastAnswer, isValidGeneratedAnswer } from '@/app/components/base/chat/utils'
@@ -586,6 +587,7 @@ function AgentPreviewChatSession({
   onSaveDraftBeforeRun?: () => Promise<AgentSoulConfig | void>
   onSendInterrupted?: () => void
 }) {
+  const { t } = useTranslation('agentV2')
   const queryClient = useQueryClient()
   const { userProfile } = useAppContext()
   const prompt = useAtomValue(agentComposerPromptAtom)
@@ -598,6 +600,7 @@ function AgentPreviewChatSession({
   const inputsForm = useMemo(() => getAgentSoulInputsForm(agentSoulConfig), [agentSoulConfig])
   const inputs = useMemo(() => getAgentSoulInputs(inputsForm), [inputsForm])
   const sendInterruptedRef = useRef(false)
+  const [isSendPending, setIsSendPending] = useState(false)
   const notifySendInterrupted = useCallback(() => {
     if (sendInterruptedRef.current)
       return
@@ -635,6 +638,8 @@ function AgentPreviewChatSession({
 
   const doSend: OnSend = useCallback(async (message, files, isRegenerate = false, parentAnswer: ChatItem | null = null) => {
     sendInterruptedRef.current = false
+    setIsSendPending(true)
+    let sendStarted = false
 
     try {
       const preparedAgentSoulConfig = await onSaveDraftBeforeRun?.()
@@ -684,6 +689,17 @@ function AgentPreviewChatSession({
             })
           },
           onGetSuggestedQuestions: responseItemId => fetchAgentSuggestedQuestions(agentId, responseItemId),
+          onUnhandledEvent: (event) => {
+            if (event.event !== 'error' || typeof event.message !== 'string')
+              return
+
+            return {
+              conversationId: typeof event.conversation_id === 'string' ? event.conversation_id : undefined,
+              messageId: typeof event.message_id === 'string' ? event.message_id : undefined,
+              errorMessage: event.message,
+              errorCode: typeof event.code === 'string' ? event.code : undefined,
+            }
+          },
           onConversationComplete: (completedConversationId, workflowRunId) => {
             if (completedConversationId && completedConversationId !== conversationId)
               onCurrentSessionConversationIdChange(completedConversationId)
@@ -691,15 +707,21 @@ function AgentPreviewChatSession({
             onConversationComplete?.(completedConversationId, workflowRunId)
           },
           onSendSettled: (hasError) => {
+            setIsSendPending(false)
             if (hasError)
               notifySendInterrupted()
           },
         },
       )
+      sendStarted = true
     }
     catch {
       return false
     }
+    finally {
+      if (!sendStarted)
+        setIsSendPending(false)
+    }
   }, [agentId, agentSoulConfig, chatList, config, conversationId, draftType, handleSend, inputs, inputsForm, notifySendInterrupted, onConversationComplete, onConversationIdChange, onCurrentSessionConversationIdChange, onSaveDraftBeforeRun, queryClient, textGenerationModelList])
 
   const doStopResponding = useCallback(() => {
@@ -722,12 +744,17 @@ function AgentPreviewChatSession({
   }, [chatList, doSend])
   const isEmptyChat = chatList.length === 0
   const hasInstructions = !!config.pre_prompt.trim()
+  const sendButtonLoading = isEmptyChat && !!sendButtonLabel && (isSendPending || isResponding)
+  const sandboxNotice = t('agentDetail.configure.preview.sandboxNotice')
+  const sandboxNoticeTooltip = t('agentDetail.configure.preview.sandboxNoticeTooltip')
+  const showSandboxNotice = isEmptyChat && !isSendPending && !isResponding
   const emptyChatInputNode = (
     <div className="pointer-events-auto mt-5 w-full">
       <ChatInputArea
         botName={agentName || 'Agent'}
         customPlaceholder={inputPlaceholder}
         disabled={isResponding}
+        sendButtonLoading={sendButtonLoading}
         showFileUpload={false}
         visionConfig={config.file_upload}
         speechToTextConfig={config.speech_to_text}
@@ -735,6 +762,8 @@ function AgentPreviewChatSession({
         inputs={inputs}
         inputsForm={inputsForm}
         sendButtonLabel={sendButtonLabel}
+        footerNotice={showSandboxNotice ? sandboxNotice : undefined}
+        footerNoticeTooltip={showSandboxNotice ? sandboxNoticeTooltip : undefined}
       />
     </div>
   )
@@ -744,6 +773,7 @@ function AgentPreviewChatSession({
       config={config}
       chatList={chatList}
       isResponding={isResponding}
+      sendButtonLoading={sendButtonLoading}
       chatNode={isEmptyChat
         ? renderEmptyState({
             agentIcon,
@@ -760,7 +790,7 @@ function AgentPreviewChatSession({
         isEmptyChat ? 'hidden' : 'px-3 pt-10',
       )}
       inputPlaceholder={inputPlaceholder}
-      sendButtonLabel={sendButtonLabel}
+      sendButtonLabel={isEmptyChat ? sendButtonLabel : undefined}
       showFileUpload={false}
       suggestedQuestions={suggestedQuestions}
       onSend={doSend}
@@ -770,7 +800,6 @@ function AgentPreviewChatSession({
       switchSibling={siblingMessageId => setTargetMessageId(siblingMessageId)}
       onStopResponding={doStopResponding}
       noChatInput={isEmptyChat}
-      showPromptLog
       questionIcon={<Avatar avatar={userProfile.avatar_url} name={userProfile.name} size="xl" />}
       onAnnotationEdited={handleAnnotationEdited}
       onAnnotationAdded={handleAnnotationAdded}
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/header.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/header.tsx
index c4f9e954f0b..baaf14c7115 100644
--- a/web/features/agent-v2/agent-detail/configure/components/preview/header.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/header.tsx
@@ -1,8 +1,8 @@
-import type { ReactNode } from 'react'
+import type { MouseEvent, ReactNode } from 'react'
 import { cn } from '@langgenius/dify-ui/cn'
+import { Popover, PopoverContent, PopoverTrigger } from '@langgenius/dify-ui/popover'
 import { SegmentedControl, SegmentedControlDivider, SegmentedControlItem } from '@langgenius/dify-ui/segmented-control'
 import { useTranslation } from 'react-i18next'
-import { Infotip } from '@/app/components/base/infotip'
 import { useDocLink } from '@/context/i18n'
 
 type AgentConfigureRightPanelMode = 'build' | 'preview'
@@ -14,7 +14,7 @@ function AgentModeTipSection({
 }: {
   iconClassName: string
   title: string
-  children: string
+  children: ReactNode
 }) {
   return (
     <div className="flex flex-col gap-1">
@@ -27,6 +27,83 @@ function AgentModeTipSection({
   )
 }
 
+function ModeInfoTip({
+  children,
+  ariaLabel,
+}: {
+  children: ReactNode
+  ariaLabel: string
+}) {
+  const handleClick = (event: MouseEvent<HTMLButtonElement>) => {
+    event.stopPropagation()
+  }
+
+  return (
+    <Popover>
+      <PopoverTrigger
+        openOnHover
+        delay={300}
+        closeDelay={200}
+        aria-label={ariaLabel}
+        onClick={handleClick}
+        className="inline-flex size-5 shrink-0 cursor-pointer items-center justify-center rounded-md border-0 bg-transparent p-0 outline-hidden focus-visible:ring-2 focus-visible:ring-state-accent-solid"
+      >
+        <span aria-hidden className="i-ri-question-line size-4 text-text-tertiary hover:text-text-secondary" />
+      </PopoverTrigger>
+      <PopoverContent
+        placement="bottom"
+        sideOffset={2}
+        popupClassName="w-60 max-w-60 rounded-xl bg-components-tooltip-bg px-4 py-3.5 text-start text-text-secondary backdrop-blur-[5px]"
+      >
+        {children}
+      </PopoverContent>
+    </Popover>
+  )
+}
+
+function PreviewModeItem({
+  previewEnabled,
+  disabledTip,
+  children,
+}: {
+  previewEnabled: boolean
+  disabledTip: string
+  children: ReactNode
+}) {
+  const item = (
+    <SegmentedControlItem<AgentConfigureRightPanelMode>
+      value="preview"
+      disabled={!previewEnabled}
+      className="uppercase"
+    >
+      {children}
+    </SegmentedControlItem>
+  )
+
+  if (previewEnabled)
+    return item
+
+  return (
+    <Popover>
+      <PopoverTrigger
+        openOnHover
+        nativeButton={false}
+        aria-label={disabledTip}
+        render={<span className="inline-flex" />}
+      >
+        {item}
+      </PopoverTrigger>
+      <PopoverContent
+        placement="bottom"
+        sideOffset={6}
+        popupClassName="max-w-[260px] rounded-md bg-components-tooltip-bg px-3 py-2 system-xs-regular text-text-tertiary shadow-lg"
+      >
+        {disabledTip}
+      </PopoverContent>
+    </Popover>
+  )
+}
+
 export function AgentPreviewHeader({
   mode,
   previewEnabled,
@@ -36,6 +113,7 @@ export function AgentPreviewHeader({
   onOpenWorkingDirectory,
   onRefresh,
   refreshDisabled,
+  showWorkingDirectoryAction = false,
   showChatFeaturesAction = true,
   trailingAction,
 }: {
@@ -47,6 +125,7 @@ export function AgentPreviewHeader({
   onOpenWorkingDirectory: () => void
   onRefresh: () => void
   refreshDisabled?: boolean
+  showWorkingDirectoryAction?: boolean
   showChatFeaturesAction?: boolean
   trailingAction?: ReactNode
 }) {
@@ -56,8 +135,9 @@ export function AgentPreviewHeader({
   const buildTipBody = t('agentDetail.configure.rightPanel.buildTipBody')
   const previewLabel = t('agentDetail.configure.rightPanel.preview')
   const previewTipBody = t('agentDetail.configure.rightPanel.previewTipBody')
+  const previewDisabledTip = t('agentDetail.configure.rightPanel.previewDisabledTip')
   const learnMoreLabel = t('agentDetail.configure.rightPanel.learnMore')
-  const modeTip = `${buildLabel}. ${buildTipBody} ${previewLabel}. ${previewTipBody} ${learnMoreLabel}`
+  const modeTip = `${buildLabel}. ${buildTipBody} ${learnMoreLabel} ${previewLabel}. ${previewTipBody}`
 
   return (
     <div className="relative z-1 flex h-12 shrink-0 items-center justify-between gap-3 px-4 py-2">
@@ -75,42 +155,34 @@ export function AgentPreviewHeader({
             <span aria-hidden className="i-custom-vender-agent-v2-configure-build size-4" />
             {t('agentDetail.configure.rightPanel.build')}
           </SegmentedControlItem>
-          <SegmentedControlItem<AgentConfigureRightPanelMode>
-            value="preview"
-            disabled={!previewEnabled}
-            className="uppercase"
-          >
+          <PreviewModeItem previewEnabled={previewEnabled} disabledTip={previewDisabledTip}>
             <span aria-hidden className="i-custom-vender-agent-v2-configure-preview size-4" />
             {t('agentDetail.configure.rightPanel.preview')}
-          </SegmentedControlItem>
+          </PreviewModeItem>
         </SegmentedControl>
-        <Infotip
-          aria-label={modeTip}
-          placement="bottom"
-          sideOffset={2}
-          className="size-5 rounded-md"
-          iconClassName="size-4 text-text-tertiary"
-          popupClassName="w-60 max-w-60 rounded-xl bg-components-tooltip-bg px-4 py-3.5 text-start text-text-secondary backdrop-blur-[5px]"
-        >
+        <ModeInfoTip ariaLabel={modeTip}>
           <div className="flex flex-col gap-2">
             <div className="flex flex-col gap-3">
               <AgentModeTipSection iconClassName="i-custom-vender-agent-v2-configure-build" title={buildLabel}>
-                {buildTipBody}
+                <>
+                  {buildTipBody}
+                  {' '}
+                  <a
+                    href={docLink('/use-dify/build/new-agent/build#build-by-chatting')}
+                    target="_blank"
+                    rel="noopener noreferrer"
+                    className="body-xs-regular text-text-accent hover:underline"
+                  >
+                    {learnMoreLabel}
+                  </a>
+                </>
               </AgentModeTipSection>
               <AgentModeTipSection iconClassName="i-custom-vender-agent-v2-configure-preview" title={previewLabel}>
                 {previewTipBody}
               </AgentModeTipSection>
             </div>
-            <a
-              href={docLink('/use-dify/build/agent')}
-              target="_blank"
-              rel="noopener noreferrer"
-              className="body-xs-regular text-text-accent hover:underline"
-            >
-              {learnMoreLabel}
-            </a>
           </div>
-        </Infotip>
+        </ModeInfoTip>
       </div>
       <div className="flex shrink-0 items-center">
         <div className="flex items-center gap-2">
@@ -123,14 +195,15 @@ export function AgentPreviewHeader({
           >
             <span aria-hidden className="i-custom-vender-other-replay-line size-4" />
           </button>
-          {mode === 'build' && (
+          {mode === 'build' && showWorkingDirectoryAction && (
             <button
               type="button"
               onClick={onOpenWorkingDirectory}
-              className="flex size-6 items-center justify-center rounded-md p-0.5 text-text-tertiary hover:bg-state-base-hover hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden"
+              className="flex h-8 items-center justify-center gap-0.5 rounded-lg px-3 py-2 text-text-tertiary hover:bg-state-base-hover hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden"
               aria-label={t('agentDetail.configure.workingDirectory.open')}
             >
               <span aria-hidden className="i-ri-folder-3-line size-4" />
+              <span className="px-0.5 system-sm-medium">{t('agentDetail.configure.workingDirectory.fileSystem')}</span>
             </button>
           )}
         </div>
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-breadcrumb.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-breadcrumb.tsx
new file mode 100644
index 00000000000..45395018258
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-breadcrumb.tsx
@@ -0,0 +1,204 @@
+'use client'
+
+import { cn } from '@langgenius/dify-ui/cn'
+import {
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+} from '@langgenius/dify-ui/dropdown-menu'
+import { useTranslation } from 'react-i18next'
+
+const AGENT_WORKING_DIRECTORY_HOME_PATH = '~'
+const AGENT_WORKING_DIRECTORY_ROOT_PATH = '.'
+
+export type AgentWorkingDirectoryPath
+  = | typeof AGENT_WORKING_DIRECTORY_HOME_PATH
+    | typeof AGENT_WORKING_DIRECTORY_ROOT_PATH
+    | string
+
+type AgentWorkingDirectoryBreadcrumbItemData = {
+  iconClassName: string
+  label: string
+  path: AgentWorkingDirectoryPath
+}
+
+const normalizeWorkingDirectoryPath = (path: AgentWorkingDirectoryPath) => {
+  if (path === AGENT_WORKING_DIRECTORY_ROOT_PATH || path === AGENT_WORKING_DIRECTORY_HOME_PATH)
+    return path
+
+  if (path.startsWith('~/'))
+    return `${AGENT_WORKING_DIRECTORY_HOME_PATH}/${path.slice(2).replace(/^\/+|\/+$/g, '')}`
+
+  return path.replace(/^\.\/+/, '').replace(/^\/+|\/+$/g, '')
+}
+
+function buildPathFromSegments(segments: string[], options: { startsFromHome: boolean }) {
+  if (options.startsFromHome)
+    return segments.length ? `${AGENT_WORKING_DIRECTORY_HOME_PATH}/${segments.join('/')}` : AGENT_WORKING_DIRECTORY_HOME_PATH
+
+  return segments.length ? segments.join('/') : AGENT_WORKING_DIRECTORY_ROOT_PATH
+}
+
+function getBreadcrumbItems({
+  homeLabel,
+  path,
+}: {
+  homeLabel: string
+  path: AgentWorkingDirectoryPath
+}): AgentWorkingDirectoryBreadcrumbItemData[] {
+  const normalizedPath = normalizeWorkingDirectoryPath(path)
+  const normalizedHomeLabel = homeLabel === 'home' ? 'Home' : homeLabel
+
+  if (normalizedPath === AGENT_WORKING_DIRECTORY_HOME_PATH) {
+    return [{
+      iconClassName: 'i-ri-folder-3-line',
+      label: normalizedHomeLabel,
+      path: AGENT_WORKING_DIRECTORY_HOME_PATH,
+    }]
+  }
+
+  if (normalizedPath === AGENT_WORKING_DIRECTORY_ROOT_PATH) {
+    return [{
+      iconClassName: 'i-ri-folder-3-line',
+      label: AGENT_WORKING_DIRECTORY_ROOT_PATH,
+      path: AGENT_WORKING_DIRECTORY_ROOT_PATH,
+    }]
+  }
+
+  const startsFromHome = normalizedPath.startsWith(`${AGENT_WORKING_DIRECTORY_HOME_PATH}/`)
+  const segments = startsFromHome
+    ? normalizedPath.slice(2).split('/').filter(Boolean)
+    : normalizedPath.split('/').filter(Boolean)
+
+  const rootItem: AgentWorkingDirectoryBreadcrumbItemData = {
+    iconClassName: 'i-ri-folder-3-line',
+    label: startsFromHome ? normalizedHomeLabel : segments[0]!,
+    path: buildPathFromSegments(startsFromHome ? [] : segments.slice(0, 1), { startsFromHome }),
+  }
+
+  return [
+    rootItem,
+    ...segments.slice(startsFromHome ? 0 : 1).map((segment, index) => {
+      const pathSegments = startsFromHome
+        ? segments.slice(0, index + 1)
+        : segments.slice(0, index + 2)
+
+      return {
+        iconClassName: 'i-ri-folder-3-line',
+        label: segment,
+        path: buildPathFromSegments(pathSegments, { startsFromHome }),
+      }
+    }),
+  ]
+}
+
+function getVisibleBreadcrumbItems(items: AgentWorkingDirectoryBreadcrumbItemData[]) {
+  if (items.length <= 3) {
+    return {
+      hiddenItems: [],
+      visibleItems: items,
+    }
+  }
+
+  return {
+    hiddenItems: items.slice(1, -2),
+    visibleItems: [items[0]!, ...items.slice(-2)],
+  }
+}
+
+function AgentWorkingDirectoryBreadcrumbItem({
+  active,
+  iconClassName,
+  label,
+  onClick,
+}: {
+  active: boolean
+  iconClassName: string
+  label: string
+  onClick: () => void
+}) {
+  return (
+    <button
+      type="button"
+      onClick={onClick}
+      aria-current={active ? 'page' : undefined}
+      className={cn(
+        'flex min-w-0 items-center justify-center gap-1 rounded-md py-0.5 pr-1 pl-0.5 text-text-tertiary hover:bg-state-base-hover hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden',
+        active && 'text-text-secondary',
+      )}
+    >
+      <span aria-hidden className={cn('size-5 shrink-0', iconClassName)} />
+      <span className="min-w-0 truncate system-sm-regular">{label}</span>
+    </button>
+  )
+}
+
+export function AgentWorkingDirectoryBreadcrumb({
+  path,
+  onPathChange,
+}: {
+  path: AgentWorkingDirectoryPath
+  onPathChange: (path: AgentWorkingDirectoryPath) => void
+}) {
+  const { t } = useTranslation('agentV2')
+  const items = getBreadcrumbItems({
+    homeLabel: t('agentDetail.configure.workingDirectory.home'),
+    path,
+  })
+  const { hiddenItems, visibleItems } = getVisibleBreadcrumbItems(items)
+
+  const renderSeparator = (key: string) => (
+    <span key={key} aria-hidden className="system-xs-regular text-divider-deep">/</span>
+  )
+
+  return (
+    <div className="mb-1 flex w-full shrink-0 flex-col border-y-[0.5px] border-divider-regular px-2.5">
+      <nav
+        aria-label={t('agentDetail.configure.workingDirectory.breadcrumbLabel')}
+        className="flex min-w-0 items-center gap-0.5 py-1"
+      >
+        {visibleItems.map((item, index) => {
+          const isLastItem = index === visibleItems.length - 1
+
+          return (
+            <div key={item.path} className="contents">
+              {index > 0 && renderSeparator(`${item.path}-separator`)}
+              {index === 1 && hiddenItems.length > 0 && (
+                <>
+                  <DropdownMenu modal={false}>
+                    <DropdownMenuTrigger
+                      aria-label="..."
+                      className="flex size-6 shrink-0 items-center justify-center rounded-md p-1 text-text-tertiary hover:bg-state-base-hover hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden data-[popup-open]:bg-state-base-hover data-[popup-open]:text-text-secondary"
+                    >
+                      <span aria-hidden className="i-ri-more-fill size-4" />
+                    </DropdownMenuTrigger>
+                    <DropdownMenuContent placement="bottom-start" sideOffset={4} popupClassName="w-[136px] p-1">
+                      {hiddenItems.map(hiddenItem => (
+                        <DropdownMenuItem
+                          key={hiddenItem.path}
+                          className="gap-1 px-2 py-1.5"
+                          onClick={() => onPathChange(hiddenItem.path)}
+                        >
+                          <span aria-hidden className={cn('size-4 shrink-0 text-text-secondary', hiddenItem.iconClassName)} />
+                          <span className="min-w-0 truncate px-1 system-md-regular">{hiddenItem.label}</span>
+                        </DropdownMenuItem>
+                      ))}
+                    </DropdownMenuContent>
+                  </DropdownMenu>
+                  {renderSeparator('hidden-items-separator')}
+                </>
+              )}
+              <AgentWorkingDirectoryBreadcrumbItem
+                active={isLastItem}
+                iconClassName={item.iconClassName}
+                label={item.label}
+                onClick={() => onPathChange(item.path)}
+              />
+            </div>
+          )
+        })}
+      </nav>
+    </div>
+  )
+}
diff --git a/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-panel.tsx b/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-panel.tsx
index 7814769e735..7e327e9c276 100644
--- a/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-panel.tsx
+++ b/web/features/agent-v2/agent-detail/configure/components/preview/working-directory-panel.tsx
@@ -1,14 +1,17 @@
 'use client'
 
 import type { SandboxFileEntryResponse, SandboxListResponse, SandboxReadResponse } from '@dify/contracts/api/console/agent/types.gen'
+import type { AgentWorkingDirectoryPath } from './working-directory-breadcrumb'
 import type { AgentFileNode } from '@/features/agent-v2/agent-composer/form-state'
 import { Dialog } from '@langgenius/dify-ui/dialog'
+import { Tooltip, TooltipContent, TooltipTrigger } from '@langgenius/dify-ui/tooltip'
 import { skipToken, useQueries, useQuery } from '@tanstack/react-query'
 import { useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { consoleQuery } from '@/service/client'
 import { getFileIconType } from '../orchestrate/files/file-icon'
 import { AgentSkillDetailDialog } from '../orchestrate/skills/detail-dialog'
+import { AgentWorkingDirectoryBreadcrumb } from './working-directory-breadcrumb'
 
 type AgentWorkingDirectoryPanelProps = {
   source: AgentWorkingDirectorySource
@@ -33,16 +36,26 @@ type SandboxErrorPayload = {
 }
 
 const normalizeSandboxPath = (path: string) => {
-  const normalizedPath = path.replace(/^\.\//, '').replace(/^\/+|\/+$/g, '')
+  const normalizedPath = path.replace(/^~(?:\/|$)/, '').replace(/^\.\//, '').replace(/^\/+|\/+$/g, '')
   return normalizedPath === '.' ? '' : normalizedPath
 }
 
+const toSandboxHomePath = (path: string) => {
+  if (path === '.')
+    return '.'
+
+  const normalizedPath = normalizeSandboxPath(path)
+  return normalizedPath ? `~/${normalizedPath}` : '~'
+}
+
+const toSandboxApiPath = toSandboxHomePath
+
 const joinSandboxPath = (basePath: string, name: string) => {
   const normalizedBasePath = normalizeSandboxPath(basePath)
   return normalizedBasePath ? `${normalizedBasePath}/${name}` : name
 }
 
-function getSandboxEntryPathSegments(entryName: string, basePath: string) {
+function getSandboxEntryRelativePathSegments(entryName: string, basePath: string) {
   const normalizedBasePath = normalizeSandboxPath(basePath)
   const normalizedEntryName = normalizeSandboxPath(entryName)
 
@@ -52,22 +65,59 @@ function getSandboxEntryPathSegments(entryName: string, basePath: string) {
   if (!normalizedBasePath)
     return normalizedEntryName.split('/').filter(Boolean)
 
-  if (normalizedEntryName === normalizedBasePath || normalizedEntryName.startsWith(`${normalizedBasePath}/`))
-    return normalizedEntryName.split('/').filter(Boolean)
+  if (normalizedEntryName === normalizedBasePath)
+    return []
 
-  return [...normalizedBasePath.split('/').filter(Boolean), ...normalizedEntryName.split('/').filter(Boolean)]
+  if (normalizedEntryName.startsWith(`${normalizedBasePath}/`))
+    return normalizedEntryName.slice(normalizedBasePath.length + 1).split('/').filter(Boolean)
+
+  return normalizedEntryName.split('/').filter(Boolean)
 }
 
-function buildSandboxFileTree(entries: SandboxFileEntryResponse[] = [], basePath = '.'): AgentFileNode[] {
+function buildSandboxFileTree(
+  entries: SandboxFileEntryResponse[] = [],
+  basePath = '.',
+  options: { nestRootPath?: string, nestUnderBasePath?: boolean } = {},
+): AgentFileNode[] {
+  const normalizedBasePath = normalizeSandboxPath(basePath)
+  const normalizedNestRootPath = normalizeSandboxPath(options.nestRootPath ?? '.')
   const rootFiles: AgentFileNode[] = []
+  let baseFolder: AgentFileNode | undefined
+
+  if (options.nestUnderBasePath && normalizedBasePath) {
+    let currentFiles = rootFiles
+    let currentPath = normalizedNestRootPath
+    const basePathSegments = normalizedBasePath.split('/').filter(Boolean)
+    const nestRootPathSegments = normalizedNestRootPath.split('/').filter(Boolean)
+    const nestedBasePathSegments = normalizedNestRootPath && (
+      normalizedBasePath === normalizedNestRootPath
+      || normalizedBasePath.startsWith(`${normalizedNestRootPath}/`)
+    )
+      ? basePathSegments.slice(nestRootPathSegments.length)
+      : basePathSegments
+
+    nestedBasePathSegments.forEach((segment) => {
+      currentPath = joinSandboxPath(currentPath, segment)
+      const folder: AgentFileNode = {
+        id: currentPath,
+        name: segment,
+        icon: 'folder',
+        children: [],
+      }
+
+      currentFiles.push(folder)
+      currentFiles = folder.children ?? []
+      baseFolder = folder
+    })
+  }
 
   for (const entry of entries) {
-    const pathSegments = getSandboxEntryPathSegments(entry.name, basePath)
+    const pathSegments = getSandboxEntryRelativePathSegments(entry.name, basePath)
     if (!pathSegments.length)
       continue
 
-    let currentFiles = rootFiles
-    let currentPath = ''
+    let currentFiles = baseFolder?.children ?? rootFiles
+    let currentPath = normalizedBasePath
 
     pathSegments.forEach((segment, index) => {
       const isLeaf = index === pathSegments.length - 1
@@ -169,6 +219,16 @@ async function isNoActiveSessionError(error: unknown) {
 
 const isNotFoundResponse = (error: unknown) => error instanceof Response && error.status === 404
 
+function isSandboxPathWithinDirectory(path: string, directory: string) {
+  const normalizedPath = normalizeSandboxPath(path)
+  const normalizedDirectory = normalizeSandboxPath(directory)
+
+  if (!normalizedDirectory)
+    return true
+
+  return normalizedPath === normalizedDirectory || normalizedPath.startsWith(`${normalizedDirectory}/`)
+}
+
 export function AgentWorkingDirectoryPanel({
   source,
   onOpenChange,
@@ -176,6 +236,7 @@ export function AgentWorkingDirectoryPanel({
 }: AgentWorkingDirectoryPanelProps) {
   const { t } = useTranslation('agentV2')
   const { t: tCommon } = useTranslation('common')
+  const [selectedDirectoryPath, setSelectedDirectoryPath] = useState<AgentWorkingDirectoryPath>()
   const [selectedFileId, setSelectedFileId] = useState<string>()
   const [loadedFolderPaths, setLoadedFolderPaths] = useState<string[]>([])
   const [openFolderPaths, setOpenFolderPaths] = useState<string[]>([])
@@ -186,16 +247,40 @@ export function AgentWorkingDirectoryPanel({
   const hasWorkingDirectorySource = source.type === 'agent'
     ? !!source.conversationId
     : !!source.appId && !!workflowNodeRunId
+  const sandboxInfoQueryOptions = consoleQuery.agent.byAgentId.sandbox.get.queryOptions({
+    input: source.type === 'agent' && source.conversationId
+      ? {
+          params: {
+            agent_id: source.agentId,
+          },
+          query: {
+            conversation_id: source.conversationId,
+          },
+        }
+      : skipToken,
+    context: {
+      silent: true,
+    },
+  })
+  const sandboxInfoQuery = useQuery({
+    ...sandboxInfoQueryOptions,
+    enabled: open && source.type === 'agent' && !!source.conversationId,
+    retry: false,
+  })
+  const isSandboxInfoLoading = source.type === 'agent' && !!source.conversationId && sandboxInfoQuery.isPending
+  const workspaceDirectoryPath = sandboxInfoQuery.data?.workspace_cwd
+  const directoryPath = selectedDirectoryPath ?? workspaceDirectoryPath ?? '.'
+  const showReturnToWorkspaceButton = !!workspaceDirectoryPath && !isSandboxPathWithinDirectory(directoryPath, workspaceDirectoryPath)
   const getFileListQueryOptions = (path: string) => source.type === 'agent'
     ? consoleQuery.agent.byAgentId.sandbox.files.get.queryOptions({
-        input: source.conversationId
+        input: source.conversationId && !isSandboxInfoLoading
           ? {
               params: {
                 agent_id: source.agentId,
               },
               query: {
                 conversation_id: source.conversationId,
-                path,
+                path: toSandboxApiPath(path),
               },
             }
           : skipToken,
@@ -212,7 +297,7 @@ export function AgentWorkingDirectoryPanel({
                 node_id: source.nodeId,
               },
               query: {
-                path,
+                path: toSandboxApiPath(path),
               },
             }
           : skipToken,
@@ -220,7 +305,14 @@ export function AgentWorkingDirectoryPanel({
           silent: true,
         },
       })
-  const fileListQueryOptions = getFileListQueryOptions('.')
+  const handleDirectoryPathChange = (path: AgentWorkingDirectoryPath) => {
+    setSelectedDirectoryPath(path)
+    setSelectedFileId(undefined)
+    setLoadedFolderPaths([])
+    setOpenFolderPaths([])
+    setPendingOpenFolderPaths([])
+  }
+  const fileListQueryOptions = getFileListQueryOptions(directoryPath)
   const fileListQuery = useQuery({
     ...fileListQueryOptions,
     queryFn: async (context): Promise<SandboxListResponse> => {
@@ -267,12 +359,19 @@ export function AgentWorkingDirectoryPanel({
         })
       : [],
   })
-  const workingDirectoryFiles = expandedFolderQueries.reduce((files, query) => {
-    return mergeSandboxFileTree(files, buildSandboxFileTree(query.data?.entries, query.data?.path))
+  const workingDirectoryFiles = expandedFolderQueries.reduce((files, query, index) => {
+    return mergeSandboxFileTree(files, buildSandboxFileTree(
+      query.data?.entries,
+      loadedFolderPaths[index] ?? query.data?.path,
+      {
+        nestRootPath: directoryPath,
+        nestUnderBasePath: true,
+      },
+    ))
   }, buildSandboxFileTree(fileListQuery.data?.entries, fileListQuery.data?.path))
   const selectedWorkingDirectoryFile = findReadableFile(workingDirectoryFiles, selectedFileId)
     ?? findFirstReadableFile(workingDirectoryFiles)
-  const isFileListLoading = hasWorkingDirectorySource && fileListQuery.isPending
+  const isFileListLoading = hasWorkingDirectorySource && (isSandboxInfoLoading || fileListQuery.isPending)
   const loadingFolderPaths = new Set(loadedFolderPaths.filter((path, index) => expandedFolderQueries[index]?.isPending))
   const loadedFolderPathIndexes = new Map(loadedFolderPaths.map((path, index) => [path, index]))
   const fileReadQueryOptions = source.type === 'agent'
@@ -284,7 +383,7 @@ export function AgentWorkingDirectoryPanel({
               },
               query: {
                 conversation_id: source.conversationId,
-                path: selectedWorkingDirectoryFile.id,
+                path: toSandboxApiPath(selectedWorkingDirectoryFile.id),
               },
             }
           : skipToken,
@@ -301,7 +400,7 @@ export function AgentWorkingDirectoryPanel({
                 node_id: source.nodeId,
               },
               query: {
-                path: selectedWorkingDirectoryFile.id,
+                path: toSandboxApiPath(selectedWorkingDirectoryFile.id),
               },
             }
           : skipToken,
@@ -340,6 +439,43 @@ export function AgentWorkingDirectoryPanel({
         detail={{
           description: t('agentDetail.configure.workingDirectory.description'),
           fileCount: countReadableFiles(workingDirectoryFiles),
+          fileListHeader: isSandboxInfoLoading
+            ? (
+                <h3 id="agent-skill-detail-files-heading" className="px-4 pt-3.5 pb-3 system-xl-semibold text-text-primary">
+                  {t('agentDetail.configure.workingDirectory.fileSystem')}
+                </h3>
+              )
+            : (
+                <div className="flex shrink-0 flex-col">
+                  <div className="flex items-center gap-1 px-4 pt-3.5 pb-3">
+                    <h3 id="agent-skill-detail-files-heading" className="min-w-0 flex-1 system-xl-semibold text-text-primary">
+                      {t('agentDetail.configure.workingDirectory.fileSystem')}
+                    </h3>
+                    {showReturnToWorkspaceButton && (
+                      <Tooltip>
+                        <TooltipTrigger
+                          aria-label={t('agentDetail.configure.workingDirectory.returnToWorkspace')}
+                          className="flex size-6 shrink-0 items-center justify-center rounded-md p-1 text-text-tertiary hover:bg-state-base-hover hover:text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden"
+                          onClick={() => handleDirectoryPathChange(workspaceDirectoryPath)}
+                        >
+                          <span aria-hidden className="i-ri-arrow-go-back-line size-3.5" />
+                        </TooltipTrigger>
+                        <TooltipContent placement="top">
+                          {t('agentDetail.configure.workingDirectory.returnToWorkspace')}
+                        </TooltipContent>
+                      </Tooltip>
+                    )}
+                  </div>
+                  <AgentWorkingDirectoryBreadcrumb
+                    path={directoryPath}
+                    onPathChange={handleDirectoryPathChange}
+                  />
+                </div>
+              ),
+          fileListLoading: isSandboxInfoLoading,
+          fileListPanelClassName: 'w-[360px]',
+          fileListTreeClassName: 'px-0',
+          fileListTreeListClassName: 'px-1',
           fileListTitle: t('agentDetail.configure.workingDirectory.title'),
           files: workingDirectoryFiles,
           filePreview: {
@@ -371,6 +507,7 @@ export function AgentWorkingDirectoryPanel({
               ? (paths.includes(file.id) ? paths : [...paths, file.id])
               : paths.filter(path => path !== file.id))
           },
+          onFolderDoubleClick: ({ file }) => handleDirectoryPathChange(toSandboxHomePath(file.id)),
           onSelectFile: selectedFile => setSelectedFileId(selectedFile.id),
           renderFolderSuffix: ({ file }) => loadingFolderPaths.has(file.id)
             ? (
diff --git a/web/features/agent-v2/agent-detail/configure/model-compatibility.ts b/web/features/agent-v2/agent-detail/configure/model-compatibility.ts
index fece602ba37..e4fa261c80f 100644
--- a/web/features/agent-v2/agent-detail/configure/model-compatibility.ts
+++ b/web/features/agent-v2/agent-detail/configure/model-compatibility.ts
@@ -1,75 +1,99 @@
 import type { Model, ModelItem } from '@/app/components/header/account-setting/model-provider-page/declarations'
 
-type ProviderModelCompatibility = {
-  providers: string[]
-  incompatibleModels: RegExp[]
-}
+const agentIncompatibleModelPatterns: RegExp[] = [
+  // openai
+  /^chatgpt-/i,
+  /^gpt-5(?:$|-|\.1$|\.2$)/i,
+  /^gpt.*(?:min|chat)/i,
+  /^gpt.*nano/i,
+  /^gpt-4/i,
+  /^gpt-3/i,
+  /^o[134](?:-|$)/i,
 
-const agentIncompatibleModelConfig: ProviderModelCompatibility[] = [
-  {
-    providers: ['openai'],
-    incompatibleModels: [
-      /^gpt-4o-mini(?:-|$)/i,
-      /^gpt-4\.1-(?:mini|nano)(?:-|$)/i,
-      /^gpt-4(?:-|$)/i,
-      /^gpt-3\.5/i,
-      /^o[34]-mini(?:-|$)/i,
-    ],
-  },
-  {
-    providers: ['anthropic'],
-    incompatibleModels: [
-      /^claude-3-(?:haiku|sonnet|opus)(?:-|$)/i,
-      /^claude-3(?:\.5|-5)-(?:haiku|sonnet)(?:-|$)/i,
-    ],
-  },
-  {
-    providers: ['gemini', 'google'],
-    incompatibleModels: [
-      /^gemini-2[.-][05]-flash(?:-lite)?(?:-|$)/i,
-      /^gemini-1[.-]5-flash(?:-8b)?(?:-|$)/i,
-    ],
-  },
-  {
-    providers: ['deepseek'],
-    incompatibleModels: [
-      /^deepseek-r1$/i,
-      /^deepseek-r1-lite$/i,
-      /^deepseek-r1-distill(?:-[a-z0-9]+)*-(?:1\.5b|7b|8b|14b|32b|70b)$/i,
-    ],
-  },
-  {
-    providers: ['minimax'],
-    incompatibleModels: [
-      /^minimax-text-01$/i,
-      /^minimax-m1$/i,
-    ],
-  },
-  {
-    providers: ['tongyi', 'qwen'],
-    incompatibleModels: [
-      /^qwen2[.-]5(?:-[a-z0-9.]+)?-instruct(?:-|$)/i,
-      /^qwen2[.-]5-coder(?:-|$)/i,
-      /^qwen3-(?:0\.6b|1\.7b|4b|8b|14b|30b)(?:-|$)/i,
-    ],
-  },
-  {
-    providers: ['chatglm', 'zhipuai'],
-    incompatibleModels: [
-      /^glm-4-(?:air|airx|flash)$/i,
-      /^glm-z1-(?:air|flash)$/i,
-    ],
-  },
+  // anthropic
+  /^claude-3/i,
+
+  // gemini
+  /^gemini.*flash[ .-]lite/i,
+  /^gemini[ .-]2(?![ .-]5[ .-]pro$)/i,
+  /^gemini[ .-]1[ .-]5[ .-]flash(?:[ .-]8b)?(?:[ .-]|$)/i,
+  /^Nano/i,
+
+  // x
+  /^grok-code-/i,
+  /^grok.*beta/i,
+  /^grok-(?:2|3)/i,
+
+  // deepseek
+  /^deepseek-(?:chat|coder|reasoner)(?:-|$)/i,
+  /^deepseek-r1$/i,
+  /^deepseek-r1-distill-/i,
+  /^deepseek-v3/i,
+
+  // qwen
+  /^qvq-/i,
+  /^qwq-/i,
+  /^qwen-/i,
+  /^qwen2/i,
+  /^qwen3-/i,
+  /^qwen3\.5/i,
+  /^qwen.*flash/i,
+  /^farui-plus$/i,
+
+  // zhipu ai
+  /^chatglm-(?:2|3)/i,
+  /^glm-3/i,
+  /^glm-4/i,
+  /^glm-z/i,
+
+  // moonshot
+  /^kimi-k2-/i,
+  /^moonshot-v1/i,
+
+  // minimax
+  /^abab/i,
+  /^minimax-text-01$/i,
+  /^minimax-m1$/i,
+  /^minimax-m2(?:-|$)/i,
 ]
 
-const normalizeProviderName = (provider: string) => provider.split('/').at(-1)?.toLowerCase() ?? provider.toLowerCase()
+const agentSuggestedModelPatterns: RegExp[] = [
+  // openai
+  /^gpt[ .-]5\.5$/i,
+  /^gpt[ .-]5\.5[ .-]pro$/i,
 
-export function isAgentCompatibleModel(provider: Model, modelItem: ModelItem) {
-  const providerName = normalizeProviderName(provider.provider)
-  const providerConfig = agentIncompatibleModelConfig.find(config => config.providers.includes(providerName))
+  // anthropic
+  /^(?:claude[ .-])?opus[ .-]4\.8$/i,
+  /^(?:claude[ .-])?opus[ .-]4\.7$/i,
+  /^(?:claude[ .-])?sonnet[ .-]4\.6$/i,
 
-  if (!providerConfig)
-    return true
+  // gemini
+  /^gemini 3\.1 pro preview$/i,
 
-  return !providerConfig.incompatibleModels.some(pattern => pattern.test(modelItem.model))
+  // x
+  /^grok[ .-]4\.3$/i,
+
+  // deepseek
+  /^deepseek[ .-]v4[ .-]pro$/i,
+
+  // qwen
+  /^qwen[ .-]?3\.7[ .-]max$/i,
+  /^qwen[ .-]?3[ .-]coder[ .-]plus$/i,
+
+  // moonshot
+  /^kimi[ .-]k2\.6$/i,
+
+  // minimax
+  /^minimax[ .-]m3$/i,
+
+  // zhipuai
+  /^glm[ .-]5\.1$/i,
+]
+
+export function isAgentCompatibleModel(_provider: Model, modelItem: ModelItem) {
+  return !agentIncompatibleModelPatterns.some(pattern => pattern.test(modelItem.label.en_US))
+}
+
+export function isAgentSuggestedModel(_provider: Model, modelItem: ModelItem) {
+  return agentSuggestedModelPatterns.some(pattern => pattern.test(modelItem.label.en_US))
 }
diff --git a/web/features/agent-v2/agent-detail/configure/use-agent-configure-build-draft.ts b/web/features/agent-v2/agent-detail/configure/use-agent-configure-build-draft.ts
index 900a960540e..85cb183204d 100644
--- a/web/features/agent-v2/agent-detail/configure/use-agent-configure-build-draft.ts
+++ b/web/features/agent-v2/agent-detail/configure/use-agent-configure-build-draft.ts
@@ -1,8 +1,9 @@
 'use client'
 
 import type { AgentSoulConfig } from '@dify/contracts/api/console/agent/types.gen'
-import type { AgentBuildDraftChangedKey } from './components/orchestrate/build-draft-changes-context'
+import type { AgentBuildDraftChangedKey, AgentBuildDraftChangeItem, AgentBuildDraftChangeSummary } from './components/orchestrate/build-draft-changes-context'
 import type { AgentConfigureSoulSource } from './state'
+import type { AgentFileNode, AgentSoulConfigFormState } from '@/features/agent-v2/agent-composer/form-state'
 import { toast } from '@langgenius/dify-ui/toast'
 import { useMutation, useQuery, useQueryClient } from '@tanstack/react-query'
 import isEqual from 'fast-deep-equal'
@@ -13,9 +14,121 @@ import { consoleQuery } from '@/service/client'
 import { usePrepareAgentBuildDraftBeforeRun } from './use-agent-build-draft-run'
 
 const isNotFoundResponse = (error: unknown) => error instanceof Response && error.status === 404
+const BUILD_NOTE_FILE_ID = '__agent_config_build_note__'
+const BUILD_NOTE_FILE_NAME = 'build_note.md'
+
 const getAgentSoulConfigFromRefetchResult = (result: unknown) => {
   return (result as { data?: { agent_soul?: AgentSoulConfig } } | undefined)?.data?.agent_soul
 }
+const flattenFileNodes = (files: AgentFileNode[]): AgentFileNode[] => files.flatMap(file => (
+  file.children?.length ? [file, ...flattenFileNodes(file.children)] : [file]
+))
+
+function getItemDiff<TItem>({
+  currentItems,
+  nextItems,
+  getIcon,
+  getKey,
+  getName,
+}: {
+  currentItems: readonly TItem[]
+  nextItems: readonly TItem[]
+  getIcon?: (item: TItem) => AgentBuildDraftChangeItem['icon']
+  getKey: (item: TItem) => string
+  getName: (item: TItem) => string
+}): AgentBuildDraftChangeItem[] {
+  const currentByKey = new Map(currentItems.map(item => [getKey(item), item]))
+  const nextByKey = new Map(nextItems.map(item => [getKey(item), item]))
+  const changes: AgentBuildDraftChangeItem[] = []
+
+  for (const item of nextItems) {
+    const key = getKey(item)
+    const currentItem = currentByKey.get(key)
+    if (!currentItem) {
+      changes.push({
+        id: key,
+        name: getName(item),
+        operation: 'added',
+        icon: getIcon?.(item),
+      })
+      continue
+    }
+
+    if (!isEqual(item, currentItem)) {
+      changes.push({
+        id: key,
+        name: getName(item),
+        operation: 'updated',
+        icon: getIcon?.(item),
+      })
+    }
+  }
+
+  for (const item of currentItems) {
+    const key = getKey(item)
+    if (nextByKey.has(key))
+      continue
+
+    changes.push({
+      id: key,
+      name: getName(item),
+      operation: 'removed',
+      icon: getIcon?.(item),
+    })
+  }
+
+  return changes
+}
+
+function getAgentBuildDraftChangeSummary({
+  buildDraft,
+  changedKeys,
+  normalAgentSoulConfig,
+  normalDraft,
+}: {
+  buildDraft: AgentSoulConfigFormState
+  changedKeys: readonly AgentBuildDraftChangedKey[]
+  normalAgentSoulConfig: AgentSoulConfig
+  normalDraft: AgentSoulConfigFormState
+}): AgentBuildDraftChangeSummary {
+  const buildNoteChange = {
+    id: BUILD_NOTE_FILE_ID,
+    name: BUILD_NOTE_FILE_NAME,
+    operation: normalAgentSoulConfig.config_note?.trim() ? 'updated' : 'added',
+    icon: 'markdown',
+    descriptionKey: 'agentDetail.configure.buildDraft.buildNoteDescription',
+  } satisfies AgentBuildDraftChangeItem
+  const fileChanges = [
+    buildNoteChange,
+    ...getItemDiff({
+      currentItems: flattenFileNodes(normalDraft.files),
+      nextItems: flattenFileNodes(buildDraft.files),
+      getIcon: file => file.icon,
+      getKey: file => file.configName ?? file.id ?? file.name,
+      getName: file => file.name,
+    }),
+  ]
+  const skillChanges = getItemDiff({
+    currentItems: normalDraft.skills,
+    nextItems: buildDraft.skills,
+    getKey: skill => skill.id || skill.name,
+    getName: skill => skill.name,
+  })
+  const envVariableChanges = getItemDiff({
+    currentItems: normalDraft.envVariables,
+    nextItems: buildDraft.envVariables,
+    getKey: variable => variable.id || variable.key,
+    getName: variable => variable.key,
+  })
+
+  return {
+    changedKeys,
+    changesCount: fileChanges.length + skillChanges.length + envVariableChanges.length,
+    skills: skillChanges,
+    files: fileChanges,
+    envVariables: envVariableChanges,
+  }
+}
 
 export function useAgentConfigureBuildDraftData({
   agentId,
@@ -92,22 +205,36 @@ export function useAgentConfigureBuildDraftData({
   const isBuildDraftActive = soulSource === 'build-draft'
   const buildDraftAgentSoulConfig = buildDraftData?.agent_soul as AgentSoulConfig | undefined
   const visibleAgentSoulConfig = isBuildDraftActive ? buildDraftAgentSoulConfig : normalAgentSoulConfig
-  const buildDraftChangedKeys = useMemo<AgentBuildDraftChangedKey[]>(() => {
-    if (!buildDraftAgentSoulConfig || !composerAgentSoulConfig)
-      return []
+  const buildDraftChangeSummary = useMemo<AgentBuildDraftChangeSummary>(() => {
+    if (!buildDraftAgentSoulConfig || !composerAgentSoulConfig) {
+      return {
+        changedKeys: [],
+        changesCount: 0,
+        skills: [],
+        files: [],
+        envVariables: [],
+      }
+    }
 
     const normalDraft = agentSoulConfigToFormState(composerAgentSoulConfig)
     const buildDraft = agentSoulConfigToFormState(buildDraftAgentSoulConfig)
-
-    return (Object.keys(buildDraft) as Array<keyof typeof buildDraft>)
+    const changedKeys = (Object.keys(buildDraft) as Array<keyof typeof buildDraft>)
       .filter(key => !isEqual(buildDraft[key], normalDraft[key]))
+
+    return getAgentBuildDraftChangeSummary({
+      buildDraft,
+      changedKeys,
+      normalAgentSoulConfig: composerAgentSoulConfig,
+      normalDraft,
+    })
   }, [buildDraftAgentSoulConfig, composerAgentSoulConfig])
 
   return {
     activeVersionId: isBuildDraftActive ? `build-draft:${buildDraftDataUpdatedAt}` : activeVersionId,
     agentSoulConfig: visibleAgentSoulConfig,
-    changedKeys: buildDraftChangedKeys,
-    changesCount: buildDraftChangedKeys.length,
+    changedKeys: buildDraftChangeSummary.changedKeys,
+    changeSummary: buildDraftChangeSummary,
+    changesCount: buildDraftChangeSummary.changesCount,
     isActive: isBuildDraftActive,
     isPending: !isViewingVersion && soulSourceOverride !== 'draft' && soulSourceOverride !== 'view-version' && isBuildDraftPending,
     refetch: refetchBuildDraft,
diff --git a/web/features/agent-v2/agent-detail/configure/use-agent-configure-sync.ts b/web/features/agent-v2/agent-detail/configure/use-agent-configure-sync.ts
index 7fa6a53537b..a3493702cf2 100644
--- a/web/features/agent-v2/agent-detail/configure/use-agent-configure-sync.ts
+++ b/web/features/agent-v2/agent-detail/configure/use-agent-configure-sync.ts
@@ -12,7 +12,7 @@ import { useCallback, useEffect, useMemo, useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { useSerialAsyncCallback } from '@/app/components/workflow/hooks/use-serial-async-callback'
 import { formStateToAgentSoulConfig } from '@/features/agent-v2/agent-composer/conversions'
-import { validateKnowledgeRetrievals } from '@/features/agent-v2/agent-composer/knowledge-validation'
+import { useKnowledgeValidationMessage, validateKnowledgeRetrievals } from '@/features/agent-v2/agent-composer/knowledge-validation'
 import {
   agentComposerDraftAtom,
   agentComposerOriginalConfigAtom,
@@ -24,12 +24,6 @@ import { consoleQuery } from '@/service/client'
 
 const DRAFT_AUTOSAVE_WAIT = 5000
 
-class InvalidKnowledgeConfigurationError extends Error {
-  constructor() {
-    super('Agent knowledge retrieval configuration is invalid.')
-  }
-}
-
 export function useAgentConfigureSync({
   agentId,
   baseConfig,
@@ -42,6 +36,7 @@ export function useAgentConfigureSync({
   enabled: boolean
 }) {
   const { t: tCommon } = useTranslation('common')
+  const getKnowledgeValidationMessage = useKnowledgeValidationMessage()
   const queryClient = useQueryClient()
   const store = useStore()
   const setOriginalConfig = useSetAtom(agentComposerOriginalConfigAtom)
@@ -127,8 +122,6 @@ export function useAgentConfigureSync({
   const latestDraftSaveRef = useRef<() => void>(() => undefined)
   latestDraftSaveRef.current = () => {
     const draft = store.get(agentComposerDraftAtom)
-    if (!validateKnowledgeRetrievals(draft.knowledgeRetrievals).isValid)
-      return
 
     void saveComposer({
       configSnapshot: getAgentSoulDraft(),
@@ -145,8 +138,6 @@ export function useAgentConfigureSync({
       return
 
     const draft = store.get(agentComposerDraftAtom)
-    if (!validateKnowledgeRetrievals(draft.knowledgeRetrievals).isValid)
-      throw new InvalidKnowledgeConfigurationError()
     const configSnapshot = getAgentSoulDraft()
     const hasEffectiveModelChange = !isEqual(configSnapshot.model, baseConfigRef.current?.model)
     debouncedSaveDraft.cancel?.()
@@ -168,7 +159,6 @@ export function useAgentConfigureSync({
     const draft = store.get(agentComposerDraftAtom)
     if (
       !store.get(isAgentComposerDirtyAtom)
-      || !validateKnowledgeRetrievals(draft.knowledgeRetrievals).isValid
     ) {
       return
     }
@@ -209,8 +199,7 @@ export function useAgentConfigureSync({
       }
 
       if (
-        !validateKnowledgeRetrievals(store.get(agentComposerDraftAtom).knowledgeRetrievals).isValid
-        || lastAutosavedDraftKeyRef.current === agentSoulDraftKey
+        lastAutosavedDraftKeyRef.current === agentSoulDraftKey
       ) {
         return
       }
@@ -248,8 +237,11 @@ export function useAgentConfigureSync({
       return
 
     const draft = store.get(agentComposerDraftAtom)
-    if (!validateKnowledgeRetrievals(draft.knowledgeRetrievals).isValid)
-      throw new InvalidKnowledgeConfigurationError()
+    const knowledgeValidation = validateKnowledgeRetrievals(draft.knowledgeRetrievals)
+    if (!knowledgeValidation.isValid) {
+      toast.error(getKnowledgeValidationMessage(knowledgeValidation.firstIssue?.code) ?? tCommon('api.actionFailed'))
+      return
+    }
 
     publishInFlightRef.current = true
     setIsPublishInFlight(true)
@@ -302,7 +294,7 @@ export function useAgentConfigureSync({
       publishInFlightRef.current = false
       setIsPublishInFlight(false)
     }
-  }, [agentId, debouncedSaveDraft, publishAgent, queryClient, saveComposer, setOriginalConfig, setOriginalDraft, setPublishedDraft, store, tCommon])
+  }, [agentId, debouncedSaveDraft, getKnowledgeValidationMessage, publishAgent, queryClient, saveComposer, setOriginalConfig, setOriginalDraft, setPublishedDraft, store, tCommon])
 
   return {
     draftSavedAt,
diff --git a/web/features/agent-v2/agent-detail/logs/__tests__/page.spec.tsx b/web/features/agent-v2/agent-detail/logs/__tests__/page.spec.tsx
index 0f625075699..4b60e27366a 100644
--- a/web/features/agent-v2/agent-detail/logs/__tests__/page.spec.tsx
+++ b/web/features/agent-v2/agent-detail/logs/__tests__/page.spec.tsx
@@ -1,4 +1,4 @@
-import type { AgentLogListResponse, AgentLogSourceListResponse } from '@dify/contracts/api/console/agent/types.gen'
+import type { AgentLogListResponse, AgentLogMessageListResponse, AgentLogSourceListResponse } from '@dify/contracts/api/console/agent/types.gen'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
 import { render, screen, waitFor } from '@testing-library/react'
 import userEvent from '@testing-library/user-event'
@@ -16,6 +16,7 @@ type AgentLogsQueryInput = {
 const mocks = vi.hoisted(() => ({
   logsQueryFn: vi.fn(),
   logSourcesQueryFn: vi.fn(),
+  messagesQueryFn: vi.fn(),
   logsQueryOptions: vi.fn((input: AgentLogsQueryInput) => ({
     queryKey: ['agent-logs', input],
     queryFn: () => mocks.logsQueryFn(input),
@@ -24,6 +25,10 @@ const mocks = vi.hoisted(() => ({
     queryKey: ['agent-log-sources', input],
     queryFn: () => mocks.logSourcesQueryFn(input),
   })),
+  messagesQueryOptions: vi.fn((input: AgentLogsQueryInput) => ({
+    queryKey: ['agent-log-messages', input],
+    queryFn: () => mocks.messagesQueryFn(input),
+  })),
 }))
 
 vi.mock('@/context/i18n', () => ({
@@ -49,6 +54,13 @@ vi.mock('@/service/client', () => ({
           get: {
             queryOptions: mocks.logsQueryOptions,
           },
+          byConversationId: {
+            messages: {
+              get: {
+                queryOptions: mocks.messagesQueryOptions,
+              },
+            },
+          },
         },
       },
     },
@@ -133,6 +145,34 @@ const logSourcesResponse: AgentLogSourceListResponse = {
   ],
 }
 
+const messagesResponse: AgentLogMessageListResponse = {
+  data: [
+    {
+      answer: 'Translated chapter summary',
+      answer_tokens: 12,
+      conversation_id: 'conversation-1',
+      created_at: 1781660001,
+      currency: 'USD',
+      error: null,
+      from_account_id: null,
+      from_end_user_id: 'end-user-1',
+      id: 'message-1',
+      latency: 1.234,
+      message_id: 'message-1',
+      message_tokens: 8,
+      query: 'Translate this chapter',
+      status: 'success',
+      total_price: '0.001',
+      total_tokens: 20,
+      updated_at: 1781660002,
+    },
+  ],
+  has_more: false,
+  limit: 100,
+  page: 1,
+  total: 1,
+}
+
 const renderPage = () => {
   const queryClient = new QueryClient({
     defaultOptions: {
@@ -165,6 +205,7 @@ describe('AgentLogsPage', () => {
     vi.clearAllMocks()
     mocks.logsQueryFn.mockResolvedValue(emptyLogsResponse)
     mocks.logSourcesQueryFn.mockResolvedValue(logSourcesResponse)
+    mocks.messagesQueryFn.mockResolvedValue(messagesResponse)
   })
 
   describe('Query contract', () => {
@@ -266,5 +307,33 @@ describe('AgentLogsPage', () => {
 
       resolveNextLogs(emptyLogsResponse)
     })
+
+    it('should open chatbot-style log detail drawer with generated messages contract when a row is clicked', async () => {
+      const user = userEvent.setup()
+      mocks.logsQueryFn.mockResolvedValue(populatedLogsResponse)
+
+      renderPage()
+
+      await user.click(await screen.findByRole('button', { name: 'Previous conversation' }))
+
+      await waitFor(() => {
+        expect(mocks.messagesQueryOptions).toHaveBeenCalledWith({
+          input: {
+            params: {
+              agent_id: 'agent-1',
+              conversation_id: 'conversation-1',
+            },
+            query: {
+              limit: 100,
+              page: 1,
+              sort_by: 'created_at',
+              sort_order: 'asc',
+              sources: ['webapp:webapp-app-id'],
+            },
+          },
+        })
+      })
+      expect(await screen.findByText('Translated chapter summary')).toBeInTheDocument()
+    })
   })
 })
diff --git a/web/features/agent-v2/agent-detail/logs/components/log-detail-panel.tsx b/web/features/agent-v2/agent-detail/logs/components/log-detail-panel.tsx
new file mode 100644
index 00000000000..610955b7492
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/logs/components/log-detail-panel.tsx
@@ -0,0 +1,164 @@
+import type { AgentLogConversationItemResponse, AgentLogMessageItemResponse } from '@dify/contracts/api/console/agent/types.gen'
+import type { IChatItem } from '@/app/components/base/chat/chat/type'
+import { Tooltip, TooltipContent, TooltipTrigger } from '@langgenius/dify-ui/tooltip'
+import { skipToken, useQuery } from '@tanstack/react-query'
+import { noop } from 'es-toolkit/function'
+import { useTranslation } from 'react-i18next'
+import ActionButton from '@/app/components/base/action-button'
+import Chat from '@/app/components/base/chat/chat'
+import CopyIcon from '@/app/components/base/copy-icon'
+import Loading from '@/app/components/base/loading'
+import useTimestamp from '@/hooks/use-timestamp'
+import { consoleQuery } from '@/service/client'
+
+export function AgentLogDetailPanel({
+  agentId,
+  log,
+  onClose,
+}: {
+  agentId: string
+  log?: AgentLogConversationItemResponse
+  onClose: () => void
+}) {
+  const { t } = useTranslation()
+  const { t: tAgentV2 } = useTranslation('agentV2')
+  const { formatTime } = useTimestamp()
+  const messagesQuery = useQuery(consoleQuery.agent.byAgentId.logs.byConversationId.messages.get.queryOptions({
+    input: log
+      ? {
+          params: {
+            agent_id: agentId,
+            conversation_id: log.conversation_id,
+          },
+          query: {
+            limit: 100,
+            page: 1,
+            sort_by: 'created_at',
+            sort_order: 'asc',
+            ...(log.source ? { sources: [log.source.id] } : {}),
+          },
+        }
+      : skipToken,
+  }))
+  const chatList = log
+    ? formatAgentLogMessages({
+        conversationId: log.conversation_id,
+        formatLogTime: value => formatTime(value, tAgentV2('roster.dateTimeFormat')),
+        messages: messagesQuery.data?.data ?? [],
+      })
+    : []
+
+  return (
+    <div className="flex h-full flex-col rounded-xl border-[0.5px] border-components-panel-border">
+      <div className="flex shrink-0 items-center gap-2 rounded-t-xl bg-components-panel-bg pt-3 pr-3 pb-2 pl-4">
+        <div className="min-w-0 shrink-0">
+          <div className="mb-0.5 system-xs-semibold-uppercase text-text-primary">{t('detail.conversationId', { ns: 'appLog' })}</div>
+          <div className="flex min-w-0 items-center system-2xs-regular-uppercase text-text-secondary">
+            {log && (
+              <>
+                <Tooltip>
+                  <TooltipTrigger
+                    render={(
+                      <div className="truncate">{log.conversation_id}</div>
+                    )}
+                  />
+                  <TooltipContent>
+                    {log.conversation_id}
+                  </TooltipContent>
+                </Tooltip>
+                <CopyIcon content={log.conversation_id} />
+              </>
+            )}
+          </div>
+        </div>
+        <div className="flex min-w-0 grow flex-wrap items-center justify-end gap-y-1">
+          <div className="min-w-0 truncate system-sm-medium text-text-secondary">
+            {log?.title || log?.conversation_id}
+          </div>
+        </div>
+        <ActionButton size="l" aria-label={t('operation.close', { ns: 'common' })} onClick={onClose}>
+          <span aria-hidden className="i-ri-close-line size-4 text-text-tertiary" />
+        </ActionButton>
+      </div>
+      <div className="shrink-0 px-1 pt-1">
+        <div className="rounded-t-xl bg-background-section-burn p-3 pb-2" />
+      </div>
+      <div className="mx-1 mb-1 grow overflow-auto rounded-b-xl bg-background-section-burn">
+        {messagesQuery.isPending && (
+          <div className="flex h-full items-center justify-center">
+            <Loading />
+          </div>
+        )}
+        {messagesQuery.isError && (
+          <div className="flex h-full items-center justify-center text-center system-sm-regular text-text-tertiary">
+            {t('agentDetail.logs.loadFailed', { ns: 'agentV2' })}
+          </div>
+        )}
+        {messagesQuery.isSuccess && chatList.length === 0 && (
+          <div className="flex h-full items-center justify-center text-center system-sm-regular text-text-tertiary">
+            {t('agentDetail.logs.empty', { ns: 'agentV2' })}
+          </div>
+        )}
+        {messagesQuery.isSuccess && chatList.length > 0 && (
+          <div className="mb-4 pt-4">
+            <Chat
+              chatList={chatList}
+              noChatInput
+              showPromptLog
+              hideProcessDetail
+              hideLogModal
+              chatContainerInnerClassName="px-3"
+              onFeedback={noop}
+            />
+          </div>
+        )}
+      </div>
+    </div>
+  )
+}
+
+function formatAgentLogMessages({
+  conversationId,
+  formatLogTime,
+  messages,
+}: {
+  conversationId: string
+  formatLogTime: (value: number) => string
+  messages: AgentLogMessageItemResponse[]
+}) {
+  const chatList: IChatItem[] = []
+
+  messages.forEach((message) => {
+    chatList.push({
+      id: `question-${message.id}`,
+      content: message.query,
+      isAnswer: false,
+      parentMessageId: undefined,
+    })
+    chatList.push({
+      id: message.id,
+      content: message.answer || message.error || '',
+      conversationId,
+      feedbackDisabled: true,
+      input: {
+        inputs: {
+          query: message.query,
+        },
+        query: message.query,
+      },
+      isAnswer: true,
+      log: [
+        { role: 'user', text: message.query },
+        { role: 'assistant', text: message.answer || message.error || '' },
+      ],
+      more: {
+        latency: message.latency.toFixed(2),
+        time: formatLogTime(message.created_at ?? message.updated_at ?? 0),
+        tokens: message.total_tokens,
+      },
+      parentMessageId: `question-${message.id}`,
+    })
+  })
+
+  return chatList
+}
diff --git a/web/features/agent-v2/agent-detail/logs/components/logs-table.tsx b/web/features/agent-v2/agent-detail/logs/components/logs-table.tsx
index fd8140be430..12e3a3b1385 100644
--- a/web/features/agent-v2/agent-detail/logs/components/logs-table.tsx
+++ b/web/features/agent-v2/agent-detail/logs/components/logs-table.tsx
@@ -18,12 +18,16 @@ export function AgentLogsTable({
   isPending,
   isError,
   isSuccess,
+  selectedLogId,
+  onOpenLog,
   onRetry,
 }: {
   logs: AgentLogConversationItemResponse[]
   isPending: boolean
   isError: boolean
   isSuccess: boolean
+  selectedLogId?: string
+  onOpenLog: (log: AgentLogConversationItemResponse) => void
   onRetry: () => void
 }) {
   const { t } = useTranslation('agentV2')
@@ -64,6 +68,8 @@ export function AgentLogsTable({
                 isPending={isPending}
                 isError={isError}
                 isSuccess={isSuccess}
+                selectedLogId={selectedLogId}
+                onOpenLog={onOpenLog}
                 onRetry={onRetry}
               />
             </table>
@@ -82,12 +88,16 @@ function AgentLogsTableBody({
   isPending,
   isError,
   isSuccess,
+  selectedLogId,
+  onOpenLog,
   onRetry,
 }: {
   logs: AgentLogConversationItemResponse[]
   isPending: boolean
   isError: boolean
   isSuccess: boolean
+  selectedLogId?: string
+  onOpenLog: (log: AgentLogConversationItemResponse) => void
   onRetry: () => void
 }) {
   const { t } = useTranslation('agentV2')
@@ -117,44 +127,63 @@ function AgentLogsTableBody({
           {t('agentDetail.logs.empty')}
         </LogsStateRow>
       )}
-      {isSuccess && logs.map(log => (
-        <tr
-          key={log.id}
-          className="h-10 border-b border-divider-subtle hover:bg-background-default-hover"
-        >
-          <td className="px-0">
-            <span className={cn(
-              'mx-auto block size-1.5 rounded-full',
-              log.unread ? 'bg-util-colors-blue-blue-500' : 'bg-transparent',
+      {isSuccess && logs.map((log) => {
+        const logTitle = log.title || log.conversation_id
+        const isSelected = selectedLogId === log.id
+
+        return (
+          <tr
+            key={log.id}
+            className={cn(
+              'h-10 cursor-pointer border-b border-divider-subtle hover:bg-background-default-hover',
+              isSelected && 'bg-background-default-hover',
             )}
-            />
-          </td>
-          <TableCell className="system-sm-medium text-text-secondary">
-            {log.title || notAvailable}
-          </TableCell>
-          <td className="px-3">
-            <LogSourceCell source={log.source} />
-          </td>
-          <TableCell translate="no">
-            {log.end_user_id || notAvailable}
-          </TableCell>
-          <TableCell className="tabular-nums">
-            {log.message_count}
-          </TableCell>
-          <TableCell className="text-text-quaternary">
-            {formatRate(log.user_rate, notAvailable)}
-          </TableCell>
-          <TableCell className="text-text-quaternary">
-            {formatRate(log.operation_rate, notAvailable)}
-          </TableCell>
-          <TableCell>
-            {formatLogTime(log.updated_at)}
-          </TableCell>
-          <TableCell>
-            {formatLogTime(log.created_at)}
-          </TableCell>
-        </tr>
-      ))}
+            onClick={() => onOpenLog(log)}
+          >
+            <td className="px-0">
+              <span className={cn(
+                'mx-auto block size-1.5 rounded-full',
+                log.unread ? 'bg-util-colors-blue-blue-500' : 'bg-transparent',
+              )}
+              />
+            </td>
+            <TableCell>
+              <button
+                type="button"
+                aria-label={logTitle}
+                className="block w-full truncate rounded-sm text-left system-sm-medium text-text-secondary focus-visible:ring-2 focus-visible:ring-state-accent-solid focus-visible:outline-hidden"
+                onClick={(event) => {
+                  event.stopPropagation()
+                  onOpenLog(log)
+                }}
+              >
+                {log.title || notAvailable}
+              </button>
+            </TableCell>
+            <td className="px-3">
+              <LogSourceCell source={log.source} />
+            </td>
+            <TableCell translate="no">
+              {log.end_user_id || notAvailable}
+            </TableCell>
+            <TableCell className="tabular-nums">
+              {log.message_count}
+            </TableCell>
+            <TableCell className="text-text-quaternary">
+              {formatRate(log.user_rate, notAvailable)}
+            </TableCell>
+            <TableCell className="text-text-quaternary">
+              {formatRate(log.operation_rate, notAvailable)}
+            </TableCell>
+            <TableCell>
+              {formatLogTime(log.updated_at)}
+            </TableCell>
+            <TableCell>
+              {formatLogTime(log.created_at)}
+            </TableCell>
+          </tr>
+        )
+      })}
     </tbody>
   )
 }
diff --git a/web/features/agent-v2/agent-detail/logs/page.tsx b/web/features/agent-v2/agent-detail/logs/page.tsx
index 311acb05912..60e13b8f8a9 100644
--- a/web/features/agent-v2/agent-detail/logs/page.tsx
+++ b/web/features/agent-v2/agent-detail/logs/page.tsx
@@ -1,6 +1,15 @@
 'use client'
 
+import type { AgentLogConversationItemResponse } from '@dify/contracts/api/console/agent/types.gen'
 import type { SourceFilterValue } from './components/source-picker'
+import {
+  Drawer,
+  DrawerBackdrop,
+  DrawerContent,
+  DrawerPopup,
+  DrawerPortal,
+  DrawerViewport,
+} from '@langgenius/dify-ui/drawer'
 import { Pagination } from '@langgenius/dify-ui/pagination'
 import { keepPreviousData, useQuery } from '@tanstack/react-query'
 import dayjs from 'dayjs'
@@ -10,7 +19,10 @@ import Chip from '@/app/components/base/chip'
 import { SearchInput } from '@/app/components/base/search-input'
 import Sort from '@/app/components/base/sort'
 import { useDocLink } from '@/context/i18n'
+import useBreakpoints, { MediaType } from '@/hooks/use-breakpoints'
 import { consoleQuery } from '@/service/client'
+import { AgentDetailSectionSurface } from '../section-surface'
+import { AgentLogDetailPanel } from './components/log-detail-panel'
 import { AgentLogsTable } from './components/logs-table'
 import { AgentLogSourcePicker } from './components/source-picker'
 
@@ -63,6 +75,8 @@ export function AgentLogsPage({
   const { t } = useTranslation('agentV2')
   const { t: tCommon } = useTranslation('common')
   const docLink = useDocLink()
+  const media = useBreakpoints()
+  const isMobile = media === MediaType.mobile
   const [period, setPeriod] = useState<PeriodKey>('last7days')
   const [source, setSource] = useState<SourceFilterValue>([])
   const [keyword, setKeyword] = useState('')
@@ -72,6 +86,7 @@ export function AgentLogsPage({
   })
   const [page, setPage] = useState(1)
   const [limit, setLimit] = useState(25)
+  const [selectedLog, setSelectedLog] = useState<AgentLogConversationItemResponse>()
   const periodItems = periodOptions.map(option => ({
     value: option.value,
     name: t(option.labelKey),
@@ -105,12 +120,13 @@ export function AgentLogsPage({
   const logs = logsQuery.data?.data ?? []
   const totalPages = Math.max(Math.ceil((logsQuery.data?.total ?? 0) / limit), 1)
   const currentPage = logsQuery.data?.page ?? page
+  const closeLogDetail = () => {
+    setSelectedLog(undefined)
+    void logsQuery.refetch()
+  }
 
   return (
-    <section
-      aria-label={t('agentDetail.sections.logs')}
-      className="flex h-full min-w-0 flex-1 flex-col overflow-hidden bg-components-panel-bg-blur"
-    >
+    <AgentDetailSectionSurface label={t('agentDetail.sections.logs')}>
       <header className="h-26.5 shrink-0 px-6 pt-3 pb-2">
         <div className="min-w-0">
           <h2 className="system-xl-semibold text-text-primary">
@@ -194,12 +210,39 @@ export function AgentLogsPage({
           isPending={logsQuery.isPending}
           isError={logsQuery.isError}
           isSuccess={logsQuery.isSuccess}
+          selectedLogId={selectedLog?.id}
+          onOpenLog={setSelectedLog}
           onRetry={() => {
             void logsQuery.refetch()
           }}
         />
       </div>
 
+      <Drawer
+        open={!!selectedLog}
+        modal
+        swipeDirection="right"
+        onOpenChange={(open) => {
+          if (!open)
+            closeLogDetail()
+        }}
+      >
+        <DrawerPortal>
+          <DrawerBackdrop className={!isMobile ? 'bg-transparent' : undefined} />
+          <DrawerViewport>
+            <DrawerPopup className="p-0! data-[swipe-direction=right]:top-16 data-[swipe-direction=right]:right-2 data-[swipe-direction=right]:bottom-3 data-[swipe-direction=right]:h-auto data-[swipe-direction=right]:w-full data-[swipe-direction=right]:max-w-150 data-[swipe-direction=right]:rounded-xl data-[swipe-direction=right]:border data-[swipe-direction=right]:border-components-panel-border">
+              <DrawerContent className="flex min-h-0 flex-1 flex-col p-0 pb-0">
+                <AgentLogDetailPanel
+                  agentId={agentId}
+                  log={selectedLog}
+                  onClose={closeLogDetail}
+                />
+              </DrawerContent>
+            </DrawerPopup>
+          </DrawerViewport>
+        </DrawerPortal>
+      </Drawer>
+
       <Pagination
         page={currentPage}
         totalPages={totalPages}
@@ -222,6 +265,6 @@ export function AgentLogsPage({
           ariaLabel: tCommon('pagination.perPage'),
         }}
       />
-    </section>
+    </AgentDetailSectionSurface>
   )
 }
diff --git a/web/features/agent-v2/agent-detail/monitoring/page.tsx b/web/features/agent-v2/agent-detail/monitoring/page.tsx
index 791f456804b..d6087336de8 100644
--- a/web/features/agent-v2/agent-detail/monitoring/page.tsx
+++ b/web/features/agent-v2/agent-detail/monitoring/page.tsx
@@ -18,6 +18,7 @@ import { useState } from 'react'
 import { useTranslation } from 'react-i18next'
 import { useDocLink } from '@/context/i18n'
 import { consoleQuery } from '@/service/client'
+import { AgentDetailSectionSurface } from '../section-surface'
 import { AgentMonitoringChart } from './chart'
 import { getAgentMonitoringMetrics } from './metrics'
 import { AgentMonitoringTimeRangePicker } from './time-range-picker'
@@ -91,10 +92,7 @@ export function AgentMonitoringPage({
   const shouldShowError = statisticsQuery.isError && !statisticsQuery.data
 
   return (
-    <section
-      aria-label={t('agentDetail.sections.monitoring')}
-      className="flex h-full min-w-0 flex-1 flex-col overflow-hidden bg-components-panel-bg-blur"
-    >
+    <AgentDetailSectionSurface label={t('agentDetail.sections.monitoring')}>
       <header className="h-26.5 shrink-0 px-6 pt-3 pb-2">
         <div className="min-w-0">
           <h2 className="system-xl-semibold text-text-primary">
@@ -175,7 +173,7 @@ export function AgentMonitoringPage({
           </div>
         )}
       </ScrollArea>
-    </section>
+    </AgentDetailSectionSurface>
   )
 }
 
diff --git a/web/features/agent-v2/agent-detail/section-surface.tsx b/web/features/agent-v2/agent-detail/section-surface.tsx
new file mode 100644
index 00000000000..229512daaa9
--- /dev/null
+++ b/web/features/agent-v2/agent-detail/section-surface.tsx
@@ -0,0 +1,29 @@
+'use client'
+
+import type { ReactNode } from 'react'
+import { cn } from '@langgenius/dify-ui/cn'
+
+type AgentDetailSectionSurfaceProps = {
+  children: ReactNode
+  className?: string
+  label: string
+  panelClassName?: string
+}
+
+export function AgentDetailSectionSurface({
+  children,
+  className,
+  label,
+  panelClassName,
+}: AgentDetailSectionSurfaceProps) {
+  return (
+    <section
+      aria-label={label}
+      className={cn('flex h-full min-w-0 flex-1 overflow-hidden py-1 pr-1 pl-0', className)}
+    >
+      <div className={cn('flex min-w-0 flex-1 flex-col overflow-hidden rounded-lg border-[0.5px] border-components-panel-border bg-components-panel-bg shadow-xl shadow-shadow-shadow-5', panelClassName)}>
+        {children}
+      </div>
+    </section>
+  )
+}
diff --git a/web/features/agent-v2/roster/components/__tests__/agent-roster-list.spec.tsx b/web/features/agent-v2/roster/components/__tests__/agent-roster-list.spec.tsx
index 43700ac1bd7..08af8f8fbf3 100644
--- a/web/features/agent-v2/roster/components/__tests__/agent-roster-list.spec.tsx
+++ b/web/features/agent-v2/roster/components/__tests__/agent-roster-list.spec.tsx
@@ -1,5 +1,5 @@
+import type { AgentAppPartial } from '@dify/contracts/api/console/agent/types.gen'
 import type { ComponentProps } from 'react'
-import type { AgentRosterListItem } from '../agent-roster-list'
 import { toast } from '@langgenius/dify-ui/toast'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
 import { render, screen, waitFor, within } from '@testing-library/react'
@@ -45,7 +45,7 @@ vi.mock('@/service/client', () => ({
   },
 }))
 
-const createAgent = (overrides: Partial<AgentRosterListItem> = {}): AgentRosterListItem => ({
+const createAgent = (overrides: Partial<AgentAppPartial> = {}): AgentAppPartial => ({
   active_config_is_published: false,
   description: 'Find and summarize market materials.',
   id: 'agent-1',
@@ -62,7 +62,7 @@ const createAgent = (overrides: Partial<AgentRosterListItem> = {}): AgentRosterL
 })
 
 const renderList = (
-  agents: AgentRosterListItem[],
+  agents: AgentAppPartial[],
   overrides: Partial<ComponentProps<typeof AgentRosterList>> = {},
 ) => {
   const queryClient = new QueryClient()
@@ -211,10 +211,21 @@ describe('AgentRosterList', () => {
     await user.click(screen.getByRole('menuitem', { name: /common\.operation\.duplicate/ }))
 
     const dialog = await screen.findByRole('dialog', { name: 'agentV2.roster.duplicateDialog.title' })
-    expect(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.nameLabel/ })).toHaveValue('')
-    expect(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.nameLabel/ })).toHaveAttribute('placeholder', 'Research Agent copy')
-    expect(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ })).toHaveValue('Research Assistant')
-    expect(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.descriptionLabel/ })).toHaveValue('Find and summarize market materials.')
+    const nameInput = within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.nameLabel.*common\.label\.optional/,
+    })
+    const roleInput = within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.roleLabel.*common\.label\.optional/,
+    })
+    const descriptionInput = within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.descriptionLabel.*common\.label\.optional/,
+    })
+    expect(nameInput).toHaveValue('')
+    expect(nameInput).toHaveAttribute('placeholder', 'Research Agent copy')
+    expect(roleInput).toHaveValue('Research Assistant')
+    expect(roleInput).not.toBeRequired()
+    expect(descriptionInput).toHaveValue('Find and summarize market materials.')
+    expect(descriptionInput).not.toBeRequired()
     expect(duplicateAgentMutationFn).not.toHaveBeenCalled()
   })
 
@@ -315,7 +326,7 @@ describe('AgentRosterList', () => {
     })
   })
 
-  it('shows a field error when duplicating with an empty role', async () => {
+  it('duplicates an agent with an empty role when the role is cleared', async () => {
     const user = userEvent.setup()
     renderList([createAgent()])
 
@@ -326,8 +337,23 @@ describe('AgentRosterList', () => {
     await user.clear(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ }))
     await user.click(within(dialog).getByRole('button', { name: 'common.operation.duplicate' }))
 
-    expect(await within(dialog).findByText('agentV2.roster.createForm.roleRequired')).toBeInTheDocument()
-    expect(duplicateAgentMutationFn).not.toHaveBeenCalled()
+    expect(duplicateAgentMutationFn).toHaveBeenCalledWith(
+      {
+        params: {
+          agent_id: 'agent-1',
+        },
+        body: {
+          description: 'Find and summarize market materials.',
+          role: '',
+          icon: '🧸',
+          icon_background: '#F5F3FF',
+          icon_type: 'emoji',
+        },
+      },
+      expect.objectContaining({
+        client: expect.any(QueryClient),
+      }),
+    )
   })
 
   it('resets the edit form draft when reopening after canceling unsaved changes', async () => {
@@ -352,7 +378,7 @@ describe('AgentRosterList', () => {
 
     const reopenedDialog = await screen.findByRole('dialog', { name: 'agentV2.roster.editDialog.title' })
     expect(within(reopenedDialog).getByRole('textbox', { name: 'agentV2.roster.createForm.nameLabel' })).toHaveValue('Research Agent')
-    expect(within(reopenedDialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' })).toHaveValue('Research Assistant')
+    expect(within(reopenedDialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ })).toHaveValue('Research Assistant')
     expect(within(reopenedDialog).getByRole('button', { name: 'common.operation.save' })).toBeDisabled()
   })
 })
diff --git a/web/features/agent-v2/roster/components/__tests__/create-agent-dialog.spec.tsx b/web/features/agent-v2/roster/components/__tests__/create-agent-dialog.spec.tsx
index 3c06e441635..54ebe62467c 100644
--- a/web/features/agent-v2/roster/components/__tests__/create-agent-dialog.spec.tsx
+++ b/web/features/agent-v2/roster/components/__tests__/create-agent-dialog.spec.tsx
@@ -55,7 +55,7 @@ describe('CreateAgentDialog', () => {
 
     const dialog = await screen.findByRole('dialog', { name: 'agentV2.roster.createDialog.title' })
     await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.nameLabel' }), ' Research Agent ')
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }), ' Research Assistant ')
+    await user.type(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ }), ' Research Assistant ')
     await user.type(within(dialog).getByPlaceholderText('agentV2.roster.createForm.descriptionPlaceholder'), ' Find and summarize market materials. ')
     await user.click(within(dialog).getByRole('button', { name: 'common.operation.create' }))
 
@@ -83,7 +83,6 @@ describe('CreateAgentDialog', () => {
 
     const dialog = await screen.findByRole('dialog', { name: 'agentV2.roster.createDialog.title' })
     await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.nameLabel' }), 'Research Agent')
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }), 'Research Assistant')
     await user.click(within(dialog).getByRole('button', { name: 'common.operation.create' }))
 
     const mutationOptions = mutationMock.mutate.mock.calls[0]?.[1]
@@ -102,7 +101,7 @@ describe('CreateAgentDialog', () => {
     await user.click(screen.getByRole('button', { name: /agentV2\.roster\.createAgent/ }))
 
     const dialog = await screen.findByRole('dialog', { name: 'agentV2.roster.createDialog.title' })
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }), 'Research Assistant')
+    await user.type(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ }), 'Research Assistant')
     await user.click(within(dialog).getByRole('button', { name: 'common.operation.create' }))
 
     expect(await within(dialog).findByText('agentV2.roster.createForm.nameRequired')).toBeInTheDocument()
@@ -110,22 +109,23 @@ describe('CreateAgentDialog', () => {
     expect(mutationMock.mutate).not.toHaveBeenCalled()
   })
 
-  it('shows a field error when creating with an empty role', async () => {
+  it('marks role and description as optional', async () => {
     const user = userEvent.setup()
     render(<CreateAgentDialog />)
 
     await user.click(screen.getByRole('button', { name: /agentV2\.roster\.createAgent/ }))
 
     const dialog = await screen.findByRole('dialog', { name: 'agentV2.roster.createDialog.title' })
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.nameLabel' }), 'Research Agent')
-    await user.click(within(dialog).getByRole('button', { name: 'common.operation.create' }))
 
-    expect(await within(dialog).findByText('agentV2.roster.createForm.roleRequired')).toBeInTheDocument()
-    expect(toastMock.error).not.toHaveBeenCalled()
-    expect(mutationMock.mutate).not.toHaveBeenCalled()
+    expect(within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.roleLabel.*common\.label\.optional/,
+    })).not.toBeRequired()
+    expect(within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.descriptionLabel.*common\.label\.optional/,
+    })).not.toBeRequired()
   })
 
-  it('shows a field error when creating with a blank role', async () => {
+  it('submits an empty role when role is left blank', async () => {
     const user = userEvent.setup()
     render(<CreateAgentDialog />)
 
@@ -133,12 +133,20 @@ describe('CreateAgentDialog', () => {
 
     const dialog = await screen.findByRole('dialog', { name: 'agentV2.roster.createDialog.title' })
     await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.nameLabel' }), 'Research Agent')
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }), '   ')
     await user.click(within(dialog).getByRole('button', { name: 'common.operation.create' }))
 
-    expect(await within(dialog).findByText('agentV2.roster.createForm.roleRequired')).toBeInTheDocument()
-    expect(toastMock.error).not.toHaveBeenCalled()
-    expect(mutationMock.mutate).not.toHaveBeenCalled()
+    expect(mutationMock.mutate).toHaveBeenCalledWith({
+      body: {
+        name: 'Research Agent',
+        description: '',
+        role: '',
+        icon_type: 'emoji',
+        icon: '🧸',
+        icon_background: '#F5F3FF',
+      },
+    }, expect.objectContaining({
+      onSuccess: expect.any(Function),
+    }))
   })
 
   it('keeps the form open when the backdrop is clicked', async () => {
diff --git a/web/features/agent-v2/roster/components/__tests__/edit-agent-dialog.spec.tsx b/web/features/agent-v2/roster/components/__tests__/edit-agent-dialog.spec.tsx
index eca96683382..84c6c4c0bec 100644
--- a/web/features/agent-v2/roster/components/__tests__/edit-agent-dialog.spec.tsx
+++ b/web/features/agent-v2/roster/components/__tests__/edit-agent-dialog.spec.tsx
@@ -123,8 +123,8 @@ describe('EditAgentDialog', () => {
     renderDialog()
 
     const dialog = screen.getByRole('dialog', { name: 'agentV2.roster.editDialog.title' })
-    await user.clear(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }))
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }), ' Market Analyst ')
+    await user.clear(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ }))
+    await user.type(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ }), ' Market Analyst ')
     await user.click(within(dialog).getByRole('button', { name: 'common.operation.save' }))
 
     expect(mutationMock.mutate).toHaveBeenCalledWith({
@@ -190,37 +190,45 @@ describe('EditAgentDialog', () => {
     expect(mutationMock.mutate).not.toHaveBeenCalled()
   })
 
-  it('shows a field error when saving with an empty role', async () => {
-    const user = userEvent.setup()
+  it('marks role and description as optional', () => {
     renderDialog()
 
     const dialog = screen.getByRole('dialog', { name: 'agentV2.roster.editDialog.title' })
-    await user.clear(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }))
 
-    const saveButton = within(dialog).getByRole('button', { name: 'common.operation.save' })
-    expect(saveButton).not.toBeDisabled()
-    await user.click(saveButton)
-
-    expect(await within(dialog).findByText('agentV2.roster.createForm.roleRequired')).toBeInTheDocument()
-    expect(toastMock.error).not.toHaveBeenCalled()
-    expect(mutationMock.mutate).not.toHaveBeenCalled()
+    expect(within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.roleLabel.*common\.label\.optional/,
+    })).not.toBeRequired()
+    expect(within(dialog).getByRole('textbox', {
+      name: /agentV2\.roster\.createForm\.descriptionLabel.*common\.label\.optional/,
+    })).not.toBeRequired()
   })
 
-  it('shows a field error when saving with a blank role', async () => {
+  it('submits an empty role when the role is cleared', async () => {
     const user = userEvent.setup()
     renderDialog()
 
     const dialog = screen.getByRole('dialog', { name: 'agentV2.roster.editDialog.title' })
-    await user.clear(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }))
-    await user.type(within(dialog).getByRole('textbox', { name: 'agentV2.roster.createForm.roleLabel' }), '   ')
+    await user.clear(within(dialog).getByRole('textbox', { name: /agentV2\.roster\.createForm\.roleLabel/ }))
 
     const saveButton = within(dialog).getByRole('button', { name: 'common.operation.save' })
     expect(saveButton).not.toBeDisabled()
     await user.click(saveButton)
 
-    expect(await within(dialog).findByText('agentV2.roster.createForm.roleRequired')).toBeInTheDocument()
-    expect(toastMock.error).not.toHaveBeenCalled()
-    expect(mutationMock.mutate).not.toHaveBeenCalled()
+    expect(mutationMock.mutate).toHaveBeenCalledWith({
+      params: {
+        agent_id: 'agent-1',
+      },
+      body: {
+        name: 'Research Agent',
+        description: 'Find and summarize market materials.',
+        role: '',
+        icon_type: 'emoji',
+        icon: '🧸',
+        icon_background: '#F5F3FF',
+      },
+    }, expect.objectContaining({
+      onSuccess: expect.any(Function),
+    }))
   })
 
   it('keeps the form open when the backdrop is clicked', async () => {
diff --git a/web/features/agent-v2/roster/components/__tests__/roster-toolbar.spec.tsx b/web/features/agent-v2/roster/components/__tests__/roster-toolbar.spec.tsx
index 7756f2cc23e..33763116c92 100644
--- a/web/features/agent-v2/roster/components/__tests__/roster-toolbar.spec.tsx
+++ b/web/features/agent-v2/roster/components/__tests__/roster-toolbar.spec.tsx
@@ -1,7 +1,7 @@
-import type { RosterFilterValue } from '../roster-filter'
 import { QueryClient, QueryClientProvider } from '@tanstack/react-query'
-import { render, screen, within } from '@testing-library/react'
+import { screen, within } from '@testing-library/react'
 import userEvent from '@testing-library/user-event'
+import { renderWithNuqs } from '@/test/nuqs-testing'
 import { RosterToolbar } from '../roster-toolbar'
 
 vi.mock('@/next/navigation', () => ({
@@ -11,34 +11,27 @@ vi.mock('@/next/navigation', () => ({
 }))
 
 const renderToolbar = ({
-  filter = 'all',
-  onFilterChange = vi.fn(),
+  searchParams = '',
 }: {
-  filter?: RosterFilterValue
-  onFilterChange?: (value: RosterFilterValue) => void
+  searchParams?: string
 } = {}) => {
   const queryClient = new QueryClient()
 
-  render(
+  return renderWithNuqs(
     <QueryClientProvider client={queryClient}>
       <RosterToolbar
         draftAgents={2}
-        filter={filter}
-        keyword=""
-        onFilterChange={onFilterChange}
-        onKeywordChange={vi.fn()}
         publishedAgents={1}
       />
     </QueryClientProvider>,
+    { searchParams },
   )
-
-  return { onFilterChange }
 }
 
 describe('RosterToolbar', () => {
   it('enables roster filters and emits the selected filter', async () => {
     const user = userEvent.setup()
-    const { onFilterChange } = renderToolbar()
+    const { onUrlUpdate } = renderToolbar()
 
     const publishedFilter = screen.getByRole('button', { name: /agentV2\.roster\.filters\.published/ })
     const draftsFilter = screen.getByRole('button', { name: /agentV2\.roster\.filters\.drafts/ })
@@ -48,7 +41,9 @@ describe('RosterToolbar', () => {
 
     await user.click(publishedFilter)
 
-    expect(onFilterChange).toHaveBeenCalledWith('published')
+    expect(onUrlUpdate).toHaveBeenCalledWith(expect.objectContaining({
+      queryString: '?filter=published',
+    }))
   })
 
   it('renders stable filter count badges and omits the all count', () => {
@@ -62,4 +57,35 @@ describe('RosterToolbar', () => {
     expect(within(publishedFilter).getByText('1')).toBeInTheDocument()
     expect(within(draftsFilter).getByText('2')).toBeInTheDocument()
   })
+
+  it('renders created-by-me filtering and emits checked state', async () => {
+    const user = userEvent.setup()
+    const { onUrlUpdate } = renderToolbar()
+
+    const createdByMeFilter = screen.getByRole('checkbox', { name: 'agentV2.roster.filters.createdByMe' })
+
+    expect(createdByMeFilter).toHaveAttribute('aria-checked', 'false')
+
+    await user.click(createdByMeFilter)
+
+    expect(onUrlUpdate).toHaveBeenCalledWith(expect.objectContaining({
+      queryString: '?created_by_me=true',
+    }))
+  })
+
+  it('renders sort options and emits the selected sort strategy', async () => {
+    const user = userEvent.setup()
+    const { onUrlUpdate } = renderToolbar()
+
+    const sortSelect = screen.getByRole('combobox', { name: 'agentV2.roster.sort.label' })
+
+    expect(screen.getByText('agentV2.roster.sort.lastModified')).toBeInTheDocument()
+
+    await user.click(sortSelect)
+    await user.click(await screen.findByRole('option', { name: 'agentV2.roster.sort.recentlyCreated' }))
+
+    expect(onUrlUpdate).toHaveBeenCalledWith(expect.objectContaining({
+      queryString: '?sort_by=recently_created',
+    }))
+  })
 })
diff --git a/web/features/agent-v2/roster/components/agent-form-fields.tsx b/web/features/agent-v2/roster/components/agent-form-fields.tsx
index c3f1fe7b025..310616385a3 100644
--- a/web/features/agent-v2/roster/components/agent-form-fields.tsx
+++ b/web/features/agent-v2/roster/components/agent-form-fields.tsx
@@ -28,6 +28,7 @@ export function AgentFormFields({
   role,
 }: AgentFormFieldsProps) {
   const { t } = useTranslation('agentV2')
+  const { t: tCommon } = useTranslation('common')
 
   return (
     <div className="space-y-5 px-6 py-3">
@@ -80,34 +81,29 @@ export function AgentFormFields({
           <FieldRoot
             name="role"
             className="relative min-w-0 flex-1"
-            validate={(value) => {
-              if (typeof value === 'string' && value.length > 0 && !value.trim())
-                return t('roster.createForm.roleRequired')
-
-              return null
-            }}
           >
             <FieldLabel>
               {t('roster.createForm.roleLabel')}
+              <span className="ml-1 system-xs-regular text-text-tertiary">
+                {tCommon('label.optional')}
+              </span>
             </FieldLabel>
             <FieldControl
               autoComplete="off"
               maxLength={255}
               onValueChange={onRoleChange}
               placeholder={t('roster.createForm.rolePlaceholder')}
-              required
               value={role}
             />
-            <div className="absolute top-full left-0 mt-1">
-              <FieldError match="valueMissing">{t('roster.createForm.roleRequired')}</FieldError>
-              <FieldError match="customError" />
-            </div>
           </FieldRoot>
         </div>
       </div>
       <FieldRoot name="description">
         <FieldLabel>
           {t('roster.createForm.descriptionLabel')}
+          <span className="ml-1 system-xs-regular text-text-tertiary">
+            {tCommon('label.optional')}
+          </span>
         </FieldLabel>
         <Textarea
           autoComplete="off"
diff --git a/web/features/agent-v2/roster/components/agent-roster-list.tsx b/web/features/agent-v2/roster/components/agent-roster-list.tsx
index 8ef3689daa4..683ee8d948c 100644
--- a/web/features/agent-v2/roster/components/agent-roster-list.tsx
+++ b/web/features/agent-v2/roster/components/agent-roster-list.tsx
@@ -21,7 +21,7 @@ import { DuplicateAgentDialog } from './duplicate-agent-dialog'
 import { EditAgentDialog } from './edit-agent-dialog'
 
 type AgentRosterListProps = {
-  agents: AgentRosterListItem[]
+  agents: AgentAppPartial[]
   hasMore: boolean
   isEmptySearch: boolean
   isError: boolean
@@ -32,8 +32,6 @@ type AgentRosterListProps = {
   onLoadMore: () => void
 }
 
-export type AgentRosterListItem = AgentAppPartial
-
 const skeletonRows = ['primary', 'secondary', 'tertiary'] as const
 const emptyPlaceholderCardIds = Array.from({ length: 16 }, (_, index) => `agent-roster-placeholder-card-${index}`)
 
@@ -96,7 +94,7 @@ function AgentRosterPlaceholderState({ title }: { title: string }) {
 function AgentRosterItem({
   agent,
 }: {
-  agent: AgentRosterListItem
+  agent: AgentAppPartial
 }) {
   const { t } = useTranslation('agentV2')
   const { t: tCommon } = useTranslation('common')
@@ -129,13 +127,13 @@ function AgentRosterItem({
   }
 
   return (
-    <article className="group relative col-span-1 h-36.5 min-w-0 overflow-hidden rounded-xl border-[0.5px] border-solid border-components-card-border bg-components-card-bg shadow-xs shadow-shadow-shadow-3 transition-shadow duration-200 ease-in-out hover:shadow-lg">
+    <article className="group relative col-span-1 h-36.5 min-w-0 overflow-hidden rounded-xl border-[0.5px] border-solid border-components-card-border bg-components-card-bg shadow-xs shadow-shadow-shadow-3 transition-shadow duration-200 ease-in-out after:pointer-events-none after:absolute after:inset-0 after:rounded-xl after:content-[''] hover:shadow-lg has-[>div>a:focus-visible]:after:inset-ring-2 has-[>div>a:focus-visible]:after:inset-ring-state-accent-solid">
       <div className="flex h-full min-w-0 flex-col">
         <Link
           href={`/roster/agent/${agent.id}/configure`}
           aria-labelledby={nameId}
           aria-describedby={agent.description ? descriptionId : undefined}
-          className="relative block shrink-0 cursor-pointer touch-manipulation rounded-xl outline-hidden after:pointer-events-none after:absolute after:inset-0 after:rounded-xl after:content-[''] focus-visible:after:inset-ring-2 focus-visible:after:inset-ring-state-accent-solid"
+          className="block shrink-0 cursor-pointer touch-manipulation outline-hidden"
         >
           <div className="flex items-center gap-3 pt-3.5 pr-4 pb-2 pl-3.5">
             <span aria-hidden className="shrink-0">
diff --git a/web/features/agent-v2/roster/components/duplicate-agent-dialog.tsx b/web/features/agent-v2/roster/components/duplicate-agent-dialog.tsx
index 4a5efae43a4..3b3b58c7ecb 100644
--- a/web/features/agent-v2/roster/components/duplicate-agent-dialog.tsx
+++ b/web/features/agent-v2/roster/components/duplicate-agent-dialog.tsx
@@ -4,7 +4,7 @@ import type { AgentAppCopyPayload, AgentAppPartial } from '@dify/contracts/api/c
 import type { AgentFormValues, AgentIconSelection } from './agent-form'
 import { Button } from '@langgenius/dify-ui/button'
 import { Dialog, DialogCloseButton, DialogContent, DialogDescription, DialogTitle } from '@langgenius/dify-ui/dialog'
-import { FieldControl, FieldError, FieldLabel, FieldRoot } from '@langgenius/dify-ui/field'
+import { FieldControl, FieldLabel, FieldRoot } from '@langgenius/dify-ui/field'
 import { Form } from '@langgenius/dify-ui/form'
 import { Textarea } from '@langgenius/dify-ui/textarea'
 import { toast } from '@langgenius/dify-ui/toast'
@@ -167,34 +167,29 @@ export function DuplicateAgentDialog({
                   <FieldRoot
                     name="role"
                     className="relative min-w-0 flex-1"
-                    validate={(value) => {
-                      if (typeof value === 'string' && value.length > 0 && !value.trim())
-                        return t('roster.createForm.roleRequired')
-
-                      return null
-                    }}
                   >
                     <FieldLabel>
                       {t('roster.createForm.roleLabel')}
+                      <span className="ml-1 system-xs-regular text-text-tertiary">
+                        {tCommon('label.optional')}
+                      </span>
                     </FieldLabel>
                     <FieldControl
                       autoComplete="off"
                       maxLength={255}
                       onValueChange={setRole}
                       placeholder={t('roster.createForm.rolePlaceholder')}
-                      required
                       value={role}
                     />
-                    <div className="absolute top-full left-0 mt-1">
-                      <FieldError match="valueMissing">{t('roster.createForm.roleRequired')}</FieldError>
-                      <FieldError match="customError" />
-                    </div>
                   </FieldRoot>
                 </div>
               </div>
               <FieldRoot name="description">
                 <FieldLabel>
                   {t('roster.createForm.descriptionLabel')}
+                  <span className="ml-1 system-xs-regular text-text-tertiary">
+                    {tCommon('label.optional')}
+                  </span>
                 </FieldLabel>
                 <Textarea
                   autoComplete="off"
diff --git a/web/features/agent-v2/roster/components/roster-sort-select.tsx b/web/features/agent-v2/roster/components/roster-sort-select.tsx
new file mode 100644
index 00000000000..95d85e717cb
--- /dev/null
+++ b/web/features/agent-v2/roster/components/roster-sort-select.tsx
@@ -0,0 +1,60 @@
+'use client'
+
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectItemIndicator,
+  SelectItemText,
+  SelectTrigger,
+} from '@langgenius/dify-ui/select'
+import { useQueryState } from 'nuqs'
+import { useTranslation } from 'react-i18next'
+import { rosterQueryParamNames, rosterSortByQueryParser } from '../query-params'
+import { DEFAULT_ROSTER_SORT_BY, rosterSortOptions } from './roster-sort'
+
+export function RosterSortSelect() {
+  const { t } = useTranslation('agentV2')
+  const [value, setValue] = useQueryState(rosterQueryParamNames.sortBy, rosterSortByQueryParser)
+  const selectedOption = rosterSortOptions.find(option => option.value === value)
+    ?? rosterSortOptions.find(option => option.value === DEFAULT_ROSTER_SORT_BY)!
+
+  return (
+    <Select
+      value={value}
+      onValueChange={(nextValue) => {
+        void setValue(nextValue)
+      }}
+    >
+      <SelectTrigger
+        aria-label={t('roster.sort.label')}
+        className="h-8 w-fit max-w-45 min-w-0 gap-0 py-1 pr-2.5 pl-2"
+      >
+        <span className="flex min-w-0 items-center gap-1 px-1">
+          <span className="shrink-0 system-sm-regular text-text-tertiary">
+            {t('roster.sort.label')}
+          </span>
+          <span className="truncate system-sm-medium text-text-secondary">
+            {t(selectedOption.labelKey)}
+          </span>
+        </span>
+      </SelectTrigger>
+      <SelectContent
+        placement="bottom-start"
+        sideOffset={4}
+        popupClassName="w-60"
+        listProps={{ 'aria-label': t('roster.sort.optionsLabel') }}
+      >
+        {rosterSortOptions.map(option => (
+          <SelectItem
+            key={option.value}
+            value={option.value}
+          >
+            <SelectItemText title={t(option.labelKey)}>{t(option.labelKey)}</SelectItemText>
+            <SelectItemIndicator />
+          </SelectItem>
+        ))}
+      </SelectContent>
+    </Select>
+  )
+}
diff --git a/web/features/agent-v2/roster/components/roster-sort.ts b/web/features/agent-v2/roster/components/roster-sort.ts
new file mode 100644
index 00000000000..7ee4604a0a7
--- /dev/null
+++ b/web/features/agent-v2/roster/components/roster-sort.ts
@@ -0,0 +1,20 @@
+import type { GetAgentData } from '@dify/contracts/api/console/agent/types.gen'
+
+export type RosterSortBy = NonNullable<NonNullable<GetAgentData['query']>['sort_by']>
+
+export const DEFAULT_ROSTER_SORT_BY = 'last_modified' satisfies RosterSortBy
+
+export const ROSTER_SORT_BY_VALUES = [
+  DEFAULT_ROSTER_SORT_BY,
+  'recently_created',
+  'earliest_created',
+] as const satisfies readonly RosterSortBy[]
+
+export const rosterSortOptions: Array<{
+  value: RosterSortBy
+  labelKey: 'roster.sort.lastModified' | 'roster.sort.recentlyCreated' | 'roster.sort.earliestCreated'
+}> = [
+  { value: 'last_modified', labelKey: 'roster.sort.lastModified' },
+  { value: 'recently_created', labelKey: 'roster.sort.recentlyCreated' },
+  { value: 'earliest_created', labelKey: 'roster.sort.earliestCreated' },
+]
diff --git a/web/features/agent-v2/roster/components/roster-toolbar.tsx b/web/features/agent-v2/roster/components/roster-toolbar.tsx
index 690d3373c30..24a3282fcc7 100644
--- a/web/features/agent-v2/roster/components/roster-toolbar.tsx
+++ b/web/features/agent-v2/roster/components/roster-toolbar.tsx
@@ -1,17 +1,22 @@
 'use client'
 
 import type { RosterFilterValue } from './roster-filter'
+import { Checkbox } from '@langgenius/dify-ui/checkbox'
 import { SegmentedControl, SegmentedControlItem } from '@langgenius/dify-ui/segmented-control'
+import { useQueryState } from 'nuqs'
 import { useTranslation } from 'react-i18next'
 import { SearchInput } from '@/app/components/base/search-input'
+import {
+  rosterCreatedByMeQueryParser,
+  rosterFilterQueryParser,
+  rosterKeywordQueryParser,
+  rosterQueryParamNames,
+} from '../query-params'
 import { CreateAgentDialog } from './create-agent-dialog'
+import { RosterSortSelect } from './roster-sort-select'
 
 type RosterToolbarProps = {
   draftAgents: number
-  filter: RosterFilterValue
-  keyword: string
-  onFilterChange: (value: RosterFilterValue) => void
-  onKeywordChange: (value: string) => void
   publishedAgents: number
 }
 
@@ -43,51 +48,94 @@ function RosterFilterItem({
   )
 }
 
-export function RosterToolbar({
+function RosterStatusFilter({
   draftAgents,
-  filter,
-  keyword,
-  onFilterChange,
-  onKeywordChange,
   publishedAgents,
 }: RosterToolbarProps) {
   const { t } = useTranslation('agentV2')
+  const [filter, setFilter] = useQueryState(rosterQueryParamNames.filter, rosterFilterQueryParser)
 
   return (
-    <div className="flex min-w-0 flex-wrap items-center gap-2">
-      <SegmentedControl
-        aria-label={t('roster.filters.label')}
-        value={[filter]}
-        onValueChange={(value) => {
-          const nextFilter = value[0]
+    <SegmentedControl
+      aria-label={t('roster.filters.label')}
+      className="shrink-0"
+      value={[filter]}
+      onValueChange={(value) => {
+        const nextFilter = value[0]
 
-          if (nextFilter)
-            onFilterChange(nextFilter)
-        }}
-      >
-        <RosterFilterItem
-          value="all"
-          label={t('roster.filters.all')}
-        />
-        <RosterFilterItem
-          value="published"
-          label={t('roster.filters.published')}
-          count={publishedAgents}
-        />
-        <RosterFilterItem
-          value="drafts"
-          label={t('roster.filters.drafts')}
-          count={draftAgents}
-        />
-      </SegmentedControl>
-      <SearchInput
-        aria-label={t('roster.searchLabel')}
-        className="h-8 w-50 max-w-full"
-        placeholder={t('roster.searchPlaceholder')}
-        value={keyword}
-        onValueChange={onKeywordChange}
+        if (nextFilter)
+          void setFilter(nextFilter)
+      }}
+    >
+      <RosterFilterItem
+        value="all"
+        label={t('roster.filters.all')}
       />
-      <div className="ml-auto flex min-w-0 items-center gap-2">
+      <RosterFilterItem
+        value="published"
+        label={t('roster.filters.published')}
+        count={publishedAgents}
+      />
+      <RosterFilterItem
+        value="drafts"
+        label={t('roster.filters.drafts')}
+        count={draftAgents}
+      />
+    </SegmentedControl>
+  )
+}
+
+function RosterSearchFilter() {
+  const { t } = useTranslation('agentV2')
+  const [keyword, setKeyword] = useQueryState(rosterQueryParamNames.keyword, rosterKeywordQueryParser)
+
+  return (
+    <SearchInput
+      aria-label={t('roster.searchLabel')}
+      className="h-8 w-50 min-w-0 shrink"
+      placeholder={t('roster.searchPlaceholder')}
+      value={keyword}
+      onValueChange={(value) => {
+        void setKeyword(value)
+      }}
+    />
+  )
+}
+
+function RosterCreatedByMeFilter() {
+  const { t } = useTranslation('agentV2')
+  const [createdByMe, setCreatedByMe] = useQueryState(rosterQueryParamNames.createdByMe, rosterCreatedByMeQueryParser)
+
+  return (
+    <label className="flex h-8 shrink-0 cursor-pointer items-center gap-1 rounded-lg bg-components-input-bg-normal px-2 py-1 whitespace-nowrap">
+      <Checkbox
+        checked={createdByMe}
+        onCheckedChange={(checked) => {
+          void setCreatedByMe(checked === true)
+        }}
+      />
+      <span className="p-1 system-sm-regular text-text-tertiary">{t('roster.filters.createdByMe')}</span>
+    </label>
+  )
+}
+
+export function RosterToolbar({
+  draftAgents,
+  publishedAgents,
+}: RosterToolbarProps) {
+  return (
+    <div className="flex min-w-0 items-center gap-2">
+      <RosterStatusFilter
+        draftAgents={draftAgents}
+        publishedAgents={publishedAgents}
+      />
+      <RosterSearchFilter />
+      <div className="flex h-4 shrink-0 px-1" aria-hidden="true">
+        <div className="h-full w-px bg-divider-regular" />
+      </div>
+      <RosterCreatedByMeFilter />
+      <div className="ml-auto flex shrink-0 items-center gap-2">
+        <RosterSortSelect />
         <CreateAgentDialog />
       </div>
     </div>
diff --git a/web/features/agent-v2/roster/page.tsx b/web/features/agent-v2/roster/page.tsx
index b207818ca6e..7a63dd1a6d7 100644
--- a/web/features/agent-v2/roster/page.tsx
+++ b/web/features/agent-v2/roster/page.tsx
@@ -1,6 +1,6 @@
 'use client'
 
-import type { AgentRosterListItem } from './components/agent-roster-list'
+import type { AgentAppPartial } from '@dify/contracts/api/console/agent/types.gen'
 import type { RosterFilterValue } from './components/roster-filter'
 import {
   ScrollAreaContent,
@@ -12,20 +12,26 @@ import {
 import { Tabs, TabsList, TabsPanel, TabsTab } from '@langgenius/dify-ui/tabs'
 import { keepPreviousData, useInfiniteQuery } from '@tanstack/react-query'
 import { useDebounce } from 'ahooks'
-import { debounce, parseAsString, parseAsStringLiteral, useQueryState } from 'nuqs'
+import { useQueryState } from 'nuqs'
 import { useTranslation } from 'react-i18next'
 import useDocumentTitle from '@/hooks/use-document-title'
 import { consoleQuery } from '@/service/client'
 import { AgentRosterList } from './components/agent-roster-list'
-import { ROSTER_FILTER_VALUES } from './components/roster-filter'
 import { RosterToolbar } from './components/roster-toolbar'
+import {
+  rosterCreatedByMeQueryParser,
+  rosterFilterQueryParser,
+  rosterKeywordQueryParser,
+  rosterQueryParamNames,
+  rosterSortByQueryParser,
+} from './query-params'
 
 const ROSTER_PAGE_SIZE = 30
-const isAgentPublished = (agent: AgentRosterListItem) => agent.active_config_is_published === true
-const rosterTabClassName = 'pt-0 pb-2 system-xl-semibold data-active:border-util-colors-blue-brand-blue-brand-500 data-disabled:opacity-100'
+const isAgentPublished = (agent: AgentAppPartial) => agent.active_config_is_published === true
+const rosterTabClassName = 'pt-0 pb-2 system-xl-semibold data-active:border-transparent data-disabled:opacity-100'
 
 const getFilteredRosterItems = (
-  agents: AgentRosterListItem[],
+  agents: AgentAppPartial[],
   filter: RosterFilterValue,
 ) => {
   if (filter === 'published')
@@ -40,18 +46,17 @@ const getFilteredRosterItems = (
 export default function RosterPage() {
   const { t } = useTranslation('agentV2')
   const { t: tCommon } = useTranslation('common')
-  const [keyword, setKeyword] = useQueryState('keyword', parseAsString.withDefault('').withOptions({
-    limitUrlUpdates: debounce(300),
-  }))
-  const [rosterFilter, setRosterFilter] = useQueryState(
-    'filter',
-    parseAsStringLiteral(ROSTER_FILTER_VALUES).withDefault('all'),
-  )
+  const [keyword] = useQueryState(rosterQueryParamNames.keyword, rosterKeywordQueryParser)
+  const [rosterFilter] = useQueryState(rosterQueryParamNames.filter, rosterFilterQueryParser)
+  const [createdByMe] = useQueryState(rosterQueryParamNames.createdByMe, rosterCreatedByMeQueryParser)
+  const [sortBy] = useQueryState(rosterQueryParamNames.sortBy, rosterSortByQueryParser)
   const debouncedKeyword = useDebounce(keyword.trim(), { wait: 300 })
 
   const rosterQueryInput = {
     limit: ROSTER_PAGE_SIZE,
+    sort_by: sortBy,
     ...(debouncedKeyword ? { name: debouncedKeyword } : {}),
+    ...(createdByMe ? { is_created_by_me: true } : {}),
   }
 
   const {
@@ -76,7 +81,7 @@ export default function RosterPage() {
     }),
   })
 
-  const rosterItems: AgentRosterListItem[] = rosterPages?.pages.flatMap(page => page.data) ?? []
+  const rosterItems: AgentAppPartial[] = rosterPages?.pages.flatMap(page => page.data) ?? []
   const publishedAgents = rosterItems.filter(isAgentPublished).length
   const draftAgents = Math.max(rosterItems.length - publishedAgents, 0)
   const filteredRosterItems = getFilteredRosterItems(rosterItems, rosterFilter)
@@ -109,14 +114,6 @@ export default function RosterPage() {
           <div className="mt-3.5">
             <RosterToolbar
               draftAgents={draftAgents}
-              filter={rosterFilter}
-              keyword={keyword}
-              onFilterChange={(value) => {
-                void setRosterFilter(value)
-              }}
-              onKeywordChange={(value) => {
-                void setKeyword(value)
-              }}
               publishedAgents={publishedAgents}
             />
           </div>
diff --git a/web/features/agent-v2/roster/query-params.ts b/web/features/agent-v2/roster/query-params.ts
new file mode 100644
index 00000000000..56a58a3e81b
--- /dev/null
+++ b/web/features/agent-v2/roster/query-params.ts
@@ -0,0 +1,20 @@
+import { debounce, parseAsBoolean, parseAsString, parseAsStringLiteral } from 'nuqs'
+import { ROSTER_FILTER_VALUES } from './components/roster-filter'
+import { DEFAULT_ROSTER_SORT_BY, ROSTER_SORT_BY_VALUES } from './components/roster-sort'
+
+export const rosterQueryParamNames = {
+  keyword: 'keyword',
+  filter: 'filter',
+  createdByMe: 'created_by_me',
+  sortBy: 'sort_by',
+} as const
+
+export const rosterKeywordQueryParser = parseAsString.withDefault('').withOptions({
+  limitUrlUpdates: debounce(300),
+})
+
+export const rosterFilterQueryParser = parseAsStringLiteral(ROSTER_FILTER_VALUES).withDefault('all')
+
+export const rosterCreatedByMeQueryParser = parseAsBoolean.withDefault(false)
+
+export const rosterSortByQueryParser = parseAsStringLiteral(ROSTER_SORT_BY_VALUES).withDefault(DEFAULT_ROSTER_SORT_BY)
diff --git a/web/features/tag-management/components/tag-selector.tsx b/web/features/tag-management/components/tag-selector.tsx
index 1ee0063e6a0..1a1f463b5eb 100644
--- a/web/features/tag-management/components/tag-selector.tsx
+++ b/web/features/tag-management/components/tag-selector.tsx
@@ -226,7 +226,7 @@ export const TagSelector = ({
         disabled={!canManageTags && !canBindOrUnbindTags}
         aria-label={triggerLabel}
         className={cn(
-          'block h-auto w-full rounded-lg border-0 bg-transparent p-0 text-left hover:bg-transparent focus:outline-hidden focus-visible:bg-transparent focus-visible:ring-2 focus-visible:ring-state-accent-solid data-popup-open:bg-state-base-hover data-popup-open:hover:bg-state-base-hover',
+          'block h-auto w-full rounded-lg border-0 bg-transparent p-0 text-left hover:bg-transparent focus:outline-hidden focus-visible:bg-transparent focus-visible:inset-ring-2 focus-visible:inset-ring-state-accent-solid data-popup-open:bg-state-base-hover data-popup-open:hover:bg-state-base-hover',
         )}
         icon={false}
       >
diff --git a/web/i18n/ar-TN/agent-v-2.json b/web/i18n/ar-TN/agent-v-2.json
index 79e5fb9c3bf..c32f0b0dac2 100644
--- a/web/i18n/ar-TN/agent-v-2.json
+++ b/web/i18n/ar-TN/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "إظهار قيمة {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "يعمل وكيلك في بيئة Linux. يمكنك إعداد التكوينات أو الأسرار مسبقًا بتعريف متغيرات البيئة هنا.\n\nبدلاً من ذلك، يمكنك دائمًا إعداد الأشياء عبر محادثة Build. راجع <docLink>المستندات</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "عيّن القيم التي يقرأها الوكيل وقت التشغيل، مثل مفتاح API تحتاجه أداة، دون وضعها في Prompt.\n\nفي وضع Build، يمكن للوكيل إعدادها لك.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "اختر نطاق متغير البيئة",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "يعمل وكيلك في بيئة Linux. يمكنك إعداد التكوينات أو الأسرار مسبقًا بتعريف متغيرات البيئة هنا. بدلاً من ذلك، يمكنك دائمًا إعداد الأشياء عبر محادثة Build. راجع المستندات.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "عيّن القيم التي يقرأها الوكيل وقت التشغيل، مثل مفتاح API تحتاجه أداة، دون وضعها في Prompt.\n\nفي وضع Build، يمكن للوكيل إعدادها لك.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "تبديل Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "ابنِ وكيلك عبر الدردشة",
   "agentDetail.configure.build.inputPlaceholder": "صِف ما يجب أن يفعله وكيلك",
   "agentDetail.configure.build.startBuild": "ابدأ البناء",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "ملاحظة مستمرة لهذا الإصدار يستخدمها الوكيل للاحتفاظ بالسياق بين الجلسات. تتم إضافتها إلى ملفاتك عند التطبيق.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} تغيير للتطبيق",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} تغييرات للتطبيق",
   "agentDetail.configure.buildDraft.discard": "تجاهل",
   "agentDetail.configure.buildDraft.modeBadge": "وضع البناء",
   "agentDetail.configure.buildDraft.modeDescription": "أنت في وضع البناء. شكّل هذا الإعداد عبر الدردشة على اليمين، ثم طبّق.",
+  "agentDetail.configure.buildDraft.rewritten": "أُعيدت صياغته",
   "agentDetail.configure.buildDraft.title": "مسودة البناء",
+  "agentDetail.configure.buildDraft.updated": "تم التحديث",
   "agentDetail.configure.chatFeatures.description": "شكّل تجربة الدردشة للمستخدم النهائي على Web app وأسطح الدردشة.",
   "agentDetail.configure.chatFeatures.title": "ميزات الدردشة",
   "agentDetail.configure.files.add": "إضافة ملف",
+  "agentDetail.configure.files.buildNote.generated": "تم إنشاؤه",
+  "agentDetail.configure.files.buildNote.richTooltip": "سجل الوكيل لما أعدّه في وضع Build. يقرأه في بداية كل محادثة إلى جانب Prompt الخاص بك. <docLink>معرفة المزيد</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "سجل الوكيل لما أعدّه في وضع Build. يقرأه في بداية كل محادثة إلى جانب Prompt الخاص بك. معرفة المزيد",
   "agentDetail.configure.files.empty.description": "قم بتحميل المستندات التي يمكن للوكيل قراءتها، مثل المواصفات أو القوالب أو الإرشادات",
   "agentDetail.configure.files.empty.title": "لا توجد ملفات بعد",
   "agentDetail.configure.files.label": "الملفات",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "فشل تحميل المعاينة.",
   "agentDetail.configure.files.preview.unsupported": "هذا الملف لا يدعم المعاينة.",
   "agentDetail.configure.files.remove": "إزالة {{name}}",
-  "agentDetail.configure.files.tip": "زوّد وكيلك بأي مثال أو إرشاد أو مجرد حامل للمعلومات يكون طويلاً جداً بحيث لا يناسب البرومبت. أشر إليه صراحةً باستخدام / في البرومبت.",
+  "agentDetail.configure.files.tip": "للمواد المرجعية الأطول من أن تُلصق في Prompt: المواصفات، القوالب، الإرشادات. أشر إليها باستخدام / في Prompt.\n\nفي وضع Build، يمكن للوكيل إعدادها لك.",
   "agentDetail.configure.files.toggle": "تبديل الملفات",
   "agentDetail.configure.files.treeLabel": "ملفات الوكيل",
   "agentDetail.configure.files.upload.action": "تحميل",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "إزالة {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "مصادر استرجاع المعرفة التي يمكن لهذا الوكيل استخدامها أثناء تنسيق المهام.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "اربط قواعد المعرفة التي يبحث فيها الوكيل أثناء عمله، لتستند إجاباته إلى محتواك. <docLink>معرفة المزيد</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "اربط قواعد المعرفة التي يبحث فيها الوكيل أثناء عمله، لتستند إجاباته إلى محتواك. معرفة المزيد",
   "agentDetail.configure.knowledgeRetrieval.toggle": "تبديل استرجاع المعرفة",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "يرجى تكوين نموذج في تصفية البيانات الوصفية",
   "agentDetail.configure.model.label": "النموذج",
   "agentDetail.configure.orchestrate": "تنسيق",
   "agentDetail.configure.preview.chatFeatures": "ميزات الدردشة",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "مصادقة المستخدم النهائي",
   "agentDetail.configure.preview.inputPlaceholder": "إرسال رسالة إلى {{name}}",
   "agentDetail.configure.preview.restart": "إعادة تشغيل المعاينة",
+  "agentDetail.configure.preview.sandboxNotice": "يعمل Agent داخل صندوق رمل Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "بالنسبة إلى إصدار Dify Community Edition، يعمل كل Agent من Agents الخاصة بك في بيئة صندوق رمل Linux 7.0.0-10060-aws داخل Docker لديك. تعديلاتك على البيئة عبر Build Chats مستمرة.",
   "agentDetail.configure.preview.title": "معاينة",
   "agentDetail.configure.prompt.copied": "تم نسخ المطالبة",
   "agentDetail.configure.prompt.copy": "نسخ المطالبة",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "معرفة المزيد",
   "agentDetail.configure.rightPanel.modeLabel": "وضع تكوين الوكيل",
   "agentDetail.configure.rightPanel.preview": "معاينة",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview متاح فقط على Dify Cloud وEnterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "يشغل Preview الوكيل المكتمل كما سيراه المستخدمون، مع ردود واضحة وميزات الدردشة.",
   "agentDetail.configure.rightPanel.previewTipTitle": "عاين وكيلك",
   "agentDetail.configure.skills.add": "إضافة مهارة",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "مهارة",
   "agentDetail.configure.skills.label": "المهارات",
   "agentDetail.configure.skills.remove": "إزالة {{name}}",
-  "agentDetail.configure.skills.richTip": "زوّد وكيلك بقدرات معيارية. راجع <docLink>الوثائق</docLink>. أشر إليها صراحةً باستخدام / في البرومبت.",
-  "agentDetail.configure.skills.tip": "زوّد وكيلك بقدرات معيارية. راجع الوثائق. أشر إليها صراحةً باستخدام / في البرومبت.",
+  "agentDetail.configure.skills.richTip": "اجمع التعليمات والملفات والبرامج النصية لمهمة متكررة في Skill. أشر إليها باستخدام / في Prompt. <docLink>معرفة المزيد</docLink>\n\nفي وضع Build، يمكن للوكيل إعدادها لك.",
+  "agentDetail.configure.skills.tip": "اجمع التعليمات والملفات والبرامج النصية لمهمة متكررة في Skill. أشر إليها باستخدام / في Prompt. معرفة المزيد\n\nفي وضع Build، يمكن للوكيل إعدادها لك.",
   "agentDetail.configure.skills.toggle": "تبديل المهارات",
   "agentDetail.configure.skills.upload.action": "تحميل",
   "agentDetail.configure.skills.upload.browse": "استعراض",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "إجراءات أخرى لـ {{name}}",
   "agentDetail.configure.tools.removeAction": "إزالة {{name}}",
   "agentDetail.configure.tools.removeProvider": "إزالة جميع الأدوات",
-  "agentDetail.configure.tools.richTip": "زوّد وكيلك بإضافات Dify. راجع <pluginDocLink>الوثائق</pluginDocLink>. أشر إليها صراحةً باستخدام / في البرومبت.\nبدلاً من ذلك، يمكنك دائماً إخبار الوكيل بتثبيت أدوات أخرى والمصادقة عليها (مثل MCPs وأدوات CLI) عبر دردشة Build. لاحظ أن الأدوات التي يتم تكوينها عبر دردشات Build لن تظهر هنا، لكنها ستظل متاحة لوكيلك لاحقاً. راجع <buildDocLink>الوثائق</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "زوّد وكيلك بإضافات Dify. راجع الوثائق. أشر إليها صراحةً باستخدام / في البرومبت. بدلاً من ذلك، يمكنك دائماً إخبار الوكيل بتثبيت أدوات أخرى والمصادقة عليها (مثل MCPs وأدوات CLI) عبر دردشة Build. لاحظ أن الأدوات التي يتم تكوينها عبر دردشات Build لن تظهر هنا، لكنها ستظل متاحة لوكيلك لاحقاً. راجع الوثائق.",
+  "agentDetail.configure.tools.richTip": "أضف أدوات Dify من مساحة عملك. أشر إليها باستخدام / في Prompt. <docLink>معرفة المزيد</docLink>\n\nالأدوات التي يثبتها الوكيل في sandbox الخاص به لن تظهر هنا، لكنها ستبقى متاحة له.",
+  "agentDetail.configure.tools.tip": "أضف أدوات Dify من مساحة عملك. أشر إليها باستخدام / في Prompt. معرفة المزيد\n\nالأدوات التي يثبتها الوكيل في sandbox الخاص به لن تظهر هنا، لكنها ستبقى متاحة له.",
   "agentDetail.configure.tools.toggle": "تبديل الأدوات",
   "agentDetail.configure.tools.toolTabs.all": "الكل",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "المكونات الإضافية",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "مسار نظام الملفات",
   "agentDetail.configure.workingDirectory.description": "الملفات التي يقرأها الوكيل ويكتبها أثناء التشغيل. كل ما ينشئه، مثل qna_report.pdf، يظهر هنا.",
+  "agentDetail.configure.workingDirectory.fileSystem": "نظام الملفات",
+  "agentDetail.configure.workingDirectory.home": "الرئيسية",
   "agentDetail.configure.workingDirectory.open": "فتح دليل العمل",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "العودة إلى دليل العمل",
   "agentDetail.configure.workingDirectory.title": "دليل العمل",
   "agentDetail.configure.workingDirectory.treeLabel": "ملفات دليل العمل",
+  "agentDetail.configure.workingDirectory.workingDirectory": "دليل العمل",
   "agentDetail.documentTitle": "وكيل",
   "agentDetail.logs.description": "تسجل السجلات الكاملة حالة تشغيل التطبيق، بما في ذلك مدخلات المستخدم وردود الوكيل والتخطيط واستخدامات الأدوات.",
   "agentDetail.logs.empty": "لم يتم العثور على سجلات",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "الاسم مطلوب.",
   "roster.createForm.roleLabel": "الدور",
   "roster.createForm.rolePlaceholder": "مثلاً مساعد بحث",
-  "roster.createForm.roleRequired": "الدور مطلوب.",
   "roster.createSuccess": "تم إنشاء الوكيل.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "ستتم إزالة هذا الوكيل من قائمة Roster النشطة.",
@@ -354,6 +372,7 @@
   "roster.empty": "لا يوجد وكيل بعد",
   "roster.emptySearch": "لا يوجد وكلاء مطابقون",
   "roster.filters.all": "الكل",
+  "roster.filters.createdByMe": "أنشأتها أنا",
   "roster.filters.drafts": "مسودات",
   "roster.filters.label": "مرشحات الوكيل",
   "roster.filters.published": "منشور",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "تم حفظ الوكيل في Roster.",
   "roster.searchLabel": "البحث عن الوكلاء",
   "roster.searchPlaceholder": "ابحث عن الوكلاء بالاسم…",
+  "roster.sort.earliestCreated": "الأقدم إنشاءً",
+  "roster.sort.label": "فرز حسب",
+  "roster.sort.lastModified": "آخر تعديل",
+  "roster.sort.optionsLabel": "خيارات الفرز",
+  "roster.sort.recentlyCreated": "الأحدث إنشاءً",
   "roster.tabs.agent": "وكيل",
-  "roster.tabs.human": "إنسان",
+  "roster.tabs.human": "جهات الاتصال",
   "roster.tabsLabel": "نوع Roster",
   "roster.updateSuccess": "تم تحديث الوكيل.",
   "roster.usageStatus.draft": "مسودة"
diff --git a/web/i18n/ar-TN/common.json b/web/i18n/ar-TN/common.json
index 5af5b93a739..74856605669 100644
--- a/web/i18n/ar-TN/common.json
+++ b/web/i18n/ar-TN/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "تصفح السوق لتثبيت مزود، أو قم بتكوين المزودين في الإعدادات.",
   "modelProvider.selector.onlyCompatibleModelsShown": "يتم عرض النماذج المتوافقة فقط",
   "modelProvider.selector.showIncompatibleModels": "عرض النماذج غير المتوافقة",
+  "modelProvider.selector.suggestion": "اقتراح",
+  "modelProvider.selector.suggestionTip": "موصى به لمهام الوكيل",
   "modelProvider.showModels": "عرض النماذج",
   "modelProvider.speechToTextModel.key": "نموذج تحويل الكلام إلى نص",
   "modelProvider.speechToTextModel.tip": "تعيين النموذج الافتراضي لإدخال تحويل الكلام إلى نص في المحادثة.",
diff --git a/web/i18n/ar-TN/workflow.json b/web/i18n/ar-TN/workflow.json
index cba85db3f15..d291f0c04f7 100644
--- a/web/i18n/ar-TN/workflow.json
+++ b/web/i18n/ar-TN/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "انشر أحدث سير عمل وتأكد من وجود عقدة إدخال مستخدم متصلة قبل تكوينها كأداة.",
   "common.workflowAsToolTip": "التكوين المطلوب للأداة بعد تحديث سير العمل.",
   "common.workflowProcess": "عملية سير العمل",
+  "common.workflowProcessFailed": "فشلت عملية سير العمل",
+  "common.workflowProcessPaused": "تم إيقاف عملية سير العمل مؤقتًا",
+  "common.workflowProcessRunning": "عملية سير العمل قيد التشغيل",
+  "common.workflowProcessSucceeded": "نجحت عملية سير العمل",
   "customWebhook": "ويب هوك مخصص",
   "debug.copyLastRun": "نسخ آخر تشغيل",
   "debug.copyLastRunError": "فشل نسخ مدخلات آخر تشغيل",
diff --git a/web/i18n/de-DE/agent-v-2.json b/web/i18n/de-DE/agent-v-2.json
index 9cf45ddeebd..ce433129037 100644
--- a/web/i18n/de-DE/agent-v-2.json
+++ b/web/i18n/de-DE/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Wert von {{key}} anzeigen",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Dein Agent läuft in einer Linux-Umgebung. Du kannst Konfigurationen oder Secrets vorab festlegen, indem du hier Umgebungsvariablen definierst.\n\nAlternativ kannst du die Einrichtung jederzeit über einen Build-Chat vornehmen. Siehe <docLink>Dokumentation</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Legen Sie Werte fest, die der Agent zur Laufzeit liest, etwa einen API-Schlüssel, den ein Tool benötigt, ohne sie in den Prompt zu schreiben.\n\nIm Build-Modus kann der Agent diese Einrichtung für Sie übernehmen.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Geltungsbereich der Umgebungsvariable auswählen",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Dein Agent läuft in einer Linux-Umgebung. Du kannst Konfigurationen oder Secrets vorab festlegen, indem du hier Umgebungsvariablen definierst. Alternativ kannst du die Einrichtung jederzeit über einen Build-Chat vornehmen. Siehe Dokumentation.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Legen Sie Werte fest, die der Agent zur Laufzeit liest, etwa einen API-Schlüssel, den ein Tool benötigt, ohne sie in den Prompt zu schreiben.\n\nIm Build-Modus kann der Agent diese Einrichtung für Sie übernehmen.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Env Editor umschalten",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Agent per Chat erstellen",
   "agentDetail.configure.build.inputPlaceholder": "Beschreiben Sie, was Ihr Agent tun soll",
   "agentDetail.configure.build.startBuild": "Build starten",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Eine fortlaufende Notiz zu diesem Build, die der Agent verwendet, um Kontext sitzungsübergreifend zu behalten. Sie wird Ihren Dateien hinzugefügt, wenn Sie anwenden.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} Änderung anzuwenden",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} Änderungen anzuwenden",
   "agentDetail.configure.buildDraft.discard": "Verwerfen",
   "agentDetail.configure.buildDraft.modeBadge": "Build-Modus",
   "agentDetail.configure.buildDraft.modeDescription": "Sie sind im Build-Modus. Formen Sie diese Einrichtung über den Chat rechts und wenden Sie sie dann an.",
+  "agentDetail.configure.buildDraft.rewritten": "Neu geschrieben",
   "agentDetail.configure.buildDraft.title": "Build-Entwurf",
+  "agentDetail.configure.buildDraft.updated": "Aktualisiert",
   "agentDetail.configure.chatFeatures.description": "Gestalten Sie das Chat-Erlebnis für Endnutzer in Ihrer Webapp und in Chat-Oberflächen.",
   "agentDetail.configure.chatFeatures.title": "Chat-Funktionen",
   "agentDetail.configure.files.add": "Datei hinzufügen",
+  "agentDetail.configure.files.buildNote.generated": "Generiert",
+  "agentDetail.configure.files.buildNote.richTooltip": "Die Aufzeichnung des Agenten darüber, was er im Build-Modus eingerichtet hat. Er liest sie zu Beginn jeder Unterhaltung zusammen mit Ihrem Prompt. <docLink>Mehr erfahren</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Die Aufzeichnung des Agenten darüber, was er im Build-Modus eingerichtet hat. Er liest sie zu Beginn jeder Unterhaltung zusammen mit Ihrem Prompt. Mehr erfahren",
   "agentDetail.configure.files.empty.description": "Laden Sie Dokumente hoch, die der Agent lesen kann, z. B. Spezifikationen, Vorlagen oder Richtlinien",
   "agentDetail.configure.files.empty.title": "Noch keine Dateien",
   "agentDetail.configure.files.label": "Dateien",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Vorschau konnte nicht geladen werden.",
   "agentDetail.configure.files.preview.unsupported": "Für diese Datei wird keine Vorschau unterstützt.",
   "agentDetail.configure.files.remove": "{{name}} entfernen",
-  "agentDetail.configure.files.tip": "Stellen Sie Ihrem Agenten Beispiele, Richtlinien oder einfach Informationen bereit, die zu lang für den Prompt sind. Verweisen Sie im Prompt explizit mit / darauf.",
+  "agentDetail.configure.files.tip": "Für Referenzmaterial, das zu lang zum Einfügen in den Prompt ist: Spezifikationen, Vorlagen, Richtlinien. Verweisen Sie im Prompt mit / darauf.\n\nIm Build-Modus kann der Agent diese Einrichtung für Sie übernehmen.",
   "agentDetail.configure.files.toggle": "Dateien umschalten",
   "agentDetail.configure.files.treeLabel": "Agentendateien",
   "agentDetail.configure.files.upload.action": "Hochladen",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "{{name}} entfernen",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Wissensabfragequellen, die dieser Agent bei der Orchestrierung von Aufgaben verwenden kann.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Verbinden Sie Wissensdatenbanken, die der Agent während der Arbeit durchsucht, damit seine Antworten auf Ihren Inhalten basieren. <docLink>Mehr erfahren</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Verbinden Sie Wissensdatenbanken, die der Agent während der Arbeit durchsucht, damit seine Antworten auf Ihren Inhalten basieren. Mehr erfahren",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Wissensabfrage umschalten",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Bitte konfigurieren Sie ein Modell in der Metadatenfilterung",
   "agentDetail.configure.model.label": "Modell",
   "agentDetail.configure.orchestrate": "Orchestrieren",
   "agentDetail.configure.preview.chatFeatures": "Chat-Funktionen",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Endbenutzer-Authentifizierung",
   "agentDetail.configure.preview.inputPlaceholder": "Nachricht an {{name}}",
   "agentDetail.configure.preview.restart": "Vorschau neu starten",
+  "agentDetail.configure.preview.sandboxNotice": "Der Agent läuft in einer Linux-Sandbox.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "In der Dify Community Edition läuft jeder deiner Agents in einer Linux 7.0.0-10060-aws-Sandbox-Umgebung innerhalb deines Docker. Änderungen an der Umgebung über Build Chats bleiben erhalten.",
   "agentDetail.configure.preview.title": "Vorschau",
   "agentDetail.configure.prompt.copied": "Prompt kopiert",
   "agentDetail.configure.prompt.copy": "Prompt kopieren",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Mehr erfahren",
   "agentDetail.configure.rightPanel.modeLabel": "Agent-Konfigurationsmodus",
   "agentDetail.configure.rightPanel.preview": "Vorschau",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview ist nur in Dify Cloud und Enterprise verfügbar.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview führt den fertigen Agenten so aus, wie deine Benutzer ihn sehen, mit klaren Antworten und Chat-Funktionen.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Agenten vorschauen",
   "agentDetail.configure.skills.add": "Skill hinzufügen",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Skill",
   "agentDetail.configure.skills.label": "Skills",
   "agentDetail.configure.skills.remove": "{{name}} entfernen",
-  "agentDetail.configure.skills.richTip": "Statten Sie Ihren Agenten mit modularen Fähigkeiten aus. Siehe <docLink>Dokumentation</docLink>. Verweisen Sie im Prompt explizit mit / darauf.",
-  "agentDetail.configure.skills.tip": "Statten Sie Ihren Agenten mit modularen Fähigkeiten aus. Siehe Dokumentation. Verweisen Sie im Prompt explizit mit / darauf.",
+  "agentDetail.configure.skills.richTip": "Bündeln Sie Anweisungen, Dateien und Skripte für eine wiederkehrende Aufgabe in einer Skill. Verweisen Sie im Prompt mit / darauf. <docLink>Mehr erfahren</docLink>\n\nIm Build-Modus kann der Agent diese Einrichtung für Sie übernehmen.",
+  "agentDetail.configure.skills.tip": "Bündeln Sie Anweisungen, Dateien und Skripte für eine wiederkehrende Aufgabe in einer Skill. Verweisen Sie im Prompt mit / darauf. Mehr erfahren\n\nIm Build-Modus kann der Agent diese Einrichtung für Sie übernehmen.",
   "agentDetail.configure.skills.toggle": "Skills umschalten",
   "agentDetail.configure.skills.upload.action": "Hochladen",
   "agentDetail.configure.skills.upload.browse": "durchsuchen",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Weitere Aktionen für {{name}}",
   "agentDetail.configure.tools.removeAction": "{{name}} entfernen",
   "agentDetail.configure.tools.removeProvider": "Alle Tools entfernen",
-  "agentDetail.configure.tools.richTip": "Statten Sie Ihren Agenten mit Dify-Plugins aus. Siehe <pluginDocLink>Dokumentation</pluginDocLink>. Verweisen Sie im Prompt explizit mit / darauf.\nAlternativ können Sie den Agenten jederzeit über einen Build-Chat anweisen, andere Tools (z. B. MCPs und CLI-Tools) zu installieren und zu authentifizieren. Beachten Sie, dass über Build-Chats konfigurierte Tools hier NICHT angezeigt werden, Ihrem Agenten künftig aber weiterhin zur Verfügung stehen. Siehe <buildDocLink>Dokumentation</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Statten Sie Ihren Agenten mit Dify-Plugins aus. Siehe Dokumentation. Verweisen Sie im Prompt explizit mit / darauf. Alternativ können Sie den Agenten jederzeit über einen Build-Chat anweisen, andere Tools (z. B. MCPs und CLI-Tools) zu installieren und zu authentifizieren. Beachten Sie, dass über Build-Chats konfigurierte Tools hier NICHT angezeigt werden, Ihrem Agenten künftig aber weiterhin zur Verfügung stehen. Siehe Dokumentation.",
+  "agentDetail.configure.tools.richTip": "Fügen Sie Dify-Tools aus Ihrem Workspace hinzu. Verweisen Sie im Prompt mit / darauf. <docLink>Mehr erfahren</docLink>\n\nTools, die der Agent in seiner Sandbox installiert, erscheinen hier nicht, bleiben ihm aber verfügbar.",
+  "agentDetail.configure.tools.tip": "Fügen Sie Dify-Tools aus Ihrem Workspace hinzu. Verweisen Sie im Prompt mit / darauf. Mehr erfahren\n\nTools, die der Agent in seiner Sandbox installiert, erscheinen hier nicht, bleiben ihm aber verfügbar.",
   "agentDetail.configure.tools.toggle": "Tools umschalten",
   "agentDetail.configure.tools.toolTabs.all": "Alle",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugins",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Dateisystempfad",
   "agentDetail.configure.workingDirectory.description": "Dateien, die der Agent während der Ausführung liest und schreibt. Alles, was er erzeugt, z. B. qna_report.pdf, erscheint hier.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Dateisystem",
+  "agentDetail.configure.workingDirectory.home": "Start",
   "agentDetail.configure.workingDirectory.open": "Arbeitsverzeichnis öffnen",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Zurück zum Arbeitsverzeichnis",
   "agentDetail.configure.workingDirectory.title": "Arbeitsverzeichnis",
   "agentDetail.configure.workingDirectory.treeLabel": "Dateien im Arbeitsverzeichnis",
+  "agentDetail.configure.workingDirectory.workingDirectory": "Arbeitsverzeichnis",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Vollständige Logs zeichnen den Laufzeitstatus der Anwendung auf, einschließlich Benutzereingaben, Agentenantworten, Planung und Tool-Nutzung.",
   "agentDetail.logs.empty": "Keine Logs gefunden",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Name ist erforderlich.",
   "roster.createForm.roleLabel": "Rolle",
   "roster.createForm.rolePlaceholder": "z. B. Research Assistant",
-  "roster.createForm.roleRequired": "Rolle ist erforderlich.",
   "roster.createSuccess": "Agent erstellt.",
   "roster.dateTimeFormat": "DD.MM.YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Dieser Agent wird aus der aktiven Roster-Liste entfernt.",
@@ -354,6 +372,7 @@
   "roster.empty": "Noch kein Agent",
   "roster.emptySearch": "Keine passenden Agenten",
   "roster.filters.all": "Alle",
+  "roster.filters.createdByMe": "Von mir erstellt",
   "roster.filters.drafts": "Entwürfe",
   "roster.filters.label": "Agentenfilter",
   "roster.filters.published": "Veröffentlicht",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent im Roster gespeichert.",
   "roster.searchLabel": "Agenten suchen",
   "roster.searchPlaceholder": "Agenten nach Namen suchen…",
+  "roster.sort.earliestCreated": "Zuerst erstellt",
+  "roster.sort.label": "Sortieren nach",
+  "roster.sort.lastModified": "Zuletzt geändert",
+  "roster.sort.optionsLabel": "Sortieroptionen",
+  "roster.sort.recentlyCreated": "Zuletzt erstellt",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Mensch",
+  "roster.tabs.human": "Kontakte",
   "roster.tabsLabel": "Roster-Typ",
   "roster.updateSuccess": "Agent aktualisiert.",
   "roster.usageStatus.draft": "Entwurf"
diff --git a/web/i18n/de-DE/common.json b/web/i18n/de-DE/common.json
index 6c8b9936e01..aaa671db3d7 100644
--- a/web/i18n/de-DE/common.json
+++ b/web/i18n/de-DE/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Durchsuchen Sie den Marketplace, um einen zu installieren, oder konfigurieren Sie Anbieter in den Einstellungen.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Es werden nur kompatible Modelle angezeigt",
   "modelProvider.selector.showIncompatibleModels": "Inkompatible Modelle anzeigen",
+  "modelProvider.selector.suggestion": "Vorschlag",
+  "modelProvider.selector.suggestionTip": "Empfohlen für Agentenaufgaben",
   "modelProvider.showModels": "Modelle anzeigen",
   "modelProvider.speechToTextModel.key": "Sprach-zu-Text-Modell",
   "modelProvider.speechToTextModel.tip": "Legen Sie das Standardmodell für die Spracheingabe in Konversationen fest.",
diff --git a/web/i18n/de-DE/workflow.json b/web/i18n/de-DE/workflow.json
index 5e84568211e..f4bbdb01f60 100644
--- a/web/i18n/de-DE/workflow.json
+++ b/web/i18n/de-DE/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Veröffentlichen Sie den neuesten Workflow und stellen Sie sicher, dass ein verbundener User-Input-Knoten vorhanden ist, bevor Sie ihn als Werkzeug konfigurieren.",
   "common.workflowAsToolTip": "Nach dem Workflow-Update ist eine Neukonfiguration des Tools erforderlich.",
   "common.workflowProcess": "Arbeitsablauf",
+  "common.workflowProcessFailed": "Arbeitsablauf fehlgeschlagen",
+  "common.workflowProcessPaused": "Arbeitsablauf pausiert",
+  "common.workflowProcessRunning": "Arbeitsablauf wird ausgeführt",
+  "common.workflowProcessSucceeded": "Arbeitsablauf erfolgreich",
   "customWebhook": "Benutzerdefinierter Webhook",
   "debug.copyLastRun": "Letzte Ausführung kopieren",
   "debug.copyLastRunError": "Fehler beim Kopieren der letzten Lauf-Eingaben",
diff --git a/web/i18n/en-US/agent-v-2.json b/web/i18n/en-US/agent-v-2.json
index 1f51d53acf0..dd489624891 100644
--- a/web/i18n/en-US/agent-v-2.json
+++ b/web/i18n/en-US/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Reveal {{key}} value",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Your agent runs in a Linux environment. You can preset configs or secrets by defining environmental variables here.\n\nAlternatively, you can always have things set up via a Build chat. Refer to <docLink>docs</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Set values the agent reads at runtime, like an API key a tool needs, without putting them in the Prompt.\n\nIn Build mode, the agent can set these up for you.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Select environment variable scope",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Your agent runs in a Linux environment. You can preset configs or secrets by defining environmental variables here. Alternatively, you can always have things set up via a Build chat. Refer to docs.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Set values the agent reads at runtime, like an API key a tool needs, without putting them in the Prompt.\n\nIn Build mode, the agent can set these up for you.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Toggle env editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Build your agent by chatting",
   "agentDetail.configure.build.inputPlaceholder": "Describe what your agent should do",
   "agentDetail.configure.build.startBuild": "Start build",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "A running note of this build the agent uses to keep context across sessions. It's added to your files when you Apply.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} change to apply",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} changes to apply",
   "agentDetail.configure.buildDraft.discard": "Discard",
   "agentDetail.configure.buildDraft.modeBadge": "Build mode",
   "agentDetail.configure.buildDraft.modeDescription": "You're in build mode. Shape this setup through the chat on the right, then Apply.",
+  "agentDetail.configure.buildDraft.rewritten": "Rewritten",
   "agentDetail.configure.buildDraft.title": "Build draft",
+  "agentDetail.configure.buildDraft.updated": "Updated",
   "agentDetail.configure.chatFeatures.description": "Shape the end-user chat experience on your web app and chat surfaces.",
   "agentDetail.configure.chatFeatures.title": "Chat Features",
   "agentDetail.configure.files.add": "Add file",
+  "agentDetail.configure.files.buildNote.generated": "Generated",
+  "agentDetail.configure.files.buildNote.richTooltip": "The agent's record of what it set up in Build mode. It reads this at the start of every conversation, alongside your Prompt. <docLink>Learn more</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "The agent's record of what it set up in Build mode. It reads this at the start of every conversation, alongside your Prompt. Learn more",
   "agentDetail.configure.files.empty.description": "Upload docs the agent can read, like specs, templates, or guidelines",
   "agentDetail.configure.files.empty.title": "No files yet",
   "agentDetail.configure.files.label": "Files",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Failed to load preview.",
   "agentDetail.configure.files.preview.unsupported": "Preview is not supported for this file.",
   "agentDetail.configure.files.remove": "Remove {{name}}",
-  "agentDetail.configure.files.tip": "Provide your agent with any example, guideline or simply a bearer of information that is too long for prompt. Make explicit reference with / in Prompt.",
+  "agentDetail.configure.files.tip": "For reference material too long to paste into the Prompt: specs, templates, guidelines. Reference with / in the Prompt.\n\nIn Build mode, the agent can set these up for you.",
   "agentDetail.configure.files.toggle": "Toggle files",
   "agentDetail.configure.files.treeLabel": "Agent files",
   "agentDetail.configure.files.upload.action": "Upload",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Remove {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Knowledge retrieval sources this agent can use while orchestrating tasks.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Connect knowledge bases the agent searches while it works, so its answers draw on your content. <docLink>Learn more</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Connect knowledge bases the agent searches while it works, so its answers draw on your content. Learn more",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Toggle knowledge retrieval",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Please config a model in metadata filtering",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Orchestrate",
   "agentDetail.configure.preview.chatFeatures": "Chat Features",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "End-user authentication",
   "agentDetail.configure.preview.inputPlaceholder": "Message {{name}}",
   "agentDetail.configure.preview.restart": "Restart preview",
+  "agentDetail.configure.preview.sandboxNotice": "Agent runs in a Linux sandbox.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "For Dify Community Edition, each of your agents runs in a Linux 7.0.0-10060-aws sandbox environment within your docker. Your edits to the environment via Build Chats are persistent.",
   "agentDetail.configure.preview.title": "Preview",
   "agentDetail.configure.prompt.copied": "Prompt copied",
   "agentDetail.configure.prompt.copy": "Copy prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Learn more",
   "agentDetail.configure.rightPanel.modeLabel": "Agent configuration mode",
   "agentDetail.configure.rightPanel.preview": "Preview",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview is available on Dify Cloud and Enterprise only.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview runs the finished agent the way your users will see it, with clean replies and chat features.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Preview your agent",
   "agentDetail.configure.skills.add": "Add skill",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Skill",
   "agentDetail.configure.skills.label": "Skills",
   "agentDetail.configure.skills.remove": "Remove {{name}}",
-  "agentDetail.configure.skills.richTip": "Equip your agent with modular capabilities. Refer to <docLink>docs</docLink>. Make explicit reference with / in Prompt.",
-  "agentDetail.configure.skills.tip": "Equip your agent with modular capabilities. Refer to docs. Make explicit reference with / in Prompt.",
+  "agentDetail.configure.skills.richTip": "Package the instructions, files, and scripts for a recurring task into a skill. Reference with / in the Prompt. <docLink>Learn more</docLink>\n\nIn Build mode, the agent can set these up for you.",
+  "agentDetail.configure.skills.tip": "Package the instructions, files, and scripts for a recurring task into a skill. Reference with / in the Prompt. Learn more\n\nIn Build mode, the agent can set these up for you.",
   "agentDetail.configure.skills.toggle": "Toggle skills",
   "agentDetail.configure.skills.upload.action": "Upload",
   "agentDetail.configure.skills.upload.browse": "browse",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "More actions for {{name}}",
   "agentDetail.configure.tools.removeAction": "Remove {{name}}",
   "agentDetail.configure.tools.removeProvider": "Remove all tools",
-  "agentDetail.configure.tools.richTip": "Equip your agent with Dify's plugins. Refer to <pluginDocLink>docs</pluginDocLink>. Make explicit reference with / in Prompt.\nAlternatively, you can always tell the agent to install and authenticate other tools (e.g. MCPs & CLI tools) via a Build chat. Note that tools configured via Build chats will NOT show up here, although they will still be available to your agent going forward. Refer to <buildDocLink>docs</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Equip your agent with Dify's plugins. Refer to docs. Make explicit reference with / in Prompt. Alternatively, you can always tell the agent to install and authenticate other tools (e.g. MCPs & CLI tools) via a Build chat. Note that tools configured via Build chats will NOT show up here, although they will still be available to your agent going forward. Refer to docs.",
+  "agentDetail.configure.tools.richTip": "Add Dify tools from your workspace. Reference with / in the Prompt. <docLink>Learn more</docLink>\n\nTools the agent installs in its sandbox won't appear here but stay available to it.",
+  "agentDetail.configure.tools.tip": "Add Dify tools from your workspace. Reference with / in the Prompt. Learn more\n\nTools the agent installs in its sandbox won't appear here but stay available to it.",
   "agentDetail.configure.tools.toggle": "Toggle tools",
   "agentDetail.configure.tools.toolTabs.all": "All",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugins",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "File system path",
   "agentDetail.configure.workingDirectory.description": "Files the agent reads and writes as it runs. Anything it generates, like qna_report.pdf, shows up here.",
+  "agentDetail.configure.workingDirectory.fileSystem": "File system",
+  "agentDetail.configure.workingDirectory.home": "home",
   "agentDetail.configure.workingDirectory.open": "Open working directory",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Back to working directory",
   "agentDetail.configure.workingDirectory.title": "Working directory",
   "agentDetail.configure.workingDirectory.treeLabel": "Working directory files",
+  "agentDetail.configure.workingDirectory.workingDirectory": "working directory",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Full logs record the running status of the application, including user inputs, agent replies, planning and tool uses.",
   "agentDetail.logs.empty": "No logs found",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Name is required.",
   "roster.createForm.roleLabel": "Role",
   "roster.createForm.rolePlaceholder": "e.g. Research Assistant",
-  "roster.createForm.roleRequired": "Role is required.",
   "roster.createSuccess": "Agent created.",
   "roster.dateTimeFormat": "MM/DD/YYYY hh:mm:ss A",
   "roster.deleteDialog.description": "This agent will be removed from the active roster list.",
@@ -354,6 +372,7 @@
   "roster.empty": "No agent yet",
   "roster.emptySearch": "No matching agents",
   "roster.filters.all": "All",
+  "roster.filters.createdByMe": "Created by me",
   "roster.filters.drafts": "Drafts",
   "roster.filters.label": "Agent filters",
   "roster.filters.published": "Published",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent saved to roster.",
   "roster.searchLabel": "Search agents",
   "roster.searchPlaceholder": "Search agents by name…",
+  "roster.sort.earliestCreated": "Earliest created",
+  "roster.sort.label": "Sort by",
+  "roster.sort.lastModified": "Last modified",
+  "roster.sort.optionsLabel": "Sort options",
+  "roster.sort.recentlyCreated": "Recently created",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Human",
+  "roster.tabs.human": "Contacts",
   "roster.tabsLabel": "Roster type",
   "roster.updateSuccess": "Agent updated.",
   "roster.usageStatus.draft": "Draft"
diff --git a/web/i18n/en-US/common.json b/web/i18n/en-US/common.json
index 8eb0ca97f23..968801a2b7e 100644
--- a/web/i18n/en-US/common.json
+++ b/web/i18n/en-US/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Browse Marketplace to install one, or configure providers in settings.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Only compatible models are shown",
   "modelProvider.selector.showIncompatibleModels": "Show incompatible models",
+  "modelProvider.selector.suggestion": "Suggestion",
+  "modelProvider.selector.suggestionTip": "Recommended for agent tasks",
   "modelProvider.showModels": "Show Models",
   "modelProvider.speechToTextModel.key": "Speech-to-Text Model",
   "modelProvider.speechToTextModel.tip": "Set the default model for speech-to-text input in conversation.",
diff --git a/web/i18n/en-US/workflow.json b/web/i18n/en-US/workflow.json
index 4a85e3be106..246493094c9 100644
--- a/web/i18n/en-US/workflow.json
+++ b/web/i18n/en-US/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Publish the latest workflow and ensure a connected User Input node before configuring it as a tool.",
   "common.workflowAsToolTip": "Tool reconfiguration is required after the workflow update.",
   "common.workflowProcess": "Workflow Process",
+  "common.workflowProcessFailed": "Workflow Process failed",
+  "common.workflowProcessPaused": "Workflow Process paused",
+  "common.workflowProcessRunning": "Workflow Process running",
+  "common.workflowProcessSucceeded": "Workflow Process succeeded",
   "customWebhook": "Custom Webhook",
   "debug.copyLastRun": "Copy Last Run",
   "debug.copyLastRunError": "Failed to copy last run inputs",
diff --git a/web/i18n/es-ES/agent-v-2.json b/web/i18n/es-ES/agent-v-2.json
index 16e3362a5e9..67394756bc7 100644
--- a/web/i18n/es-ES/agent-v-2.json
+++ b/web/i18n/es-ES/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Mostrar valor de {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Tu agente se ejecuta en un entorno Linux. Puedes predefinir configuraciones o secretos definiendo variables de entorno aquí.\n\nComo alternativa, siempre puedes configurarlo mediante un chat de Build. Consulta la <docLink>documentación</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Define valores que el agente lee en tiempo de ejecución, como una clave API que necesita una herramienta, sin ponerlos en el Prompt.\n\nEn modo Build, el agente puede configurarlos por ti.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Seleccionar alcance de la variable de entorno",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Tu agente se ejecuta en un entorno Linux. Puedes predefinir configuraciones o secretos definiendo variables de entorno aquí. Como alternativa, siempre puedes configurarlo mediante un chat de Build. Consulta la documentación.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Define valores que el agente lee en tiempo de ejecución, como una clave API que necesita una herramienta, sin ponerlos en el Prompt.\n\nEn modo Build, el agente puede configurarlos por ti.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Alternar Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Crea tu agente chateando",
   "agentDetail.configure.build.inputPlaceholder": "Describe qué debe hacer tu agente",
   "agentDetail.configure.build.startBuild": "Iniciar compilación",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Una nota continua de esta compilación que el agente usa para conservar el contexto entre sesiones. Se añade a tus archivos al aplicar.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} cambio por aplicar",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} cambios por aplicar",
   "agentDetail.configure.buildDraft.discard": "Descartar",
   "agentDetail.configure.buildDraft.modeBadge": "Modo build",
   "agentDetail.configure.buildDraft.modeDescription": "Estás en modo build. Ajusta esta configuración con el chat de la derecha y luego aplica los cambios.",
+  "agentDetail.configure.buildDraft.rewritten": "Reescrito",
   "agentDetail.configure.buildDraft.title": "Borrador de compilación",
+  "agentDetail.configure.buildDraft.updated": "Actualizado",
   "agentDetail.configure.chatFeatures.description": "Da forma a la experiencia de chat del usuario final en tu webapp y superficies de chat.",
   "agentDetail.configure.chatFeatures.title": "Funciones de chat",
   "agentDetail.configure.files.add": "Agregar archivo",
+  "agentDetail.configure.files.buildNote.generated": "Generado",
+  "agentDetail.configure.files.buildNote.richTooltip": "El registro del agente sobre lo que configuró en modo Build. Lo lee al inicio de cada conversación, junto con tu Prompt. <docLink>Más información</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "El registro del agente sobre lo que configuró en modo Build. Lo lee al inicio de cada conversación, junto con tu Prompt. Más información",
   "agentDetail.configure.files.empty.description": "Sube documentos que el agente pueda leer, como especificaciones, plantillas o guías",
   "agentDetail.configure.files.empty.title": "Aún no hay archivos",
   "agentDetail.configure.files.label": "Archivos",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Error al cargar la vista previa.",
   "agentDetail.configure.files.preview.unsupported": "Este archivo no admite vista previa.",
   "agentDetail.configure.files.remove": "Eliminar {{name}}",
-  "agentDetail.configure.files.tip": "Proporciona a tu agente cualquier ejemplo, guía o simplemente un soporte de información que sea demasiado largo para el prompt. Haz referencia explícita con / en el prompt.",
+  "agentDetail.configure.files.tip": "Para material de referencia demasiado largo para pegar en el Prompt: especificaciones, plantillas, guías. Haz referencia con / en el Prompt.\n\nEn modo Build, el agente puede configurarlos por ti.",
   "agentDetail.configure.files.toggle": "Alternar archivos",
   "agentDetail.configure.files.treeLabel": "Archivos del agente",
   "agentDetail.configure.files.upload.action": "Subir",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Eliminar {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Fuentes de recuperación de conocimiento que este agente puede usar al orquestar tareas.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Conecta bases de conocimiento que el agente consulta mientras trabaja, para que sus respuestas se basen en tu contenido. <docLink>Más información</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Conecta bases de conocimiento que el agente consulta mientras trabaja, para que sus respuestas se basen en tu contenido. Más información",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Alternar recuperación de conocimiento",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Configura un modelo en el filtrado de metadatos",
   "agentDetail.configure.model.label": "Modelo",
   "agentDetail.configure.orchestrate": "Orquestar",
   "agentDetail.configure.preview.chatFeatures": "Funciones de chat",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Autenticación de usuario final",
   "agentDetail.configure.preview.inputPlaceholder": "Enviar mensaje a {{name}}",
   "agentDetail.configure.preview.restart": "Reiniciar vista previa",
+  "agentDetail.configure.preview.sandboxNotice": "El agente se ejecuta en un sandbox de Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "En Dify Community Edition, cada uno de tus agentes se ejecuta en un entorno sandbox Linux 7.0.0-10060-aws dentro de tu Docker. Tus cambios en el entorno mediante Build Chats son persistentes.",
   "agentDetail.configure.preview.title": "Vista previa",
   "agentDetail.configure.prompt.copied": "Prompt copiado",
   "agentDetail.configure.prompt.copy": "Copiar prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Más información",
   "agentDetail.configure.rightPanel.modeLabel": "Modo de configuración del agente",
   "agentDetail.configure.rightPanel.preview": "Vista previa",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview solo está disponible en Dify Cloud y Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview ejecuta el agente terminado como lo verán tus usuarios, con respuestas claras y funciones de chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Previsualiza tu agente",
   "agentDetail.configure.skills.add": "Agregar habilidad",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Habilidad",
   "agentDetail.configure.skills.label": "Habilidades",
   "agentDetail.configure.skills.remove": "Eliminar {{name}}",
-  "agentDetail.configure.skills.richTip": "Equipa a tu agente con capacidades modulares. Consulta la <docLink>documentación</docLink>. Haz referencia explícita con / en el prompt.",
-  "agentDetail.configure.skills.tip": "Equipa a tu agente con capacidades modulares. Consulta la documentación. Haz referencia explícita con / en el prompt.",
+  "agentDetail.configure.skills.richTip": "Empaqueta las instrucciones, archivos y scripts de una tarea recurrente en una skill. Haz referencia con / en el Prompt. <docLink>Más información</docLink>\n\nEn modo Build, el agente puede configurarlas por ti.",
+  "agentDetail.configure.skills.tip": "Empaqueta las instrucciones, archivos y scripts de una tarea recurrente en una skill. Haz referencia con / en el Prompt. Más información\n\nEn modo Build, el agente puede configurarlas por ti.",
   "agentDetail.configure.skills.toggle": "Alternar habilidades",
   "agentDetail.configure.skills.upload.action": "Subir",
   "agentDetail.configure.skills.upload.browse": "examinar",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Más acciones para {{name}}",
   "agentDetail.configure.tools.removeAction": "Eliminar {{name}}",
   "agentDetail.configure.tools.removeProvider": "Eliminar todas las herramientas",
-  "agentDetail.configure.tools.richTip": "Equipa a tu agente con los plugins de Dify. Consulta la <pluginDocLink>documentación</pluginDocLink>. Haz referencia explícita con / en el prompt.\nComo alternativa, siempre puedes indicar al agente que instale y autentique otras herramientas (por ejemplo, MCPs y herramientas CLI) mediante un chat de Build. Ten en cuenta que las herramientas configuradas mediante chats de Build NO aparecerán aquí, aunque seguirán estando disponibles para tu agente en adelante. Consulta la <buildDocLink>documentación</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Equipa a tu agente con los plugins de Dify. Consulta la documentación. Haz referencia explícita con / en el prompt. Como alternativa, siempre puedes indicar al agente que instale y autentique otras herramientas (por ejemplo, MCPs y herramientas CLI) mediante un chat de Build. Ten en cuenta que las herramientas configuradas mediante chats de Build NO aparecerán aquí, aunque seguirán estando disponibles para tu agente en adelante. Consulta la documentación.",
+  "agentDetail.configure.tools.richTip": "Añade herramientas de Dify desde tu espacio de trabajo. Haz referencia con / en el Prompt. <docLink>Más información</docLink>\n\nLas herramientas que el agente instale en su sandbox no aparecerán aquí, pero seguirán disponibles para él.",
+  "agentDetail.configure.tools.tip": "Añade herramientas de Dify desde tu espacio de trabajo. Haz referencia con / en el Prompt. Más información\n\nLas herramientas que el agente instale en su sandbox no aparecerán aquí, pero seguirán disponibles para él.",
   "agentDetail.configure.tools.toggle": "Alternar herramientas",
   "agentDetail.configure.tools.toolTabs.all": "Todas",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugins",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Ruta del sistema de archivos",
   "agentDetail.configure.workingDirectory.description": "Archivos que el agente lee y escribe mientras se ejecuta. Todo lo que genere, como qna_report.pdf, aparecerá aquí.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Sistema de archivos",
+  "agentDetail.configure.workingDirectory.home": "inicio",
   "agentDetail.configure.workingDirectory.open": "Abrir directorio de trabajo",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Volver al directorio de trabajo",
   "agentDetail.configure.workingDirectory.title": "Directorio de trabajo",
   "agentDetail.configure.workingDirectory.treeLabel": "Archivos del directorio de trabajo",
+  "agentDetail.configure.workingDirectory.workingDirectory": "directorio de trabajo",
   "agentDetail.documentTitle": "Agente",
   "agentDetail.logs.description": "Los registros completos registran el estado de ejecución de la aplicación, incluidas las entradas del usuario, las respuestas del agente, la planificación y los usos de herramientas.",
   "agentDetail.logs.empty": "No se encontraron registros",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "El nombre es obligatorio.",
   "roster.createForm.roleLabel": "Rol",
   "roster.createForm.rolePlaceholder": "p. ej. Asistente de investigación",
-  "roster.createForm.roleRequired": "El rol es obligatorio.",
   "roster.createSuccess": "Agente creado.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Este agente se eliminará de la lista del roster activo.",
@@ -354,6 +372,7 @@
   "roster.empty": "Aún no hay agentes",
   "roster.emptySearch": "No hay agentes coincidentes",
   "roster.filters.all": "Todos",
+  "roster.filters.createdByMe": "Creado por mí",
   "roster.filters.drafts": "Borradores",
   "roster.filters.label": "Filtros de agente",
   "roster.filters.published": "Publicados",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agente guardado en el roster.",
   "roster.searchLabel": "Buscar agentes",
   "roster.searchPlaceholder": "Buscar agentes por nombre…",
+  "roster.sort.earliestCreated": "Creados primero",
+  "roster.sort.label": "Ordenar por",
+  "roster.sort.lastModified": "Última modificación",
+  "roster.sort.optionsLabel": "Opciones de ordenación",
+  "roster.sort.recentlyCreated": "Creados recientemente",
   "roster.tabs.agent": "Agente",
-  "roster.tabs.human": "Humano",
+  "roster.tabs.human": "Contactos",
   "roster.tabsLabel": "Tipo de roster",
   "roster.updateSuccess": "Agente actualizado.",
   "roster.usageStatus.draft": "Borrador"
diff --git a/web/i18n/es-ES/common.json b/web/i18n/es-ES/common.json
index 990ebba61b6..ce0d0803ba7 100644
--- a/web/i18n/es-ES/common.json
+++ b/web/i18n/es-ES/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Explora el Marketplace para instalar uno, o configura proveedores en los ajustes.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Solo se muestran los modelos compatibles",
   "modelProvider.selector.showIncompatibleModels": "Mostrar modelos incompatibles",
+  "modelProvider.selector.suggestion": "Sugerencia",
+  "modelProvider.selector.suggestionTip": "Recomendado para tareas de agente",
   "modelProvider.showModels": "Mostrar Modelos",
   "modelProvider.speechToTextModel.key": "Modelo de Voz a Texto",
   "modelProvider.speechToTextModel.tip": "Establece el modelo predeterminado para la entrada de voz a texto en la conversación.",
diff --git a/web/i18n/es-ES/workflow.json b/web/i18n/es-ES/workflow.json
index 2f128eb161d..65f9d2db0bc 100644
--- a/web/i18n/es-ES/workflow.json
+++ b/web/i18n/es-ES/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Publica el flujo de trabajo más reciente y asegúrate de que haya un nodo de Entrada de Usuario conectado antes de configurarlo como una herramienta.",
   "common.workflowAsToolTip": "Se requiere la reconfiguración de la herramienta después de la actualización del flujo de trabajo.",
   "common.workflowProcess": "Proceso de flujo de trabajo",
+  "common.workflowProcessFailed": "Proceso de flujo de trabajo fallido",
+  "common.workflowProcessPaused": "Proceso de flujo de trabajo pausado",
+  "common.workflowProcessRunning": "Proceso de flujo de trabajo en ejecución",
+  "common.workflowProcessSucceeded": "Proceso de flujo de trabajo completado correctamente",
   "customWebhook": "Webhook personalizado",
   "debug.copyLastRun": "Copiar última ejecución",
   "debug.copyLastRunError": "No se pudo copiar las entradas de la última ejecución",
diff --git a/web/i18n/fa-IR/agent-v-2.json b/web/i18n/fa-IR/agent-v-2.json
index c9789d50909..1dcab1bff12 100644
--- a/web/i18n/fa-IR/agent-v-2.json
+++ b/web/i18n/fa-IR/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "نمایش مقدار {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "عامل شما در یک محیط Linux اجرا می‌شود. می‌توانید با تعریف متغیرهای محیطی در اینجا، پیکربندی‌ها یا اسرار را از قبل تنظیم کنید.\n\nهمچنین همیشه می‌توانید تنظیمات را از طریق گفتگوی Build انجام دهید. به <docLink>مستندات</docLink> مراجعه کنید.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "مقادیرى را تنظیم کنید که عامل در زمان اجرا می‌خواند، مانند کلید API موردنیاز یک ابزار، بدون اینکه آن‌ها را در Prompt قرار دهید.\n\nدر حالت Build، عامل می‌تواند این موارد را برای شما تنظیم کند.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "انتخاب دامنه متغیر محیطی",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "عامل شما در یک محیط Linux اجرا می‌شود. می‌توانید با تعریف متغیرهای محیطی در اینجا، پیکربندی‌ها یا اسرار را از قبل تنظیم کنید. همچنین همیشه می‌توانید تنظیمات را از طریق گفتگوی Build انجام دهید. به مستندات مراجعه کنید.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "مقادیرى را تنظیم کنید که عامل در زمان اجرا می‌خواند، مانند کلید API موردنیاز یک ابزار، بدون اینکه آن‌ها را در Prompt قرار دهید.\n\nدر حالت Build، عامل می‌تواند این موارد را برای شما تنظیم کند.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "تغییر Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "عامل خود را با چت بسازید",
   "agentDetail.configure.build.inputPlaceholder": "توضیح دهید عامل شما باید چه کاری انجام دهد",
   "agentDetail.configure.build.startBuild": "شروع ساخت",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "یک یادداشت پیوسته از این ساخت که عامل برای نگه‌داشتن زمینه بین نشست‌ها از آن استفاده می‌کند. هنگام اعمال، به فایل‌های شما اضافه می‌شود.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} تغییر برای اعمال",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} تغییر برای اعمال",
   "agentDetail.configure.buildDraft.discard": "رد کردن",
   "agentDetail.configure.buildDraft.modeBadge": "حالت ساخت",
   "agentDetail.configure.buildDraft.modeDescription": "شما در حالت ساخت هستید. این تنظیمات را از طریق چت سمت راست شکل دهید، سپس اعمال کنید.",
+  "agentDetail.configure.buildDraft.rewritten": "بازنویسی شد",
   "agentDetail.configure.buildDraft.title": "پیش نویس ساخت",
+  "agentDetail.configure.buildDraft.updated": "به‌روزرسانی شد",
   "agentDetail.configure.chatFeatures.description": "تجربه چت کاربر نهایی را در Web app و سطوح چت خود شکل دهید.",
   "agentDetail.configure.chatFeatures.title": "ویژگی‌های چت",
   "agentDetail.configure.files.add": "افزودن فایل",
+  "agentDetail.configure.files.buildNote.generated": "تولید شده",
+  "agentDetail.configure.files.buildNote.richTooltip": "رکورد عامل از چیزهایی که در حالت Build تنظیم کرده است. در آغاز هر گفتگو، آن را همراه با Prompt شما می‌خواند. <docLink>بیشتر بدانید</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "رکورد عامل از چیزهایی که در حالت Build تنظیم کرده است. در آغاز هر گفتگو، آن را همراه با Prompt شما می‌خواند. بیشتر بدانید",
   "agentDetail.configure.files.empty.description": "اسنادی را که عامل می‌تواند بخواند بارگذاری کنید، مانند مشخصات، قالب‌ها یا دستورالعمل‌ها",
   "agentDetail.configure.files.empty.title": "هنوز فایلی وجود ندارد",
   "agentDetail.configure.files.label": "فایل‌ها",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "بارگذاری پیش‌نمایش ناموفق بود.",
   "agentDetail.configure.files.preview.unsupported": "این فایل از پیش‌نمایش پشتیبانی نمی‌کند.",
   "agentDetail.configure.files.remove": "حذف {{name}}",
-  "agentDetail.configure.files.tip": "هر نمونه، دستورالعمل یا حامل اطلاعاتی را که برای پرامپت بیش از حد طولانی است در اختیار عامل قرار دهید. با / در پرامپت به‌صورت صریح به آن ارجاع دهید.",
+  "agentDetail.configure.files.tip": "برای منابع مرجعی که برای چسباندن در Prompt بیش از حد طولانی‌اند: مشخصات، قالب‌ها، دستورالعمل‌ها. در Prompt با / به آن‌ها ارجاع دهید.\n\nدر حالت Build، عامل می‌تواند این موارد را برای شما تنظیم کند.",
   "agentDetail.configure.files.toggle": "تغییر فایل‌ها",
   "agentDetail.configure.files.treeLabel": "فایل‌های عامل",
   "agentDetail.configure.files.upload.action": "بارگذاری",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "حذف {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "منابع بازیابی دانش که این عامل هنگام هماهنگی وظایف می‌تواند استفاده کند.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "پایگاه‌های دانشی را وصل کنید که عامل هنگام کار جستجو می‌کند تا پاسخ‌هایش بر محتوای شما تکیه کند. <docLink>بیشتر بدانید</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "پایگاه‌های دانشی را وصل کنید که عامل هنگام کار جستجو می‌کند تا پاسخ‌هایش بر محتوای شما تکیه کند. بیشتر بدانید",
   "agentDetail.configure.knowledgeRetrieval.toggle": "تغییر بازیابی دانش",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "لطفاً یک مدل در فیلتر کردن فراداده پیکربندی کنید",
   "agentDetail.configure.model.label": "مدل",
   "agentDetail.configure.orchestrate": "هماهنگ‌سازی",
   "agentDetail.configure.preview.chatFeatures": "ویژگی‌های چت",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "احراز هویت کاربر نهایی",
   "agentDetail.configure.preview.inputPlaceholder": "ارسال پیام به {{name}}",
   "agentDetail.configure.preview.restart": "راه‌اندازی مجدد پیش‌نمایش",
+  "agentDetail.configure.preview.sandboxNotice": "Agent در یک سندباکس لینوکس اجرا می‌شود.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "در Dify Community Edition، هر یک از Agentهای شما در یک محیط سندباکس Linux 7.0.0-10060-aws داخل Docker شما اجرا می‌شود. ویرایش‌های شما روی محیط از طریق Build Chats پایدار می‌ماند.",
   "agentDetail.configure.preview.title": "پیش‌نمایش",
   "agentDetail.configure.prompt.copied": "پرامپت کپی شد",
   "agentDetail.configure.prompt.copy": "کپی پرامپت",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "بیشتر بدانید",
   "agentDetail.configure.rightPanel.modeLabel": "حالت پیکربندی عامل",
   "agentDetail.configure.rightPanel.preview": "پیش‌نمایش",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview فقط در Dify Cloud و Enterprise در دسترس است.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview عامل تکمیل‌شده را همان‌طور که کاربران می‌بینند اجرا می‌کند، با پاسخ‌های تمیز و قابلیت‌های گفتگو.",
   "agentDetail.configure.rightPanel.previewTipTitle": "پیش‌نمایش عامل",
   "agentDetail.configure.skills.add": "افزودن مهارت",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "مهارت",
   "agentDetail.configure.skills.label": "مهارت‌ها",
   "agentDetail.configure.skills.remove": "حذف {{name}}",
-  "agentDetail.configure.skills.richTip": "عامل خود را به قابلیت‌های ماژولار مجهز کنید. به <docLink>مستندات</docLink> مراجعه کنید. با / در پرامپت به‌صورت صریح ارجاع دهید.",
-  "agentDetail.configure.skills.tip": "عامل خود را به قابلیت‌های ماژولار مجهز کنید. به مستندات مراجعه کنید. با / در پرامپت به‌صورت صریح ارجاع دهید.",
+  "agentDetail.configure.skills.richTip": "دستورالعمل‌ها، فایل‌ها و اسکریپت‌های یک کار تکرارشونده را در یک Skill بسته‌بندی کنید. در Prompt با / به آن ارجاع دهید. <docLink>بیشتر بدانید</docLink>\n\nدر حالت Build، عامل می‌تواند این موارد را برای شما تنظیم کند.",
+  "agentDetail.configure.skills.tip": "دستورالعمل‌ها، فایل‌ها و اسکریپت‌های یک کار تکرارشونده را در یک Skill بسته‌بندی کنید. در Prompt با / به آن ارجاع دهید. بیشتر بدانید\n\nدر حالت Build، عامل می‌تواند این موارد را برای شما تنظیم کند.",
   "agentDetail.configure.skills.toggle": "تغییر مهارت‌ها",
   "agentDetail.configure.skills.upload.action": "بارگذاری",
   "agentDetail.configure.skills.upload.browse": "مرور",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "اقدامات بیشتر برای {{name}}",
   "agentDetail.configure.tools.removeAction": "حذف {{name}}",
   "agentDetail.configure.tools.removeProvider": "حذف همه ابزارها",
-  "agentDetail.configure.tools.richTip": "عامل خود را به افزونه‌های Dify مجهز کنید. به <pluginDocLink>مستندات</pluginDocLink> مراجعه کنید. با / در پرامپت به‌صورت صریح ارجاع دهید.\nهمچنین همیشه می‌توانید از طریق چت Build به عامل بگویید ابزارهای دیگر (مانند MCPها و ابزارهای CLI) را نصب و احراز هویت کند. توجه کنید ابزارهایی که از طریق چت‌های Build پیکربندی می‌شوند اینجا نمایش داده نمی‌شوند، اما همچنان در ادامه برای عامل شما در دسترس خواهند بود. به <buildDocLink>مستندات</buildDocLink> مراجعه کنید.",
-  "agentDetail.configure.tools.tip": "عامل خود را به افزونه‌های Dify مجهز کنید. به مستندات مراجعه کنید. با / در پرامپت به‌صورت صریح ارجاع دهید. همچنین همیشه می‌توانید از طریق چت Build به عامل بگویید ابزارهای دیگر (مانند MCPها و ابزارهای CLI) را نصب و احراز هویت کند. توجه کنید ابزارهایی که از طریق چت‌های Build پیکربندی می‌شوند اینجا نمایش داده نمی‌شوند، اما همچنان در ادامه برای عامل شما در دسترس خواهند بود. به مستندات مراجعه کنید.",
+  "agentDetail.configure.tools.richTip": "ابزارهای Dify را از workspace خود اضافه کنید. در Prompt با / به آن‌ها ارجاع دهید. <docLink>بیشتر بدانید</docLink>\n\nابزارهایی که عامل در sandbox خود نصب می‌کند اینجا نمایش داده نمی‌شوند، اما برای او در دسترس می‌مانند.",
+  "agentDetail.configure.tools.tip": "ابزارهای Dify را از workspace خود اضافه کنید. در Prompt با / به آن‌ها ارجاع دهید. بیشتر بدانید\n\nابزارهایی که عامل در sandbox خود نصب می‌کند اینجا نمایش داده نمی‌شوند، اما برای او در دسترس می‌مانند.",
   "agentDetail.configure.tools.toggle": "تغییر ابزارها",
   "agentDetail.configure.tools.toolTabs.all": "همه",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "افزونه‌ها",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "مسیر سیستم فایل",
   "agentDetail.configure.workingDirectory.description": "فایل‌هایی که عامل هنگام اجرا می‌خواند و می‌نویسد. هر چیزی که تولید کند، مثل qna_report.pdf، اینجا نمایش داده می‌شود.",
+  "agentDetail.configure.workingDirectory.fileSystem": "سیستم فایل",
+  "agentDetail.configure.workingDirectory.home": "خانه",
   "agentDetail.configure.workingDirectory.open": "باز کردن پوشه کاری",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "بازگشت به پوشه کاری",
   "agentDetail.configure.workingDirectory.title": "پوشه کاری",
   "agentDetail.configure.workingDirectory.treeLabel": "فایل‌های پوشه کاری",
+  "agentDetail.configure.workingDirectory.workingDirectory": "پوشه کاری",
   "agentDetail.documentTitle": "عامل",
   "agentDetail.logs.description": "گزارش‌های کامل وضعیت اجرای برنامه را ثبت می‌کنند، از جمله ورودی‌های کاربر، پاسخ‌های عامل، برنامه‌ریزی و استفاده از ابزار.",
   "agentDetail.logs.empty": "هیچ گزارشی یافت نشد",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "نام الزامی است.",
   "roster.createForm.roleLabel": "نقش",
   "roster.createForm.rolePlaceholder": "مثلاً دستیار پژوهش",
-  "roster.createForm.roleRequired": "نقش الزامی است.",
   "roster.createSuccess": "عامل ایجاد شد.",
   "roster.dateTimeFormat": "YYYY/MM/DD HH:mm:ss",
   "roster.deleteDialog.description": "این عامل از فهرست Roster فعال حذف خواهد شد.",
@@ -354,6 +372,7 @@
   "roster.empty": "هنوز عاملی وجود ندارد",
   "roster.emptySearch": "هیچ عامل مطابقی نیست",
   "roster.filters.all": "همه",
+  "roster.filters.createdByMe": "ایجادشده توسط من",
   "roster.filters.drafts": "پیش‌نویس‌ها",
   "roster.filters.label": "فیلترهای عامل",
   "roster.filters.published": "منتشر شده",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "عامل در Roster ذخیره شد.",
   "roster.searchLabel": "جستجوی عوامل",
   "roster.searchPlaceholder": "جستجوی عوامل بر اساس نام…",
+  "roster.sort.earliestCreated": "قدیمی‌ترین ایجاد",
+  "roster.sort.label": "مرتب‌سازی بر اساس",
+  "roster.sort.lastModified": "آخرین تغییر",
+  "roster.sort.optionsLabel": "گزینه‌های مرتب‌سازی",
+  "roster.sort.recentlyCreated": "تازه‌ترین ایجاد",
   "roster.tabs.agent": "عامل",
-  "roster.tabs.human": "انسان",
+  "roster.tabs.human": "مخاطبین",
   "roster.tabsLabel": "نوع Roster",
   "roster.updateSuccess": "عامل به‌روزرسانی شد.",
   "roster.usageStatus.draft": "پیش‌نویس"
diff --git a/web/i18n/fa-IR/common.json b/web/i18n/fa-IR/common.json
index a6b079e6f15..4c9dc6821b9 100644
--- a/web/i18n/fa-IR/common.json
+++ b/web/i18n/fa-IR/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "در Marketplace جستجو کنید تا یکی نصب کنید، یا ارائه‌دهندگان را در تنظیمات پیکربندی کنید.",
   "modelProvider.selector.onlyCompatibleModelsShown": "فقط مدل‌های سازگار نمایش داده می‌شوند",
   "modelProvider.selector.showIncompatibleModels": "نمایش مدل‌های ناسازگار",
+  "modelProvider.selector.suggestion": "پیشنهاد",
+  "modelProvider.selector.suggestionTip": "پیشنهادی برای وظایف عامل",
   "modelProvider.showModels": "نمایش مدل‌ها",
   "modelProvider.speechToTextModel.key": "مدل تبدیل گفتار به متن",
   "modelProvider.speechToTextModel.tip": "مدل پیش‌فرض را برای ورودی گفتار به متن در مکالمه تنظیم کنید.",
diff --git a/web/i18n/fa-IR/workflow.json b/web/i18n/fa-IR/workflow.json
index 20b56d009a4..e6136fe301f 100644
--- a/web/i18n/fa-IR/workflow.json
+++ b/web/i18n/fa-IR/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "برای تنظیم به عنوان ابزار، ابتدا گردش کار را منتشر کنید و مطمئن شوید که گره ورودی کاربر متصل است.",
   "common.workflowAsToolTip": "پس از به‌روزرسانی گردش کار، پیکربندی مجدد ابزار الزامی است.",
   "common.workflowProcess": "فرآیند گردش کار",
+  "common.workflowProcessFailed": "فرآیند گردش کار ناموفق بود",
+  "common.workflowProcessPaused": "فرآیند گردش کار متوقف شده است",
+  "common.workflowProcessRunning": "فرآیند گردش کار در حال اجراست",
+  "common.workflowProcessSucceeded": "فرآیند گردش کار با موفقیت انجام شد",
   "customWebhook": "وب‌هوک سفارشی",
   "debug.copyLastRun": "کپی آخرین اجرا",
   "debug.copyLastRunError": "کپی ورودی‌های آخرین اجرا ناموفق بود",
diff --git a/web/i18n/fr-FR/agent-v-2.json b/web/i18n/fr-FR/agent-v-2.json
index de47d784c63..697e50cca8c 100644
--- a/web/i18n/fr-FR/agent-v-2.json
+++ b/web/i18n/fr-FR/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Afficher la valeur de {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Votre agent s’exécute dans un environnement Linux. Vous pouvez prédéfinir des configurations ou des secrets en définissant ici des variables d’environnement.\n\nVous pouvez aussi toujours faire la configuration via un chat Build. Consultez la <docLink>documentation</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Définissez les valeurs que l'agent lit à l'exécution, comme une clé API nécessaire à un outil, sans les mettre dans le Prompt.\n\nEn mode Build, l'agent peut les configurer pour vous.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Sélectionner la portée de la variable d’environnement",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Votre agent s’exécute dans un environnement Linux. Vous pouvez prédéfinir des configurations ou des secrets en définissant ici des variables d’environnement. Vous pouvez aussi toujours faire la configuration via un chat Build. Consultez la documentation.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Définissez les valeurs que l'agent lit à l'exécution, comme une clé API nécessaire à un outil, sans les mettre dans le Prompt.\n\nEn mode Build, l'agent peut les configurer pour vous.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Basculer Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Créez votre agent par chat",
   "agentDetail.configure.build.inputPlaceholder": "Décrivez ce que votre agent doit faire",
   "agentDetail.configure.build.startBuild": "Démarrer la création",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Une note continue de ce build que l'agent utilise pour conserver le contexte entre les sessions. Elle est ajoutée à vos fichiers lorsque vous appliquez.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} modification à appliquer",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} modifications à appliquer",
   "agentDetail.configure.buildDraft.discard": "Ignorer",
   "agentDetail.configure.buildDraft.modeBadge": "Mode build",
   "agentDetail.configure.buildDraft.modeDescription": "Vous êtes en mode build. Ajustez cette configuration avec le chat à droite, puis appliquez.",
+  "agentDetail.configure.buildDraft.rewritten": "Réécrit",
   "agentDetail.configure.buildDraft.title": "Brouillon de build",
+  "agentDetail.configure.buildDraft.updated": "Mis à jour",
   "agentDetail.configure.chatFeatures.description": "Façonnez l’expérience de chat de l’utilisateur final sur votre webapp et vos surfaces de chat.",
   "agentDetail.configure.chatFeatures.title": "Fonctionnalités de chat",
   "agentDetail.configure.files.add": "Ajouter un fichier",
+  "agentDetail.configure.files.buildNote.generated": "Généré",
+  "agentDetail.configure.files.buildNote.richTooltip": "Le registre de l'agent sur ce qu'il a configuré en mode Build. Il le lit au début de chaque conversation, avec votre Prompt. <docLink>En savoir plus</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Le registre de l'agent sur ce qu'il a configuré en mode Build. Il le lit au début de chaque conversation, avec votre Prompt. En savoir plus",
   "agentDetail.configure.files.empty.description": "Téléchargez des documents que l’agent peut lire, comme des spécifications, des modèles ou des directives",
   "agentDetail.configure.files.empty.title": "Pas encore de fichiers",
   "agentDetail.configure.files.label": "Fichiers",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Échec du chargement de l’aperçu.",
   "agentDetail.configure.files.preview.unsupported": "Ce fichier ne prend pas en charge l’aperçu.",
   "agentDetail.configure.files.remove": "Supprimer {{name}}",
-  "agentDetail.configure.files.tip": "Fournissez à votre agent tout exemple, consigne ou simple support d’information trop long pour le prompt. Référencez-le explicitement avec / dans le prompt.",
+  "agentDetail.configure.files.tip": "Pour les documents de référence trop longs à coller dans le Prompt : spécifications, modèles, consignes. Référencez-les avec / dans le Prompt.\n\nEn mode Build, l'agent peut les configurer pour vous.",
   "agentDetail.configure.files.toggle": "Basculer les fichiers",
   "agentDetail.configure.files.treeLabel": "Fichiers de l’agent",
   "agentDetail.configure.files.upload.action": "Téléverser",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Supprimer {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Sources de récupération de connaissances que cet agent peut utiliser lors de l’orchestration des tâches.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Connectez les bases de connaissances que l'agent consulte pendant son travail, afin que ses réponses s'appuient sur votre contenu. <docLink>En savoir plus</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Connectez les bases de connaissances que l'agent consulte pendant son travail, afin que ses réponses s'appuient sur votre contenu. En savoir plus",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Basculer la récupération de connaissances",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Veuillez configurer un modèle dans le filtrage des métadonnées",
   "agentDetail.configure.model.label": "Modèle",
   "agentDetail.configure.orchestrate": "Orchestrer",
   "agentDetail.configure.preview.chatFeatures": "Fonctionnalités de chat",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Authentification de l’utilisateur final",
   "agentDetail.configure.preview.inputPlaceholder": "Envoyer un message à {{name}}",
   "agentDetail.configure.preview.restart": "Redémarrer l’aperçu",
+  "agentDetail.configure.preview.sandboxNotice": "L’agent s’exécute dans un bac à sable Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Dans Dify Community Edition, chacun de vos agents s’exécute dans un environnement bac à sable Linux 7.0.0-10060-aws au sein de votre Docker. Les modifications apportées à l’environnement via Build Chats sont persistantes.",
   "agentDetail.configure.preview.title": "Aperçu",
   "agentDetail.configure.prompt.copied": "Prompt copié",
   "agentDetail.configure.prompt.copy": "Copier le prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "En savoir plus",
   "agentDetail.configure.rightPanel.modeLabel": "Mode de configuration de l’agent",
   "agentDetail.configure.rightPanel.preview": "Aperçu",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview est disponible uniquement sur Dify Cloud et Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview exécute l’agent final comme vos utilisateurs le verront, avec des réponses claires et les fonctions de chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Prévisualiser votre agent",
   "agentDetail.configure.skills.add": "Ajouter une compétence",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Compétence",
   "agentDetail.configure.skills.label": "Compétences",
   "agentDetail.configure.skills.remove": "Supprimer {{name}}",
-  "agentDetail.configure.skills.richTip": "Équipez votre agent de capacités modulaires. Consultez la <docLink>documentation</docLink>. Référencez-les explicitement avec / dans le prompt.",
-  "agentDetail.configure.skills.tip": "Équipez votre agent de capacités modulaires. Consultez la documentation. Référencez-les explicitement avec / dans le prompt.",
+  "agentDetail.configure.skills.richTip": "Regroupez les instructions, fichiers et scripts d'une tâche récurrente dans une skill. Référencez-la avec / dans le Prompt. <docLink>En savoir plus</docLink>\n\nEn mode Build, l'agent peut les configurer pour vous.",
+  "agentDetail.configure.skills.tip": "Regroupez les instructions, fichiers et scripts d'une tâche récurrente dans une skill. Référencez-la avec / dans le Prompt. En savoir plus\n\nEn mode Build, l'agent peut les configurer pour vous.",
   "agentDetail.configure.skills.toggle": "Basculer les compétences",
   "agentDetail.configure.skills.upload.action": "Téléverser",
   "agentDetail.configure.skills.upload.browse": "parcourir",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Plus d’actions pour {{name}}",
   "agentDetail.configure.tools.removeAction": "Supprimer {{name}}",
   "agentDetail.configure.tools.removeProvider": "Supprimer tous les outils",
-  "agentDetail.configure.tools.richTip": "Équipez votre agent avec les plugins de Dify. Consultez la <pluginDocLink>documentation</pluginDocLink>. Référencez-les explicitement avec / dans le prompt.\nVous pouvez aussi toujours demander à l’agent d’installer et d’authentifier d’autres outils (par exemple des MCPs et des outils CLI) via un chat Build. Notez que les outils configurés via les chats Build n’apparaîtront PAS ici, même s’ils resteront disponibles pour votre agent par la suite. Consultez la <buildDocLink>documentation</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Équipez votre agent avec les plugins de Dify. Consultez la documentation. Référencez-les explicitement avec / dans le prompt. Vous pouvez aussi toujours demander à l’agent d’installer et d’authentifier d’autres outils (par exemple des MCPs et des outils CLI) via un chat Build. Notez que les outils configurés via les chats Build n’apparaîtront PAS ici, même s’ils resteront disponibles pour votre agent par la suite. Consultez la documentation.",
+  "agentDetail.configure.tools.richTip": "Ajoutez les outils Dify de votre espace de travail. Référencez-les avec / dans le Prompt. <docLink>En savoir plus</docLink>\n\nLes outils que l'agent installe dans son sandbox n'apparaîtront pas ici, mais resteront disponibles.",
+  "agentDetail.configure.tools.tip": "Ajoutez les outils Dify de votre espace de travail. Référencez-les avec / dans le Prompt. En savoir plus\n\nLes outils que l'agent installe dans son sandbox n'apparaîtront pas ici, mais resteront disponibles.",
   "agentDetail.configure.tools.toggle": "Basculer les outils",
   "agentDetail.configure.tools.toolTabs.all": "Tous",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugins",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Chemin du système de fichiers",
   "agentDetail.configure.workingDirectory.description": "Fichiers que l’agent lit et écrit pendant son exécution. Tout ce qu’il génère, comme qna_report.pdf, apparaît ici.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Système de fichiers",
+  "agentDetail.configure.workingDirectory.home": "accueil",
   "agentDetail.configure.workingDirectory.open": "Ouvrir le répertoire de travail",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Retour au répertoire de travail",
   "agentDetail.configure.workingDirectory.title": "Répertoire de travail",
   "agentDetail.configure.workingDirectory.treeLabel": "Fichiers du répertoire de travail",
+  "agentDetail.configure.workingDirectory.workingDirectory": "répertoire de travail",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Les journaux complets enregistrent l’état d’exécution de l’application, y compris les entrées des utilisateurs, les réponses de l’agent, la planification et l’utilisation des outils.",
   "agentDetail.logs.empty": "Aucun journal trouvé",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Le nom est requis.",
   "roster.createForm.roleLabel": "Rôle",
   "roster.createForm.rolePlaceholder": "p. ex. Assistant de recherche",
-  "roster.createForm.roleRequired": "Le rôle est requis.",
   "roster.createSuccess": "Agent créé.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Cet agent sera supprimé de la liste du roster actif.",
@@ -354,6 +372,7 @@
   "roster.empty": "Pas encore d’agent",
   "roster.emptySearch": "Aucun agent correspondant",
   "roster.filters.all": "Tous",
+  "roster.filters.createdByMe": "Créés par moi",
   "roster.filters.drafts": "Brouillons",
   "roster.filters.label": "Filtres d’agent",
   "roster.filters.published": "Publiés",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent enregistré dans le roster.",
   "roster.searchLabel": "Rechercher des agents",
   "roster.searchPlaceholder": "Rechercher des agents par nom…",
+  "roster.sort.earliestCreated": "Créés en premier",
+  "roster.sort.label": "Trier par",
+  "roster.sort.lastModified": "Dernière modification",
+  "roster.sort.optionsLabel": "Options de tri",
+  "roster.sort.recentlyCreated": "Récemment créés",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Humain",
+  "roster.tabs.human": "Contacts",
   "roster.tabsLabel": "Type de roster",
   "roster.updateSuccess": "Agent mis à jour.",
   "roster.usageStatus.draft": "Brouillon"
diff --git a/web/i18n/fr-FR/common.json b/web/i18n/fr-FR/common.json
index d0bafc616df..4693c5b88fa 100644
--- a/web/i18n/fr-FR/common.json
+++ b/web/i18n/fr-FR/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Parcourez le Marketplace pour en installer un, ou configurez les fournisseurs dans les paramètres.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Seuls les modèles compatibles sont affichés",
   "modelProvider.selector.showIncompatibleModels": "Afficher les modèles incompatibles",
+  "modelProvider.selector.suggestion": "Suggestion",
+  "modelProvider.selector.suggestionTip": "Recommandé pour les tâches d’agent",
   "modelProvider.showModels": "Montrer les modèles",
   "modelProvider.speechToTextModel.key": "Modèle de Texte-à-Parole",
   "modelProvider.speechToTextModel.tip": "Définissez le modèle par défaut pour l'entrée de texte par la parole dans la conversation.",
diff --git a/web/i18n/fr-FR/workflow.json b/web/i18n/fr-FR/workflow.json
index 018fc7d9c17..a7f9ea73a95 100644
--- a/web/i18n/fr-FR/workflow.json
+++ b/web/i18n/fr-FR/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Publiez le dernier flux de travail et assurez-vous qu'un nœud d'entrée utilisateur est connecté avant de le configurer comme outil.",
   "common.workflowAsToolTip": "Reconfiguration de l'outil requise après la mise à jour du flux de travail.",
   "common.workflowProcess": "Processus de flux de travail",
+  "common.workflowProcessFailed": "Processus de flux de travail échoué",
+  "common.workflowProcessPaused": "Processus de flux de travail en pause",
+  "common.workflowProcessRunning": "Processus de flux de travail en cours d'exécution",
+  "common.workflowProcessSucceeded": "Processus de flux de travail réussi",
   "customWebhook": "Webhook personnalisé",
   "debug.copyLastRun": "Copier la dernière exécution",
   "debug.copyLastRunError": "Échec de la copie des entrées de la dernière exécution",
diff --git a/web/i18n/hi-IN/agent-v-2.json b/web/i18n/hi-IN/agent-v-2.json
index b4e94709e68..55acc170321 100644
--- a/web/i18n/hi-IN/agent-v-2.json
+++ b/web/i18n/hi-IN/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "{{key}} मान दिखाएँ",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "आपका एजेंट Linux environment में चलता है। आप यहां environmental variables परिभाषित करके configs या secrets preset कर सकते हैं।\n\nवैकल्पिक रूप से, आप Build chat के माध्यम से भी चीजें set up करवा सकते हैं। <docLink>docs</docLink> देखें।",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "ऐसे values सेट करें जिन्हें एजेंट runtime पर पढ़ता है, जैसे किसी tool को चाहिए API key, उन्हें Prompt में डाले बिना।\n\nBuild mode में, एजेंट इन्हें आपके लिए सेट कर सकता है।",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "पर्यावरण चर का दायरा चुनें",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "आपका एजेंट Linux environment में चलता है। आप यहां environmental variables परिभाषित करके configs या secrets preset कर सकते हैं। वैकल्पिक रूप से, आप Build chat के माध्यम से भी चीजें set up करवा सकते हैं। docs देखें।",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "ऐसे values सेट करें जिन्हें एजेंट runtime पर पढ़ता है, जैसे किसी tool को चाहिए API key, उन्हें Prompt में डाले बिना।\n\nBuild mode में, एजेंट इन्हें आपके लिए सेट कर सकता है।",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Env Editor टॉगल करें",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "चैट करके अपना एजेंट बनाएं",
   "agentDetail.configure.build.inputPlaceholder": "बताएं कि आपका एजेंट क्या करे",
   "agentDetail.configure.build.startBuild": "बिल्ड शुरू करें",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "इस बिल्ड का एक सतत नोट, जिसका उपयोग एजेंट सत्रों के बीच संदर्भ बनाए रखने के लिए करता है। लागू करने पर यह आपकी फ़ाइलों में जोड़ दिया जाता है।",
+  "agentDetail.configure.buildDraft.changesToApply_one": "लागू करने के लिए {{count}} बदलाव",
+  "agentDetail.configure.buildDraft.changesToApply_other": "लागू करने के लिए {{count}} बदलाव",
   "agentDetail.configure.buildDraft.discard": "छोड़ें",
   "agentDetail.configure.buildDraft.modeBadge": "बिल्ड मोड",
   "agentDetail.configure.buildDraft.modeDescription": "आप बिल्ड मोड में हैं। दाईं ओर की चैट से इस सेटअप को आकार दें, फिर लागू करें।",
+  "agentDetail.configure.buildDraft.rewritten": "फिर से लिखा गया",
   "agentDetail.configure.buildDraft.title": "बिल्ड ड्राफ्ट",
+  "agentDetail.configure.buildDraft.updated": "अपडेट किया गया",
   "agentDetail.configure.chatFeatures.description": "अपने Web app और चैट सतहों पर अंतिम-उपयोगकर्ता चैट अनुभव को आकार दें।",
   "agentDetail.configure.chatFeatures.title": "चैट सुविधाएँ",
   "agentDetail.configure.files.add": "फ़ाइल जोड़ें",
+  "agentDetail.configure.files.buildNote.generated": "जनरेट किया गया",
+  "agentDetail.configure.files.buildNote.richTooltip": "Build mode में एजेंट ने जो सेट अप किया उसका रिकॉर्ड। हर बातचीत की शुरुआत में यह इसे आपके Prompt के साथ पढ़ता है। <docLink>और जानें</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Build mode में एजेंट ने जो सेट अप किया उसका रिकॉर्ड। हर बातचीत की शुरुआत में यह इसे आपके Prompt के साथ पढ़ता है। और जानें",
   "agentDetail.configure.files.empty.description": "ऐसे दस्तावेज़ अपलोड करें जिन्हें एजेंट पढ़ सके, जैसे विनिर्देश, टेम्पलेट या दिशानिर्देश",
   "agentDetail.configure.files.empty.title": "अभी तक कोई फ़ाइल नहीं",
   "agentDetail.configure.files.label": "फ़ाइलें",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "पूर्वावलोकन लोड करने में विफल।",
   "agentDetail.configure.files.preview.unsupported": "यह फ़ाइल पूर्वावलोकन का समर्थन नहीं करती।",
   "agentDetail.configure.files.remove": "{{name}} हटाएँ",
-  "agentDetail.configure.files.tip": "अपने एजेंट को कोई उदाहरण, दिशा-निर्देश या ऐसी जानकारी दें जो प्रॉम्प्ट के लिए बहुत लंबी हो। प्रॉम्प्ट में / का उपयोग करके स्पष्ट संदर्भ दें।",
+  "agentDetail.configure.files.tip": "Prompt में चिपकाने के लिए बहुत लंबी संदर्भ सामग्री के लिए: स्पेक्स, टेम्पलेट, दिशानिर्देश। Prompt में / से संदर्भ दें।\n\nBuild mode में, एजेंट इन्हें आपके लिए सेट कर सकता है।",
   "agentDetail.configure.files.toggle": "फ़ाइलें टॉगल करें",
   "agentDetail.configure.files.treeLabel": "एजेंट फ़ाइलें",
   "agentDetail.configure.files.upload.action": "अपलोड करें",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "{{name}} हटाएँ",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "इस एजेंट द्वारा कार्यों को व्यवस्थित करते समय उपयोग किए जा सकने वाले ज्ञान पुनर्प्राप्ति स्रोत।",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "ऐसे knowledge bases कनेक्ट करें जिन्हें एजेंट काम करते समय खोजता है, ताकि उसके उत्तर आपके content पर आधारित हों। <docLink>और जानें</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "ऐसे knowledge bases कनेक्ट करें जिन्हें एजेंट काम करते समय खोजता है, ताकि उसके उत्तर आपके content पर आधारित हों। और जानें",
   "agentDetail.configure.knowledgeRetrieval.toggle": "ज्ञान पुनर्प्राप्ति टॉगल करें",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "मेटाडेटा फ़िल्टरिंग में एक मॉडल कॉन्फ़िगर करें",
   "agentDetail.configure.model.label": "मॉडल",
   "agentDetail.configure.orchestrate": "व्यवस्थित करें",
   "agentDetail.configure.preview.chatFeatures": "चैट सुविधाएँ",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "अंतिम-उपयोगकर्ता प्रमाणीकरण",
   "agentDetail.configure.preview.inputPlaceholder": "{{name}} को संदेश भेजें",
   "agentDetail.configure.preview.restart": "पूर्वावलोकन पुनः आरंभ करें",
+  "agentDetail.configure.preview.sandboxNotice": "Agent Linux सैंडबॉक्स में चलता है.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Dify Community Edition में, आपके हर Agent आपके Docker के अंदर Linux 7.0.0-10060-aws सैंडबॉक्स वातावरण में चलता है. Build Chats के ज़रिए वातावरण में किए गए आपके बदलाव स्थायी रहते हैं.",
   "agentDetail.configure.preview.title": "पूर्वावलोकन",
   "agentDetail.configure.prompt.copied": "प्रॉम्प्ट कॉपी हो गया",
   "agentDetail.configure.prompt.copy": "प्रॉम्प्ट कॉपी करें",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "और जानें",
   "agentDetail.configure.rightPanel.modeLabel": "एजेंट कॉन्फ़िगरेशन मोड",
   "agentDetail.configure.rightPanel.preview": "पूर्वावलोकन",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview केवल Dify Cloud और Enterprise पर उपलब्ध है।",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview तैयार agent को वैसे चलाता है जैसे आपके users उसे देखेंगे, साफ replies और chat features के साथ।",
   "agentDetail.configure.rightPanel.previewTipTitle": "अपने agent का preview करें",
   "agentDetail.configure.skills.add": "कौशल जोड़ें",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "कौशल",
   "agentDetail.configure.skills.label": "कौशल",
   "agentDetail.configure.skills.remove": "{{name}} हटाएँ",
-  "agentDetail.configure.skills.richTip": "अपने एजेंट को मॉड्यूलर क्षमताओं से लैस करें। <docLink>दस्तावेज़</docLink> देखें। प्रॉम्प्ट में / का उपयोग करके स्पष्ट संदर्भ दें।",
-  "agentDetail.configure.skills.tip": "अपने एजेंट को मॉड्यूलर क्षमताओं से लैस करें। दस्तावेज़ देखें। प्रॉम्प्ट में / का उपयोग करके स्पष्ट संदर्भ दें।",
+  "agentDetail.configure.skills.richTip": "किसी दोहराए जाने वाले कार्य के निर्देशों, फ़ाइलों और स्क्रिप्ट को एक Skill में पैकेज करें। Prompt में / से संदर्भ दें। <docLink>और जानें</docLink>\n\nBuild mode में, एजेंट इन्हें आपके लिए सेट कर सकता है।",
+  "agentDetail.configure.skills.tip": "किसी दोहराए जाने वाले कार्य के निर्देशों, फ़ाइलों और स्क्रिप्ट को एक Skill में पैकेज करें। Prompt में / से संदर्भ दें। और जानें\n\nBuild mode में, एजेंट इन्हें आपके लिए सेट कर सकता है।",
   "agentDetail.configure.skills.toggle": "कौशल टॉगल करें",
   "agentDetail.configure.skills.upload.action": "अपलोड करें",
   "agentDetail.configure.skills.upload.browse": "ब्राउज़ करें",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "{{name}} के लिए और क्रियाएँ",
   "agentDetail.configure.tools.removeAction": "{{name}} हटाएँ",
   "agentDetail.configure.tools.removeProvider": "सभी उपकरण हटाएँ",
-  "agentDetail.configure.tools.richTip": "अपने एजेंट को Dify के प्लगइन से लैस करें। <pluginDocLink>दस्तावेज़</pluginDocLink> देखें। प्रॉम्प्ट में / का उपयोग करके स्पष्ट संदर्भ दें।\nवैकल्पिक रूप से, आप Build chat के माध्यम से एजेंट को अन्य टूल (जैसे MCPs और CLI tools) इंस्टॉल और प्रमाणित करने के लिए हमेशा कह सकते हैं। ध्यान दें कि Build chats के माध्यम से कॉन्फ़िगर किए गए टूल यहां दिखाई नहीं देंगे, हालांकि वे आगे भी आपके एजेंट के लिए उपलब्ध रहेंगे। <buildDocLink>दस्तावेज़</buildDocLink> देखें।",
-  "agentDetail.configure.tools.tip": "अपने एजेंट को Dify के प्लगइन से लैस करें। दस्तावेज़ देखें। प्रॉम्प्ट में / का उपयोग करके स्पष्ट संदर्भ दें। वैकल्पिक रूप से, आप Build chat के माध्यम से एजेंट को अन्य टूल (जैसे MCPs और CLI tools) इंस्टॉल और प्रमाणित करने के लिए हमेशा कह सकते हैं। ध्यान दें कि Build chats के माध्यम से कॉन्फ़िगर किए गए टूल यहां दिखाई नहीं देंगे, हालांकि वे आगे भी आपके एजेंट के लिए उपलब्ध रहेंगे। दस्तावेज़ देखें।",
+  "agentDetail.configure.tools.richTip": "अपने workspace से Dify tools जोड़ें। Prompt में / से संदर्भ दें। <docLink>और जानें</docLink>\n\nएजेंट अपने sandbox में जो tools इंस्टॉल करता है वे यहां दिखाई नहीं देंगे, लेकिन उसके लिए उपलब्ध रहेंगे।",
+  "agentDetail.configure.tools.tip": "अपने workspace से Dify tools जोड़ें। Prompt में / से संदर्भ दें। और जानें\n\nएजेंट अपने sandbox में जो tools इंस्टॉल करता है वे यहां दिखाई नहीं देंगे, लेकिन उसके लिए उपलब्ध रहेंगे।",
   "agentDetail.configure.tools.toggle": "उपकरण टॉगल करें",
   "agentDetail.configure.tools.toolTabs.all": "सभी",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "एकीकरण",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "फ़ाइल सिस्टम पथ",
   "agentDetail.configure.workingDirectory.description": "एजेंट चलते समय जिन फाइलों को पढ़ता और लिखता है। वह जो भी बनाता है, जैसे qna_report.pdf, यहां दिखता है।",
+  "agentDetail.configure.workingDirectory.fileSystem": "फ़ाइल सिस्टम",
+  "agentDetail.configure.workingDirectory.home": "होम",
   "agentDetail.configure.workingDirectory.open": "वर्किंग डायरेक्टरी खोलें",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "वर्किंग डायरेक्टरी पर वापस जाएँ",
   "agentDetail.configure.workingDirectory.title": "वर्किंग डायरेक्टरी",
   "agentDetail.configure.workingDirectory.treeLabel": "वर्किंग डायरेक्टरी फाइलें",
+  "agentDetail.configure.workingDirectory.workingDirectory": "वर्किंग डायरेक्टरी",
   "agentDetail.documentTitle": "एजेंट",
   "agentDetail.logs.description": "पूर्ण लॉग एप्लिकेशन की चल रही स्थिति को रिकॉर्ड करते हैं, जिसमें उपयोगकर्ता इनपुट, एजेंट उत्तर, योजना और उपकरण उपयोग शामिल हैं।",
   "agentDetail.logs.empty": "कोई लॉग नहीं मिला",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "नाम आवश्यक है।",
   "roster.createForm.roleLabel": "भूमिका",
   "roster.createForm.rolePlaceholder": "उदा. शोध सहायक",
-  "roster.createForm.roleRequired": "भूमिका आवश्यक है।",
   "roster.createSuccess": "एजेंट बन गया।",
   "roster.dateTimeFormat": "DD/MM/YYYY hh:mm:ss A",
   "roster.deleteDialog.description": "इस एजेंट को सक्रिय Roster सूची से हटा दिया जाएगा।",
@@ -354,6 +372,7 @@
   "roster.empty": "अभी तक कोई एजेंट नहीं",
   "roster.emptySearch": "कोई मेल खाते एजेंट नहीं",
   "roster.filters.all": "सभी",
+  "roster.filters.createdByMe": "मेरे द्वारा बनाए गए",
   "roster.filters.drafts": "ड्राफ्ट",
   "roster.filters.label": "एजेंट फ़िल्टर",
   "roster.filters.published": "प्रकाशित",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "एजेंट Roster में सहेजा गया।",
   "roster.searchLabel": "एजेंट खोजें",
   "roster.searchPlaceholder": "नाम से एजेंट खोजें…",
+  "roster.sort.earliestCreated": "सबसे पहले बनाए गए",
+  "roster.sort.label": "क्रमबद्ध करें",
+  "roster.sort.lastModified": "अंतिम संशोधन",
+  "roster.sort.optionsLabel": "क्रमबद्ध विकल्प",
+  "roster.sort.recentlyCreated": "हाल ही में बनाए गए",
   "roster.tabs.agent": "एजेंट",
-  "roster.tabs.human": "मानव",
+  "roster.tabs.human": "संपर्क",
   "roster.tabsLabel": "Roster प्रकार",
   "roster.updateSuccess": "एजेंट अपडेट हो गया।",
   "roster.usageStatus.draft": "ड्राफ्ट"
diff --git a/web/i18n/hi-IN/common.json b/web/i18n/hi-IN/common.json
index 7cd3d6d23c8..5a82a70c482 100644
--- a/web/i18n/hi-IN/common.json
+++ b/web/i18n/hi-IN/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "इंस्टॉल करने के लिए Marketplace ब्राउज़ करें, या सेटिंग्स में प्रदाता कॉन्फ़िगर करें।",
   "modelProvider.selector.onlyCompatibleModelsShown": "केवल संगत मॉडल दिखाए गए हैं",
   "modelProvider.selector.showIncompatibleModels": "असंगत मॉडल दिखाएं",
+  "modelProvider.selector.suggestion": "सुझाव",
+  "modelProvider.selector.suggestionTip": "एजेंट कार्यों के लिए अनुशंसित",
   "modelProvider.showModels": "मॉडल्स दिखाएं",
   "modelProvider.speechToTextModel.key": "भाषण-से-पाठ मॉडल",
   "modelProvider.speechToTextModel.tip": "संवाद में भाषण-से-पाठ इनपुट के लिए डिफ़ॉल्ट मॉडल सेट करें।",
diff --git a/web/i18n/hi-IN/workflow.json b/web/i18n/hi-IN/workflow.json
index 1da71617556..48c584c2018 100644
--- a/web/i18n/hi-IN/workflow.json
+++ b/web/i18n/hi-IN/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "सबसे नया वर्कफ़्लो प्रकाशित करें और इसे टूल के रूप में कॉन्फ़िगर करने से पहले एक कनेक्टेड यूज़र इनपुट नोड सुनिश्चित करें।",
   "common.workflowAsToolTip": "कार्यप्रवाह अपडेट के बाद टूल पुनः कॉन्फ़िगरेशन आवश्यक है।",
   "common.workflowProcess": "कार्यप्रवाह प्रक्रिया",
+  "common.workflowProcessFailed": "कार्यप्रवाह प्रक्रिया विफल रही",
+  "common.workflowProcessPaused": "कार्यप्रवाह प्रक्रिया रुकी हुई है",
+  "common.workflowProcessRunning": "कार्यप्रवाह प्रक्रिया चल रही है",
+  "common.workflowProcessSucceeded": "कार्यप्रवाह प्रक्रिया सफल रही",
   "customWebhook": "कस्टम वेबहुक",
   "debug.copyLastRun": "अंतिम रन कॉपी करें",
   "debug.copyLastRunError": "अंतिम रन इनपुट को कॉपी करने में विफल",
diff --git a/web/i18n/id-ID/agent-v-2.json b/web/i18n/id-ID/agent-v-2.json
index 78761342714..983bea96f3e 100644
--- a/web/i18n/id-ID/agent-v-2.json
+++ b/web/i18n/id-ID/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Tampilkan nilai {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Agen Anda berjalan di lingkungan Linux. Anda dapat menyiapkan konfigurasi atau rahasia dengan mendefinisikan variabel lingkungan di sini.\n\nSebagai alternatif, Anda selalu dapat menyiapkannya melalui chat Build. Lihat <docLink>dokumentasi</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Tetapkan nilai yang dibaca agen saat runtime, seperti API key yang dibutuhkan tool, tanpa memasukkannya ke Prompt.\n\nDalam mode Build, agen dapat menyiapkannya untuk Anda.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Pilih cakupan variabel lingkungan",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Agen Anda berjalan di lingkungan Linux. Anda dapat menyiapkan konfigurasi atau rahasia dengan mendefinisikan variabel lingkungan di sini. Sebagai alternatif, Anda selalu dapat menyiapkannya melalui chat Build. Lihat dokumentasi.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Tetapkan nilai yang dibaca agen saat runtime, seperti API key yang dibutuhkan tool, tanpa memasukkannya ke Prompt.\n\nDalam mode Build, agen dapat menyiapkannya untuk Anda.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Alihkan Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Bangun agen Anda lewat chat",
   "agentDetail.configure.build.inputPlaceholder": "Jelaskan apa yang harus dilakukan agen Anda",
   "agentDetail.configure.build.startBuild": "Mulai build",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Catatan berjalan untuk build ini yang digunakan agen untuk mempertahankan konteks antar sesi. Catatan ini ditambahkan ke file Anda saat diterapkan.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} perubahan untuk diterapkan",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} perubahan untuk diterapkan",
   "agentDetail.configure.buildDraft.discard": "Buang",
   "agentDetail.configure.buildDraft.modeBadge": "Mode build",
   "agentDetail.configure.buildDraft.modeDescription": "Anda berada dalam mode build. Bentuk pengaturan ini lewat chat di kanan, lalu terapkan.",
+  "agentDetail.configure.buildDraft.rewritten": "Ditulis ulang",
   "agentDetail.configure.buildDraft.title": "Draf build",
+  "agentDetail.configure.buildDraft.updated": "Diperbarui",
   "agentDetail.configure.chatFeatures.description": "Bentuk pengalaman chat pengguna akhir di Web app dan permukaan chat Anda.",
   "agentDetail.configure.chatFeatures.title": "Fitur Chat",
   "agentDetail.configure.files.add": "Tambahkan file",
+  "agentDetail.configure.files.buildNote.generated": "Dihasilkan",
+  "agentDetail.configure.files.buildNote.richTooltip": "Catatan agen tentang apa yang disiapkannya dalam mode Build. Agen membaca ini di awal setiap percakapan, bersama Prompt Anda. <docLink>Pelajari selengkapnya</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Catatan agen tentang apa yang disiapkannya dalam mode Build. Agen membaca ini di awal setiap percakapan, bersama Prompt Anda. Pelajari selengkapnya",
   "agentDetail.configure.files.empty.description": "Unggah dokumen yang dapat dibaca agen, seperti spesifikasi, templat, atau pedoman",
   "agentDetail.configure.files.empty.title": "Belum ada file",
   "agentDetail.configure.files.label": "File",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Gagal memuat pratinjau.",
   "agentDetail.configure.files.preview.unsupported": "File ini tidak mendukung pratinjau.",
   "agentDetail.configure.files.remove": "Hapus {{name}}",
-  "agentDetail.configure.files.tip": "Berikan agen Anda contoh, panduan, atau sekadar pembawa informasi yang terlalu panjang untuk prompt. Buat referensi eksplisit dengan / di Prompt.",
+  "agentDetail.configure.files.tip": "Untuk materi referensi yang terlalu panjang untuk ditempel ke Prompt: spesifikasi, templat, panduan. Referensikan dengan / di Prompt.\n\nDalam mode Build, agen dapat menyiapkannya untuk Anda.",
   "agentDetail.configure.files.toggle": "Alihkan file",
   "agentDetail.configure.files.treeLabel": "File agen",
   "agentDetail.configure.files.upload.action": "Unggah",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Hapus {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Sumber pengambilan pengetahuan yang dapat digunakan agen ini saat mengorkestrasi tugas.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Hubungkan knowledge base yang dicari agen saat bekerja, sehingga jawabannya mengambil dari konten Anda. <docLink>Pelajari selengkapnya</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Hubungkan knowledge base yang dicari agen saat bekerja, sehingga jawabannya mengambil dari konten Anda. Pelajari selengkapnya",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Alihkan pengambilan pengetahuan",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Konfigurasikan model dalam pemfilteran metadata",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Orkestrasi",
   "agentDetail.configure.preview.chatFeatures": "Fitur Chat",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Autentikasi pengguna akhir",
   "agentDetail.configure.preview.inputPlaceholder": "Kirim pesan ke {{name}}",
   "agentDetail.configure.preview.restart": "Mulai ulang pratinjau",
+  "agentDetail.configure.preview.sandboxNotice": "Agent berjalan di sandbox Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Di Dify Community Edition, setiap Agent Anda berjalan di lingkungan sandbox Linux 7.0.0-10060-aws di dalam Docker Anda. Perubahan yang Anda buat pada lingkungan melalui Build Chats bersifat persisten.",
   "agentDetail.configure.preview.title": "Pratinjau",
   "agentDetail.configure.prompt.copied": "Prompt disalin",
   "agentDetail.configure.prompt.copy": "Salin prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Pelajari selengkapnya",
   "agentDetail.configure.rightPanel.modeLabel": "Mode konfigurasi agen",
   "agentDetail.configure.rightPanel.preview": "Pratinjau",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview hanya tersedia di Dify Cloud dan Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview menjalankan agent yang sudah selesai seperti yang akan dilihat pengguna, dengan balasan rapi dan fitur chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Pratinjau agen Anda",
   "agentDetail.configure.skills.add": "Tambahkan keterampilan",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Keterampilan",
   "agentDetail.configure.skills.label": "Keterampilan",
   "agentDetail.configure.skills.remove": "Hapus {{name}}",
-  "agentDetail.configure.skills.richTip": "Lengkapi agen Anda dengan kemampuan modular. Lihat <docLink>dokumentasi</docLink>. Buat referensi eksplisit dengan / di Prompt.",
-  "agentDetail.configure.skills.tip": "Lengkapi agen Anda dengan kemampuan modular. Lihat dokumentasi. Buat referensi eksplisit dengan / di Prompt.",
+  "agentDetail.configure.skills.richTip": "Kemas instruksi, file, dan skrip untuk tugas berulang menjadi skill. Referensikan dengan / di Prompt. <docLink>Pelajari selengkapnya</docLink>\n\nDalam mode Build, agen dapat menyiapkannya untuk Anda.",
+  "agentDetail.configure.skills.tip": "Kemas instruksi, file, dan skrip untuk tugas berulang menjadi skill. Referensikan dengan / di Prompt. Pelajari selengkapnya\n\nDalam mode Build, agen dapat menyiapkannya untuk Anda.",
   "agentDetail.configure.skills.toggle": "Alihkan keterampilan",
   "agentDetail.configure.skills.upload.action": "Unggah",
   "agentDetail.configure.skills.upload.browse": "telusuri",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Tindakan lainnya untuk {{name}}",
   "agentDetail.configure.tools.removeAction": "Hapus {{name}}",
   "agentDetail.configure.tools.removeProvider": "Hapus semua alat",
-  "agentDetail.configure.tools.richTip": "Lengkapi agen Anda dengan plugin Dify. Lihat <pluginDocLink>dokumentasi</pluginDocLink>. Buat referensi eksplisit dengan / di Prompt.\nAtau, Anda selalu dapat meminta agen untuk menginstal dan mengautentikasi alat lain (misalnya MCPs dan CLI tools) melalui Build chat. Perhatikan bahwa alat yang dikonfigurasi melalui Build chats TIDAK akan muncul di sini, meskipun tetap tersedia untuk agen Anda ke depannya. Lihat <buildDocLink>dokumentasi</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Lengkapi agen Anda dengan plugin Dify. Lihat dokumentasi. Buat referensi eksplisit dengan / di Prompt. Atau, Anda selalu dapat meminta agen untuk menginstal dan mengautentikasi alat lain (misalnya MCPs dan CLI tools) melalui Build chat. Perhatikan bahwa alat yang dikonfigurasi melalui Build chats TIDAK akan muncul di sini, meskipun tetap tersedia untuk agen Anda ke depannya. Lihat dokumentasi.",
+  "agentDetail.configure.tools.richTip": "Tambahkan Dify tools dari workspace Anda. Referensikan dengan / di Prompt. <docLink>Pelajari selengkapnya</docLink>\n\nTools yang diinstal agen di sandbox-nya tidak akan muncul di sini, tetapi tetap tersedia baginya.",
+  "agentDetail.configure.tools.tip": "Tambahkan Dify tools dari workspace Anda. Referensikan dengan / di Prompt. Pelajari selengkapnya\n\nTools yang diinstal agen di sandbox-nya tidak akan muncul di sini, tetapi tetap tersedia baginya.",
   "agentDetail.configure.tools.toggle": "Alihkan alat",
   "agentDetail.configure.tools.toolTabs.all": "Semua",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugin",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Jalur sistem file",
   "agentDetail.configure.workingDirectory.description": "File yang dibaca dan ditulis agent saat berjalan. Semua yang dibuatnya, seperti qna_report.pdf, muncul di sini.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Sistem file",
+  "agentDetail.configure.workingDirectory.home": "beranda",
   "agentDetail.configure.workingDirectory.open": "Buka direktori kerja",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Kembali ke direktori kerja",
   "agentDetail.configure.workingDirectory.title": "Direktori kerja",
   "agentDetail.configure.workingDirectory.treeLabel": "File direktori kerja",
+  "agentDetail.configure.workingDirectory.workingDirectory": "direktori kerja",
   "agentDetail.documentTitle": "Agen",
   "agentDetail.logs.description": "Log lengkap mencatat status berjalan aplikasi, termasuk input pengguna, balasan agen, perencanaan, dan penggunaan alat.",
   "agentDetail.logs.empty": "Tidak ada log ditemukan",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Nama diperlukan.",
   "roster.createForm.roleLabel": "Peran",
   "roster.createForm.rolePlaceholder": "mis. Asisten Riset",
-  "roster.createForm.roleRequired": "Peran diperlukan.",
   "roster.createSuccess": "Agen dibuat.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Agen ini akan dihapus dari daftar Roster aktif.",
@@ -354,6 +372,7 @@
   "roster.empty": "Belum ada agen",
   "roster.emptySearch": "Tidak ada agen yang cocok",
   "roster.filters.all": "Semua",
+  "roster.filters.createdByMe": "Dibuat oleh saya",
   "roster.filters.drafts": "Draf",
   "roster.filters.label": "Filter agen",
   "roster.filters.published": "Diterbitkan",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agen disimpan ke Roster.",
   "roster.searchLabel": "Cari agen",
   "roster.searchPlaceholder": "Cari agen berdasarkan nama…",
+  "roster.sort.earliestCreated": "Paling awal dibuat",
+  "roster.sort.label": "Urutkan berdasarkan",
+  "roster.sort.lastModified": "Terakhir diubah",
+  "roster.sort.optionsLabel": "Opsi pengurutan",
+  "roster.sort.recentlyCreated": "Baru dibuat",
   "roster.tabs.agent": "Agen",
-  "roster.tabs.human": "Manusia",
+  "roster.tabs.human": "Kontak",
   "roster.tabsLabel": "Tipe Roster",
   "roster.updateSuccess": "Agen diperbarui.",
   "roster.usageStatus.draft": "Draf"
diff --git a/web/i18n/id-ID/common.json b/web/i18n/id-ID/common.json
index 6f783a52d77..d37302bae9f 100644
--- a/web/i18n/id-ID/common.json
+++ b/web/i18n/id-ID/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Jelajahi Marketplace untuk menginstal, atau konfigurasikan penyedia di pengaturan.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Hanya model yang kompatibel yang ditampilkan",
   "modelProvider.selector.showIncompatibleModels": "Tampilkan model yang tidak kompatibel",
+  "modelProvider.selector.suggestion": "Saran",
+  "modelProvider.selector.suggestionTip": "Direkomendasikan untuk tugas agen",
   "modelProvider.showModels": "Tampilkan Model",
   "modelProvider.speechToTextModel.key": "Model Ucapan-ke-Teks",
   "modelProvider.speechToTextModel.tip": "Atur model default untuk input ucapan-ke-teks dalam percakapan.",
diff --git a/web/i18n/id-ID/workflow.json b/web/i18n/id-ID/workflow.json
index d42c24b9c0b..fd38608711e 100644
--- a/web/i18n/id-ID/workflow.json
+++ b/web/i18n/id-ID/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Terbitkan alur kerja terbaru dan pastikan ada node Input Pengguna yang terhubung sebelum mengonfigurasinya sebagai alat.",
   "common.workflowAsToolTip": "Konfigurasi ulang alat diperlukan setelah pembaruan alur kerja.",
   "common.workflowProcess": "Proses Alur Kerja",
+  "common.workflowProcessFailed": "Proses Alur Kerja gagal",
+  "common.workflowProcessPaused": "Proses Alur Kerja dijeda",
+  "common.workflowProcessRunning": "Proses Alur Kerja sedang berjalan",
+  "common.workflowProcessSucceeded": "Proses Alur Kerja berhasil",
   "customWebhook": "Webhook Kustom",
   "debug.copyLastRun": "Salin Eksekusi Terakhir",
   "debug.copyLastRunError": "Gagal menyalin input eksekusi terakhir",
diff --git a/web/i18n/it-IT/agent-v-2.json b/web/i18n/it-IT/agent-v-2.json
index 69725d1d7b2..4f0bb90010a 100644
--- a/web/i18n/it-IT/agent-v-2.json
+++ b/web/i18n/it-IT/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Mostra valore di {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Il tuo agente viene eseguito in un ambiente Linux. Puoi preimpostare configurazioni o segreti definendo qui le variabili d’ambiente.\n\nIn alternativa, puoi sempre configurare tutto tramite una chat Build. Consulta la <docLink>documentazione</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Imposta valori che l'agente legge a runtime, come una chiave API necessaria a uno strumento, senza inserirli nel Prompt.\n\nIn modalità Build, l'agente può configurarli per te.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Seleziona ambito della variabile d’ambiente",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Il tuo agente viene eseguito in un ambiente Linux. Puoi preimpostare configurazioni o segreti definendo qui le variabili d’ambiente. In alternativa, puoi sempre configurare tutto tramite una chat Build. Consulta la documentazione.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Imposta valori che l'agente legge a runtime, come una chiave API necessaria a uno strumento, senza inserirli nel Prompt.\n\nIn modalità Build, l'agente può configurarli per te.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Attiva/disattiva Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Crea il tuo agente con la chat",
   "agentDetail.configure.build.inputPlaceholder": "Descrivi cosa dovrebbe fare il tuo agente",
   "agentDetail.configure.build.startBuild": "Avvia build",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Una nota continua di questa build che l'agente usa per mantenere il contesto tra le sessioni. Viene aggiunta ai tuoi file quando applichi.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} modifica da applicare",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} modifiche da applicare",
   "agentDetail.configure.buildDraft.discard": "Scarta",
   "agentDetail.configure.buildDraft.modeBadge": "Modalità build",
   "agentDetail.configure.buildDraft.modeDescription": "Sei in modalità build. Modella questa configurazione tramite la chat a destra, poi applica.",
+  "agentDetail.configure.buildDraft.rewritten": "Riscritto",
   "agentDetail.configure.buildDraft.title": "Bozza di build",
+  "agentDetail.configure.buildDraft.updated": "Aggiornato",
   "agentDetail.configure.chatFeatures.description": "Definisci l’esperienza di chat per l’utente finale sulla tua webapp e sulle superfici di chat.",
   "agentDetail.configure.chatFeatures.title": "Funzionalità chat",
   "agentDetail.configure.files.add": "Aggiungi file",
+  "agentDetail.configure.files.buildNote.generated": "Generato",
+  "agentDetail.configure.files.buildNote.richTooltip": "Il registro dell'agente di ciò che ha configurato in modalità Build. Lo legge all'inizio di ogni conversazione, insieme al tuo Prompt. <docLink>Scopri di più</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Il registro dell'agente di ciò che ha configurato in modalità Build. Lo legge all'inizio di ogni conversazione, insieme al tuo Prompt. Scopri di più",
   "agentDetail.configure.files.empty.description": "Carica documenti che l’agente possa leggere, come specifiche, modelli o linee guida",
   "agentDetail.configure.files.empty.title": "Nessun file al momento",
   "agentDetail.configure.files.label": "File",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Impossibile caricare l’anteprima.",
   "agentDetail.configure.files.preview.unsupported": "Questo file non supporta l’anteprima.",
   "agentDetail.configure.files.remove": "Rimuovi {{name}}",
-  "agentDetail.configure.files.tip": "Fornisci al tuo agente esempi, linee guida o semplicemente informazioni troppo lunghe per il prompt. Fai riferimento esplicito con / nel prompt.",
+  "agentDetail.configure.files.tip": "Per materiale di riferimento troppo lungo da incollare nel Prompt: specifiche, modelli, linee guida. Fai riferimento con / nel Prompt.\n\nIn modalità Build, l'agente può configurarli per te.",
   "agentDetail.configure.files.toggle": "Attiva/disattiva file",
   "agentDetail.configure.files.treeLabel": "File dell’agente",
   "agentDetail.configure.files.upload.action": "Carica",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Rimuovi {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Sorgenti di recupero della conoscenza che questo agente può utilizzare durante l’orchestrazione delle attività.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Collega basi di conoscenza che l'agente consulta mentre lavora, così le sue risposte si basano sui tuoi contenuti. <docLink>Scopri di più</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Collega basi di conoscenza che l'agente consulta mentre lavora, così le sue risposte si basano sui tuoi contenuti. Scopri di più",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Attiva/disattiva recupero della conoscenza",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Configura un modello nel filtro dei metadati",
   "agentDetail.configure.model.label": "Modello",
   "agentDetail.configure.orchestrate": "Orchestra",
   "agentDetail.configure.preview.chatFeatures": "Funzionalità chat",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Autenticazione utente finale",
   "agentDetail.configure.preview.inputPlaceholder": "Invia un messaggio a {{name}}",
   "agentDetail.configure.preview.restart": "Riavvia anteprima",
+  "agentDetail.configure.preview.sandboxNotice": "L'agent viene eseguito in una sandbox Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "In Dify Community Edition, ciascuno dei tuoi agent viene eseguito in un ambiente sandbox Linux 7.0.0-10060-aws all’interno del tuo Docker. Le modifiche all’ambiente tramite Build Chats sono persistenti.",
   "agentDetail.configure.preview.title": "Anteprima",
   "agentDetail.configure.prompt.copied": "Prompt copiato",
   "agentDetail.configure.prompt.copy": "Copia prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Scopri di più",
   "agentDetail.configure.rightPanel.modeLabel": "Modalità di configurazione agente",
   "agentDetail.configure.rightPanel.preview": "Anteprima",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview è disponibile solo su Dify Cloud ed Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview esegue l’agente completato come lo vedranno gli utenti, con risposte pulite e funzionalità di chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Anteprima del tuo agente",
   "agentDetail.configure.skills.add": "Aggiungi abilità",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Abilità",
   "agentDetail.configure.skills.label": "Abilità",
   "agentDetail.configure.skills.remove": "Rimuovi {{name}}",
-  "agentDetail.configure.skills.richTip": "Dota il tuo agente di capacità modulari. Consulta la <docLink>documentazione</docLink>. Fai riferimento esplicito con / nel prompt.",
-  "agentDetail.configure.skills.tip": "Dota il tuo agente di capacità modulari. Consulta la documentazione. Fai riferimento esplicito con / nel prompt.",
+  "agentDetail.configure.skills.richTip": "Raggruppa istruzioni, file e script per un'attività ricorrente in una skill. Fai riferimento con / nel Prompt. <docLink>Scopri di più</docLink>\n\nIn modalità Build, l'agente può configurarli per te.",
+  "agentDetail.configure.skills.tip": "Raggruppa istruzioni, file e script per un'attività ricorrente in una skill. Fai riferimento con / nel Prompt. Scopri di più\n\nIn modalità Build, l'agente può configurarli per te.",
   "agentDetail.configure.skills.toggle": "Attiva/disattiva abilità",
   "agentDetail.configure.skills.upload.action": "Carica",
   "agentDetail.configure.skills.upload.browse": "sfoglia",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Altre azioni per {{name}}",
   "agentDetail.configure.tools.removeAction": "Rimuovi {{name}}",
   "agentDetail.configure.tools.removeProvider": "Rimuovi tutti gli strumenti",
-  "agentDetail.configure.tools.richTip": "Dota il tuo agente dei plugin di Dify. Consulta la <pluginDocLink>documentazione</pluginDocLink>. Fai riferimento esplicito con / nel prompt.\nIn alternativa, puoi sempre dire all’agente di installare e autenticare altri strumenti (ad esempio MCPs e strumenti CLI) tramite una chat Build. Nota che gli strumenti configurati tramite chat Build NON appariranno qui, anche se resteranno disponibili per il tuo agente in futuro. Consulta la <buildDocLink>documentazione</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Dota il tuo agente dei plugin di Dify. Consulta la documentazione. Fai riferimento esplicito con / nel prompt. In alternativa, puoi sempre dire all’agente di installare e autenticare altri strumenti (ad esempio MCPs e strumenti CLI) tramite una chat Build. Nota che gli strumenti configurati tramite chat Build NON appariranno qui, anche se resteranno disponibili per il tuo agente in futuro. Consulta la documentazione.",
+  "agentDetail.configure.tools.richTip": "Aggiungi strumenti Dify dal tuo workspace. Fai riferimento con / nel Prompt. <docLink>Scopri di più</docLink>\n\nGli strumenti che l'agente installa nel suo sandbox non appariranno qui, ma resteranno disponibili.",
+  "agentDetail.configure.tools.tip": "Aggiungi strumenti Dify dal tuo workspace. Fai riferimento con / nel Prompt. Scopri di più\n\nGli strumenti che l'agente installa nel suo sandbox non appariranno qui, ma resteranno disponibili.",
   "agentDetail.configure.tools.toggle": "Attiva/disattiva strumenti",
   "agentDetail.configure.tools.toolTabs.all": "Tutti",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugin",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Percorso del file system",
   "agentDetail.configure.workingDirectory.description": "File che l’agente legge e scrive durante l’esecuzione. Tutto ciò che genera, come qna_report.pdf, appare qui.",
+  "agentDetail.configure.workingDirectory.fileSystem": "File system",
+  "agentDetail.configure.workingDirectory.home": "home",
   "agentDetail.configure.workingDirectory.open": "Apri directory di lavoro",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Torna alla directory di lavoro",
   "agentDetail.configure.workingDirectory.title": "Directory di lavoro",
   "agentDetail.configure.workingDirectory.treeLabel": "File della directory di lavoro",
+  "agentDetail.configure.workingDirectory.workingDirectory": "directory di lavoro",
   "agentDetail.documentTitle": "Agente",
   "agentDetail.logs.description": "I log completi registrano lo stato di esecuzione dell’applicazione, inclusi input degli utenti, risposte dell’agente, pianificazione e uso degli strumenti.",
   "agentDetail.logs.empty": "Nessun log trovato",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Il nome è obbligatorio.",
   "roster.createForm.roleLabel": "Ruolo",
   "roster.createForm.rolePlaceholder": "ad es. Assistente di ricerca",
-  "roster.createForm.roleRequired": "Il ruolo è obbligatorio.",
   "roster.createSuccess": "Agente creato.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Questo agente verrà rimosso dall’elenco del roster attivo.",
@@ -354,6 +372,7 @@
   "roster.empty": "Nessun agente al momento",
   "roster.emptySearch": "Nessun agente corrispondente",
   "roster.filters.all": "Tutti",
+  "roster.filters.createdByMe": "Creati da me",
   "roster.filters.drafts": "Bozze",
   "roster.filters.label": "Filtri agente",
   "roster.filters.published": "Pubblicati",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agente salvato nel roster.",
   "roster.searchLabel": "Cerca agenti",
   "roster.searchPlaceholder": "Cerca agenti per nome…",
+  "roster.sort.earliestCreated": "Creati per primi",
+  "roster.sort.label": "Ordina per",
+  "roster.sort.lastModified": "Ultima modifica",
+  "roster.sort.optionsLabel": "Opzioni di ordinamento",
+  "roster.sort.recentlyCreated": "Creati di recente",
   "roster.tabs.agent": "Agente",
-  "roster.tabs.human": "Umano",
+  "roster.tabs.human": "Contatti",
   "roster.tabsLabel": "Tipo di roster",
   "roster.updateSuccess": "Agente aggiornato.",
   "roster.usageStatus.draft": "Bozza"
diff --git a/web/i18n/it-IT/common.json b/web/i18n/it-IT/common.json
index 35cb57442da..b65ce4d48d2 100644
--- a/web/i18n/it-IT/common.json
+++ b/web/i18n/it-IT/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Cerca nel Marketplace per installarne uno o configura i fornitori nelle impostazioni.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Vengono mostrati solo i modelli compatibili",
   "modelProvider.selector.showIncompatibleModels": "Mostra modelli incompatibili",
+  "modelProvider.selector.suggestion": "Suggerimento",
+  "modelProvider.selector.suggestionTip": "Consigliato per le attività degli agenti",
   "modelProvider.showModels": "Mostra Modelli",
   "modelProvider.speechToTextModel.key": "Modello da Voce a Testo",
   "modelProvider.speechToTextModel.tip": "Imposta il modello predefinito per l'input da voce a testo nella conversazione.",
diff --git a/web/i18n/it-IT/workflow.json b/web/i18n/it-IT/workflow.json
index 67cd1ca8aff..cd0c18c066f 100644
--- a/web/i18n/it-IT/workflow.json
+++ b/web/i18n/it-IT/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Pubblica il flusso di lavoro più recente e assicurati che ci sia un nodo di Input Utente collegato prima di configurarlo come strumento.",
   "common.workflowAsToolTip": "È richiesta una nuova configurazione dello strumento dopo l'aggiornamento del flusso di lavoro.",
   "common.workflowProcess": "Processo di flusso di lavoro",
+  "common.workflowProcessFailed": "Processo di flusso di lavoro non riuscito",
+  "common.workflowProcessPaused": "Processo di flusso di lavoro in pausa",
+  "common.workflowProcessRunning": "Processo di flusso di lavoro in esecuzione",
+  "common.workflowProcessSucceeded": "Processo di flusso di lavoro riuscito",
   "customWebhook": "Webhook personalizzato",
   "debug.copyLastRun": "Copia ultimo eseguito",
   "debug.copyLastRunError": "Impossibile copiare gli input dell'ultima esecuzione",
diff --git a/web/i18n/ja-JP/agent-v-2.json b/web/i18n/ja-JP/agent-v-2.json
index 7fa72d4bb12..a45519caf11 100644
--- a/web/i18n/ja-JP/agent-v-2.json
+++ b/web/i18n/ja-JP/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "{{key}} の値を表示",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "エージェントは Linux 環境で実行されます。ここで環境変数を定義して、設定やシークレットを事前に指定できます。\n\nまた、Build チャット経由で設定することもできます。<docLink>ドキュメント</docLink>を参照してください。",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "ツールに必要な API キーなど、エージェントが実行時に読み取る値を Prompt に入れずに設定します。\n\nBuild モードでは、エージェントがこれらを設定できます。",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "環境変数のスコープを選択",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "エージェントは Linux 環境で実行されます。ここで環境変数を定義して、設定やシークレットを事前に指定できます。また、Build チャット経由で設定することもできます。ドキュメントを参照してください。",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "ツールに必要な API キーなど、エージェントが実行時に読み取る値を Prompt に入れずに設定します。\n\nBuild モードでは、エージェントがこれらを設定できます。",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Env Editor の表示を切り替え",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "チャットでエージェントを作成",
   "agentDetail.configure.build.inputPlaceholder": "エージェントに実行させたいことを説明",
   "agentDetail.configure.build.startBuild": "ビルドを開始",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "このビルドの継続的なメモで、エージェントがセッション間でコンテキストを保持するために使用します。適用するとファイルに追加されます。",
+  "agentDetail.configure.buildDraft.changesToApply_one": "適用する変更 {{count}} 件",
+  "agentDetail.configure.buildDraft.changesToApply_other": "適用する変更 {{count}} 件",
   "agentDetail.configure.buildDraft.discard": "破棄",
   "agentDetail.configure.buildDraft.modeBadge": "ビルドモード",
   "agentDetail.configure.buildDraft.modeDescription": "ビルドモードです。右側のチャットでこの設定を調整してから適用してください。",
+  "agentDetail.configure.buildDraft.rewritten": "書き換え済み",
   "agentDetail.configure.buildDraft.title": "ビルドドラフト",
+  "agentDetail.configure.buildDraft.updated": "更新済み",
   "agentDetail.configure.chatFeatures.description": "Web app やチャット画面でのエンドユーザー向けチャット体験を設定します。",
   "agentDetail.configure.chatFeatures.title": "チャット機能",
   "agentDetail.configure.files.add": "ファイルを追加",
+  "agentDetail.configure.files.buildNote.generated": "生成済み",
+  "agentDetail.configure.files.buildNote.richTooltip": "Build モードでエージェントが設定した内容の記録です。各会話の開始時に、Prompt と一緒にこれを読み取ります。<docLink>詳しく見る</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Build モードでエージェントが設定した内容の記録です。各会話の開始時に、Prompt と一緒にこれを読み取ります。詳しく見る",
   "agentDetail.configure.files.empty.description": "仕様、テンプレート、ガイドラインなど、エージェントが読めるドキュメントをアップロード",
   "agentDetail.configure.files.empty.title": "ファイルはまだありません",
   "agentDetail.configure.files.label": "ファイル",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "プレビューの読み込みに失敗しました。",
   "agentDetail.configure.files.preview.unsupported": "ファイルはプレビューに対応していません。",
   "agentDetail.configure.files.remove": "{{name}} を削除",
-  "agentDetail.configure.files.tip": "プロンプトには長すぎる例、ガイドライン、または情報そのものをエージェントに提供します。プロンプト内で / を使って明示的に参照してください。",
+  "agentDetail.configure.files.tip": "Prompt に貼り付けるには長すぎる参考資料向けです: 仕様、テンプレート、ガイドライン。Prompt で / を使って参照します。\n\nBuild モードでは、エージェントがこれらを設定できます。",
   "agentDetail.configure.files.toggle": "ファイルの表示を切り替え",
   "agentDetail.configure.files.treeLabel": "エージェントファイル",
   "agentDetail.configure.files.upload.action": "アップロード",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "{{name}} を削除",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "このエージェントがタスクのオーケストレーション中に利用できるナレッジ検索ソース。",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "エージェントが作業中に検索するナレッジベースを接続し、回答があなたのコンテンツに基づくようにします。<docLink>詳しく見る</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "エージェントが作業中に検索するナレッジベースを接続し、回答があなたのコンテンツに基づくようにします。詳しく見る",
   "agentDetail.configure.knowledgeRetrieval.toggle": "ナレッジ検索の表示を切り替え",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "メタデータフィルタリングでモデルを設定してください",
   "agentDetail.configure.model.label": "モデル",
   "agentDetail.configure.orchestrate": "オーケストレーション",
   "agentDetail.configure.preview.chatFeatures": "チャット機能",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "エンドユーザー認証",
   "agentDetail.configure.preview.inputPlaceholder": "{{name}} にメッセージを送信",
   "agentDetail.configure.preview.restart": "プレビューを再開",
+  "agentDetail.configure.preview.sandboxNotice": "Agent は Linux サンドボックスで実行されます。",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Dify Community Edition では、各 Agent は Docker 内の Linux 7.0.0-10060-aws サンドボックス環境で実行されます。Build Chats で行った環境への編集は永続化されます。",
   "agentDetail.configure.preview.title": "プレビュー",
   "agentDetail.configure.prompt.copied": "プロンプトをコピーしました",
   "agentDetail.configure.prompt.copy": "プロンプトをコピー",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "詳しく見る",
   "agentDetail.configure.rightPanel.modeLabel": "エージェント設定モード",
   "agentDetail.configure.rightPanel.preview": "プレビュー",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview は Dify Cloud と Enterprise でのみ利用できます。",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview は完成したエージェントをユーザーに見える形で実行し、整った返信とチャット機能を確認できます。",
   "agentDetail.configure.rightPanel.previewTipTitle": "エージェントをプレビュー",
   "agentDetail.configure.skills.add": "スキルを追加",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "スキル",
   "agentDetail.configure.skills.label": "スキル",
   "agentDetail.configure.skills.remove": "{{name}} を削除",
-  "agentDetail.configure.skills.richTip": "エージェントにモジュール式の機能を装備します。<docLink>ドキュメント</docLink>を参照してください。プロンプト内で / を使って明示的に参照してください。",
-  "agentDetail.configure.skills.tip": "エージェントにモジュール式の機能を装備します。ドキュメントを参照してください。プロンプト内で / を使って明示的に参照してください。",
+  "agentDetail.configure.skills.richTip": "繰り返し行うタスクに必要な指示、ファイル、スクリプトを Skill にまとめます。Prompt で / を使って参照します。<docLink>詳しく見る</docLink>\n\nBuild モードでは、エージェントがこれらを設定できます。",
+  "agentDetail.configure.skills.tip": "繰り返し行うタスクに必要な指示、ファイル、スクリプトを Skill にまとめます。Prompt で / を使って参照します。詳しく見る\n\nBuild モードでは、エージェントがこれらを設定できます。",
   "agentDetail.configure.skills.toggle": "スキルの表示を切り替え",
   "agentDetail.configure.skills.upload.action": "アップロード",
   "agentDetail.configure.skills.upload.browse": "選択",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "{{name}} のその他の操作",
   "agentDetail.configure.tools.removeAction": "{{name}} を削除",
   "agentDetail.configure.tools.removeProvider": "すべてのツールを削除",
-  "agentDetail.configure.tools.richTip": "エージェントに Dify のプラグインを装備します。<pluginDocLink>ドキュメント</pluginDocLink>を参照してください。プロンプト内で / を使って明示的に参照してください。\nまた、Build チャットを通じて、他のツール（MCPs や CLI ツールなど）のインストールと認証をエージェントに指示することもできます。Build チャットで設定したツールはここには表示されませんが、今後もエージェントで利用できます。<buildDocLink>ドキュメント</buildDocLink>を参照してください。",
-  "agentDetail.configure.tools.tip": "エージェントに Dify のプラグインを装備します。ドキュメントを参照してください。プロンプト内で / を使って明示的に参照してください。また、Build チャットを通じて、他のツール（MCPs や CLI ツールなど）のインストールと認証をエージェントに指示することもできます。Build チャットで設定したツールはここには表示されませんが、今後もエージェントで利用できます。ドキュメントを参照してください。",
+  "agentDetail.configure.tools.richTip": "ワークスペースの Dify ツールを追加します。Prompt で / を使って参照します。<docLink>詳しく見る</docLink>\n\nエージェントがサンドボックス内にインストールしたツールはここには表示されませんが、引き続き利用できます。",
+  "agentDetail.configure.tools.tip": "ワークスペースの Dify ツールを追加します。Prompt で / を使って参照します。詳しく見る\n\nエージェントがサンドボックス内にインストールしたツールはここには表示されませんが、引き続き利用できます。",
   "agentDetail.configure.tools.toggle": "ツールの表示を切り替え",
   "agentDetail.configure.tools.toolTabs.all": "すべて",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "プラグイン",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "ファイルシステムのパス",
   "agentDetail.configure.workingDirectory.description": "エージェントが実行中に読み書きするファイルです。qna_report.pdf など、生成されたものはここに表示されます。",
+  "agentDetail.configure.workingDirectory.fileSystem": "ファイルシステム",
+  "agentDetail.configure.workingDirectory.home": "ホーム",
   "agentDetail.configure.workingDirectory.open": "作業ディレクトリを開く",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "作業ディレクトリに戻る",
   "agentDetail.configure.workingDirectory.title": "作業ディレクトリ",
   "agentDetail.configure.workingDirectory.treeLabel": "作業ディレクトリのファイル",
+  "agentDetail.configure.workingDirectory.workingDirectory": "作業ディレクトリ",
   "agentDetail.documentTitle": "エージェント",
   "agentDetail.logs.description": "完全なログには、ユーザー入力、エージェントの応答、計画、ツールの使用など、アプリケーションの実行状況が記録されます。",
   "agentDetail.logs.empty": "ログが見つかりません",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "名前を入力してください。",
   "roster.createForm.roleLabel": "ロール",
   "roster.createForm.rolePlaceholder": "例: リサーチアシスタント",
-  "roster.createForm.roleRequired": "ロールを入力してください。",
   "roster.createSuccess": "エージェントを作成しました。",
   "roster.dateTimeFormat": "YYYY/MM/DD HH:mm:ss",
   "roster.deleteDialog.description": "このエージェントはアクティブな Roster リストから削除されます。",
@@ -354,6 +372,7 @@
   "roster.empty": "エージェントはまだありません",
   "roster.emptySearch": "一致するエージェントがありません",
   "roster.filters.all": "すべて",
+  "roster.filters.createdByMe": "自分が作成",
   "roster.filters.drafts": "ドラフト",
   "roster.filters.label": "エージェントフィルター",
   "roster.filters.published": "公開済み",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "エージェントを Roster に保存しました。",
   "roster.searchLabel": "エージェントを検索",
   "roster.searchPlaceholder": "エージェントを名前で検索…",
+  "roster.sort.earliestCreated": "作成が古い順",
+  "roster.sort.label": "並び替え",
+  "roster.sort.lastModified": "最終更新",
+  "roster.sort.optionsLabel": "並び替えオプション",
+  "roster.sort.recentlyCreated": "最近作成",
   "roster.tabs.agent": "エージェント",
-  "roster.tabs.human": "人間",
+  "roster.tabs.human": "連絡先",
   "roster.tabsLabel": "Roster タイプ",
   "roster.updateSuccess": "エージェントを更新しました。",
   "roster.usageStatus.draft": "ドラフト"
diff --git a/web/i18n/ja-JP/common.json b/web/i18n/ja-JP/common.json
index ea9ec1f876e..3758341f94c 100644
--- a/web/i18n/ja-JP/common.json
+++ b/web/i18n/ja-JP/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "マーケットプレイスでインストールするか、設定でプロバイダーを設定してください。",
   "modelProvider.selector.onlyCompatibleModelsShown": "互換性のあるモデルのみが表示されます",
   "modelProvider.selector.showIncompatibleModels": "互換性のないモデルを表示",
+  "modelProvider.selector.suggestion": "提案",
+  "modelProvider.selector.suggestionTip": "エージェントタスクに推奨",
   "modelProvider.showModels": "モデルの表示",
   "modelProvider.speechToTextModel.key": "音声-to-テキストモデル",
   "modelProvider.speechToTextModel.tip": "会話での音声-to-テキスト入力に使用するデフォルトモデルを設定します。",
diff --git a/web/i18n/ja-JP/workflow.json b/web/i18n/ja-JP/workflow.json
index ea822b78e10..7b875ff1c79 100644
--- a/web/i18n/ja-JP/workflow.json
+++ b/web/i18n/ja-JP/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "最新のワークフローを公開し、接続済みの User Input ノードを用意してからツールとして設定してください。",
   "common.workflowAsToolTip": "ワークフロー更新後はツールの再設定が必要です",
   "common.workflowProcess": "ワークフロー処理",
+  "common.workflowProcessFailed": "ワークフロー処理が失敗しました",
+  "common.workflowProcessPaused": "ワークフロー処理は一時停止中です",
+  "common.workflowProcessRunning": "ワークフロー処理を実行中です",
+  "common.workflowProcessSucceeded": "ワークフロー処理が成功しました",
   "customWebhook": "カスタムWebhook",
   "debug.copyLastRun": "最後の実行をコピー",
   "debug.copyLastRunError": "最後の実行の入力をコピーできませんでした",
diff --git a/web/i18n/ko-KR/agent-v-2.json b/web/i18n/ko-KR/agent-v-2.json
index 29924bfca9f..08cf45d7c70 100644
--- a/web/i18n/ko-KR/agent-v-2.json
+++ b/web/i18n/ko-KR/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "{{key}} 값 표시",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "에이전트는 Linux 환경에서 실행됩니다. 여기에서 환경 변수를 정의하여 구성이나 시크릿을 미리 설정할 수 있습니다.\n\n또는 Build 채팅을 통해 언제든지 설정할 수 있습니다. <docLink>문서</docLink>를 참조하세요.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "도구에 필요한 API 키처럼 에이전트가 런타임에 읽는 값을 Prompt에 넣지 않고 설정합니다.\n\nBuild mode에서는 에이전트가 이를 대신 설정할 수 있습니다.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "환경 변수 범위 선택",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "에이전트는 Linux 환경에서 실행됩니다. 여기에서 환경 변수를 정의하여 구성이나 시크릿을 미리 설정할 수 있습니다. 또는 Build 채팅을 통해 언제든지 설정할 수 있습니다. 문서를 참조하세요.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "도구에 필요한 API 키처럼 에이전트가 런타임에 읽는 값을 Prompt에 넣지 않고 설정합니다.\n\nBuild mode에서는 에이전트가 이를 대신 설정할 수 있습니다.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Env Editor 전환",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "채팅으로 에이전트 만들기",
   "agentDetail.configure.build.inputPlaceholder": "에이전트가 해야 할 일을 설명하세요",
   "agentDetail.configure.build.startBuild": "빌드 시작",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "이 빌드의 지속적인 노트로, 에이전트가 세션 간 컨텍스트를 유지하는 데 사용합니다. 적용하면 파일에 추가됩니다.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "적용할 변경 사항 {{count}}개",
+  "agentDetail.configure.buildDraft.changesToApply_other": "적용할 변경 사항 {{count}}개",
   "agentDetail.configure.buildDraft.discard": "폐기",
   "agentDetail.configure.buildDraft.modeBadge": "빌드 모드",
   "agentDetail.configure.buildDraft.modeDescription": "빌드 모드입니다. 오른쪽 채팅으로 이 설정을 다듬은 뒤 적용하세요.",
+  "agentDetail.configure.buildDraft.rewritten": "다시 작성됨",
   "agentDetail.configure.buildDraft.title": "빌드 초안",
+  "agentDetail.configure.buildDraft.updated": "업데이트됨",
   "agentDetail.configure.chatFeatures.description": "Web app 및 채팅 화면에서의 최종 사용자 채팅 경험을 구성합니다.",
   "agentDetail.configure.chatFeatures.title": "채팅 기능",
   "agentDetail.configure.files.add": "파일 추가",
+  "agentDetail.configure.files.buildNote.generated": "생성됨",
+  "agentDetail.configure.files.buildNote.richTooltip": "에이전트가 Build mode에서 설정한 내용의 기록입니다. 모든 대화 시작 시 Prompt와 함께 이 기록을 읽습니다. <docLink>자세히 알아보기</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "에이전트가 Build mode에서 설정한 내용의 기록입니다. 모든 대화 시작 시 Prompt와 함께 이 기록을 읽습니다. 자세히 알아보기",
   "agentDetail.configure.files.empty.description": "사양, 템플릿, 가이드라인 등 에이전트가 읽을 수 있는 문서를 업로드하세요",
   "agentDetail.configure.files.empty.title": "아직 파일이 없습니다",
   "agentDetail.configure.files.label": "파일",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "미리보기를 불러오지 못했습니다.",
   "agentDetail.configure.files.preview.unsupported": "이 파일은 미리보기를 지원하지 않습니다.",
   "agentDetail.configure.files.remove": "{{name}} 제거",
-  "agentDetail.configure.files.tip": "프롬프트에 넣기에는 너무 긴 예시, 가이드라인 또는 정보 자체를 에이전트에 제공합니다. 프롬프트에서 /로 명시적으로 참조하세요.",
+  "agentDetail.configure.files.tip": "Prompt에 붙여넣기에는 너무 긴 참고 자료용입니다: 사양, 템플릿, 가이드라인. Prompt에서 /로 참조하세요.\n\nBuild mode에서는 에이전트가 이를 대신 설정할 수 있습니다.",
   "agentDetail.configure.files.toggle": "파일 전환",
   "agentDetail.configure.files.treeLabel": "에이전트 파일",
   "agentDetail.configure.files.upload.action": "업로드",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "{{name}} 제거",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "이 에이전트가 작업을 오케스트레이션할 때 사용할 수 있는 지식 검색 소스입니다.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "에이전트가 작업 중 검색할 지식 베이스를 연결하여 답변이 내 콘텐츠를 기반으로 하도록 합니다. <docLink>자세히 알아보기</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "에이전트가 작업 중 검색할 지식 베이스를 연결하여 답변이 내 콘텐츠를 기반으로 하도록 합니다. 자세히 알아보기",
   "agentDetail.configure.knowledgeRetrieval.toggle": "지식 검색 전환",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "메타데이터 필터링에서 모델을 구성하세요",
   "agentDetail.configure.model.label": "모델",
   "agentDetail.configure.orchestrate": "오케스트레이션",
   "agentDetail.configure.preview.chatFeatures": "채팅 기능",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "최종 사용자 인증",
   "agentDetail.configure.preview.inputPlaceholder": "{{name}}에게 메시지 보내기",
   "agentDetail.configure.preview.restart": "미리보기 다시 시작",
+  "agentDetail.configure.preview.sandboxNotice": "Agent는 Linux 샌드박스에서 실행됩니다.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Dify Community Edition에서는 각 Agent가 Docker 내부의 Linux 7.0.0-10060-aws 샌드박스 환경에서 실행됩니다. Build Chats를 통해 환경에 적용한 변경 사항은 유지됩니다.",
   "agentDetail.configure.preview.title": "미리보기",
   "agentDetail.configure.prompt.copied": "프롬프트를 복사했습니다",
   "agentDetail.configure.prompt.copy": "프롬프트 복사",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "자세히 알아보기",
   "agentDetail.configure.rightPanel.modeLabel": "에이전트 구성 모드",
   "agentDetail.configure.rightPanel.preview": "미리보기",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview는 Dify Cloud 및 Enterprise에서만 사용할 수 있습니다.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview는 사용자가 보게 될 방식으로 완성된 에이전트를 실행하며, 깔끔한 답변과 채팅 기능을 보여줍니다.",
   "agentDetail.configure.rightPanel.previewTipTitle": "에이전트 미리보기",
   "agentDetail.configure.skills.add": "스킬 추가",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "스킬",
   "agentDetail.configure.skills.label": "스킬",
   "agentDetail.configure.skills.remove": "{{name}} 제거",
-  "agentDetail.configure.skills.richTip": "에이전트에 모듈형 기능을 갖추세요. <docLink>문서</docLink>를 참조하세요. 프롬프트에서 /로 명시적으로 참조하세요.",
-  "agentDetail.configure.skills.tip": "에이전트에 모듈형 기능을 갖추세요. 문서를 참조하세요. 프롬프트에서 /로 명시적으로 참조하세요.",
+  "agentDetail.configure.skills.richTip": "반복 작업에 필요한 지침, 파일, 스크립트를 Skill로 패키징하세요. Prompt에서 /로 참조하세요. <docLink>자세히 알아보기</docLink>\n\nBuild mode에서는 에이전트가 이를 대신 설정할 수 있습니다.",
+  "agentDetail.configure.skills.tip": "반복 작업에 필요한 지침, 파일, 스크립트를 Skill로 패키징하세요. Prompt에서 /로 참조하세요. 자세히 알아보기\n\nBuild mode에서는 에이전트가 이를 대신 설정할 수 있습니다.",
   "agentDetail.configure.skills.toggle": "스킬 전환",
   "agentDetail.configure.skills.upload.action": "업로드",
   "agentDetail.configure.skills.upload.browse": "찾아보기",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "{{name}}의 추가 작업",
   "agentDetail.configure.tools.removeAction": "{{name}} 제거",
   "agentDetail.configure.tools.removeProvider": "모든 도구 제거",
-  "agentDetail.configure.tools.richTip": "에이전트에 Dify 플러그인을 갖추세요. <pluginDocLink>문서</pluginDocLink>를 참조하세요. 프롬프트에서 /로 명시적으로 참조하세요.\n또는 Build 채팅을 통해 에이전트에게 다른 도구(예: MCPs 및 CLI 도구)를 설치하고 인증하도록 지시할 수도 있습니다. Build 채팅에서 구성한 도구는 여기에 표시되지 않지만, 앞으로도 에이전트에서 계속 사용할 수 있습니다. <buildDocLink>문서</buildDocLink>를 참조하세요.",
-  "agentDetail.configure.tools.tip": "에이전트에 Dify 플러그인을 갖추세요. 문서를 참조하세요. 프롬프트에서 /로 명시적으로 참조하세요. 또는 Build 채팅을 통해 에이전트에게 다른 도구(예: MCPs 및 CLI 도구)를 설치하고 인증하도록 지시할 수도 있습니다. Build 채팅에서 구성한 도구는 여기에 표시되지 않지만, 앞으로도 에이전트에서 계속 사용할 수 있습니다. 문서를 참조하세요.",
+  "agentDetail.configure.tools.richTip": "워크스페이스의 Dify 도구를 추가하세요. Prompt에서 /로 참조하세요. <docLink>자세히 알아보기</docLink>\n\n에이전트가 sandbox에 설치한 도구는 여기에 표시되지 않지만 계속 사용할 수 있습니다.",
+  "agentDetail.configure.tools.tip": "워크스페이스의 Dify 도구를 추가하세요. Prompt에서 /로 참조하세요. 자세히 알아보기\n\n에이전트가 sandbox에 설치한 도구는 여기에 표시되지 않지만 계속 사용할 수 있습니다.",
   "agentDetail.configure.tools.toggle": "도구 전환",
   "agentDetail.configure.tools.toolTabs.all": "전체",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "플러그인",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "파일 시스템 경로",
   "agentDetail.configure.workingDirectory.description": "에이전트가 실행 중에 읽고 쓰는 파일입니다. qna_report.pdf 같은 생성 결과가 여기에 표시됩니다.",
+  "agentDetail.configure.workingDirectory.fileSystem": "파일 시스템",
+  "agentDetail.configure.workingDirectory.home": "홈",
   "agentDetail.configure.workingDirectory.open": "작업 디렉터리 열기",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "작업 디렉터리로 돌아가기",
   "agentDetail.configure.workingDirectory.title": "작업 디렉터리",
   "agentDetail.configure.workingDirectory.treeLabel": "작업 디렉터리 파일",
+  "agentDetail.configure.workingDirectory.workingDirectory": "작업 디렉터리",
   "agentDetail.documentTitle": "에이전트",
   "agentDetail.logs.description": "전체 로그는 사용자 입력, 에이전트 응답, 계획 및 도구 사용을 포함한 애플리케이션의 실행 상태를 기록합니다.",
   "agentDetail.logs.empty": "로그가 없습니다",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "이름을 입력하세요.",
   "roster.createForm.roleLabel": "역할",
   "roster.createForm.rolePlaceholder": "예: 리서치 어시스턴트",
-  "roster.createForm.roleRequired": "역할을 입력하세요.",
   "roster.createSuccess": "에이전트가 생성되었습니다.",
   "roster.dateTimeFormat": "YYYY-MM-DD HH:mm:ss",
   "roster.deleteDialog.description": "이 에이전트가 활성 Roster 목록에서 제거됩니다.",
@@ -354,6 +372,7 @@
   "roster.empty": "아직 에이전트가 없습니다",
   "roster.emptySearch": "일치하는 에이전트가 없습니다",
   "roster.filters.all": "전체",
+  "roster.filters.createdByMe": "내가 만든 항목",
   "roster.filters.drafts": "초안",
   "roster.filters.label": "에이전트 필터",
   "roster.filters.published": "게시됨",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "에이전트가 Roster에 저장되었습니다.",
   "roster.searchLabel": "에이전트 검색",
   "roster.searchPlaceholder": "이름으로 에이전트 검색…",
+  "roster.sort.earliestCreated": "오래된 생성순",
+  "roster.sort.label": "정렬 기준",
+  "roster.sort.lastModified": "마지막 수정",
+  "roster.sort.optionsLabel": "정렬 옵션",
+  "roster.sort.recentlyCreated": "최근 생성",
   "roster.tabs.agent": "에이전트",
-  "roster.tabs.human": "사람",
+  "roster.tabs.human": "연락처",
   "roster.tabsLabel": "Roster 유형",
   "roster.updateSuccess": "에이전트가 업데이트되었습니다.",
   "roster.usageStatus.draft": "초안"
diff --git a/web/i18n/ko-KR/common.json b/web/i18n/ko-KR/common.json
index f1635c32dad..a0040398bfe 100644
--- a/web/i18n/ko-KR/common.json
+++ b/web/i18n/ko-KR/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "마켓플레이스에서 설치하거나 설정에서 공급자를 구성하세요.",
   "modelProvider.selector.onlyCompatibleModelsShown": "호환되는 모델만 표시됩니다",
   "modelProvider.selector.showIncompatibleModels": "호환되지 않는 모델 표시",
+  "modelProvider.selector.suggestion": "추천",
+  "modelProvider.selector.suggestionTip": "에이전트 작업에 권장",
   "modelProvider.showModels": "모델 표시",
   "modelProvider.speechToTextModel.key": "음성-to-텍스트 모델",
   "modelProvider.speechToTextModel.tip": "대화에서의 음성-to-텍스트 입력에 사용되는 기본 모델을 설정합니다.",
diff --git a/web/i18n/ko-KR/workflow.json b/web/i18n/ko-KR/workflow.json
index 283020a7f04..83dfcba07d3 100644
--- a/web/i18n/ko-KR/workflow.json
+++ b/web/i18n/ko-KR/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "최신 워크플로를 게시하고 도구로 구성하기 전에 연결된 사용자 입력 노드가 있는지 확인하세요.",
   "common.workflowAsToolTip": "워크플로우 업데이트 후 도구 재구성이 필요합니다.",
   "common.workflowProcess": "워크플로우 과정",
+  "common.workflowProcessFailed": "워크플로우 과정 실패",
+  "common.workflowProcessPaused": "워크플로우 과정 일시 중지됨",
+  "common.workflowProcessRunning": "워크플로우 과정 실행 중",
+  "common.workflowProcessSucceeded": "워크플로우 과정 성공",
   "customWebhook": "맞춤 웹훅",
   "debug.copyLastRun": "마지막 실행 복사",
   "debug.copyLastRunError": "마지막 실행 입력을 복사하는 데 실패했습니다.",
diff --git a/web/i18n/nl-NL/agent-v-2.json b/web/i18n/nl-NL/agent-v-2.json
index ce86f4e4928..d4d48eb77e6 100644
--- a/web/i18n/nl-NL/agent-v-2.json
+++ b/web/i18n/nl-NL/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Waarde van {{key}} tonen",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Je agent draait in een Linux-omgeving. Je kunt configuraties of geheimen vooraf instellen door hier omgevingsvariabelen te definiëren.\n\nJe kunt dit ook altijd laten instellen via een Build-chat. Zie de <docLink>documentatie</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Stel waarden in die de agent tijdens runtime leest, zoals een API-sleutel die een tool nodig heeft, zonder ze in de Prompt te zetten.\n\nIn de Build-modus kan de agent dit voor u instellen.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Selecteer scope van omgevingsvariabele",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Je agent draait in een Linux-omgeving. Je kunt configuraties of geheimen vooraf instellen door hier omgevingsvariabelen te definiëren. Je kunt dit ook altijd laten instellen via een Build-chat. Zie de documentatie.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Stel waarden in die de agent tijdens runtime leest, zoals een API-sleutel die een tool nodig heeft, zonder ze in de Prompt te zetten.\n\nIn de Build-modus kan de agent dit voor u instellen.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Env Editor in/uit",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Bouw je agent via chat",
   "agentDetail.configure.build.inputPlaceholder": "Beschrijf wat je agent moet doen",
   "agentDetail.configure.build.startBuild": "Build starten",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Een doorlopende notitie van deze build die de agent gebruikt om context tussen sessies te behouden. Deze wordt aan je bestanden toegevoegd wanneer je toepast.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} wijziging om toe te passen",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} wijzigingen om toe te passen",
   "agentDetail.configure.buildDraft.discard": "Negeren",
   "agentDetail.configure.buildDraft.modeBadge": "Buildmodus",
   "agentDetail.configure.buildDraft.modeDescription": "Je bent in buildmodus. Werk deze configuratie bij via de chat rechts en pas daarna toe.",
+  "agentDetail.configure.buildDraft.rewritten": "Herschreven",
   "agentDetail.configure.buildDraft.title": "Buildconcept",
+  "agentDetail.configure.buildDraft.updated": "Bijgewerkt",
   "agentDetail.configure.chatFeatures.description": "Geef vorm aan de chatervaring voor eindgebruikers in je webapp en chatoppervlakken.",
   "agentDetail.configure.chatFeatures.title": "Chatfuncties",
   "agentDetail.configure.files.add": "Bestand toevoegen",
+  "agentDetail.configure.files.buildNote.generated": "Gegenereerd",
+  "agentDetail.configure.files.buildNote.richTooltip": "Het verslag van de agent van wat hij in de Build-modus heeft ingesteld. Hij leest dit aan het begin van elk gesprek, samen met uw Prompt. <docLink>Meer informatie</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Het verslag van de agent van wat hij in de Build-modus heeft ingesteld. Hij leest dit aan het begin van elk gesprek, samen met uw Prompt. Meer informatie",
   "agentDetail.configure.files.empty.description": "Upload documenten die de agent kan lezen, zoals specificaties, sjablonen of richtlijnen",
   "agentDetail.configure.files.empty.title": "Nog geen bestanden",
   "agentDetail.configure.files.label": "Bestanden",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Laden van voorbeeld mislukt.",
   "agentDetail.configure.files.preview.unsupported": "Dit bestand ondersteunt geen voorbeeldweergave.",
   "agentDetail.configure.files.remove": "{{name}} verwijderen",
-  "agentDetail.configure.files.tip": "Geef uw agent voorbeelden, richtlijnen of simpelweg informatie die te lang is voor de prompt. Verwijs er expliciet naar met / in de prompt.",
+  "agentDetail.configure.files.tip": "Voor referentiemateriaal dat te lang is om in de Prompt te plakken: specificaties, sjablonen, richtlijnen. Verwijs ernaar met / in de Prompt.\n\nIn de Build-modus kan de agent dit voor u instellen.",
   "agentDetail.configure.files.toggle": "Bestanden in/uit",
   "agentDetail.configure.files.treeLabel": "Agentbestanden",
   "agentDetail.configure.files.upload.action": "Uploaden",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "{{name}} verwijderen",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Kennisophalingsbronnen die deze agent kan gebruiken tijdens het orkestreren van taken.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Koppel kennisbanken die de agent tijdens het werken doorzoekt, zodat zijn antwoorden op uw content zijn gebaseerd. <docLink>Meer informatie</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Koppel kennisbanken die de agent tijdens het werken doorzoekt, zodat zijn antwoorden op uw content zijn gebaseerd. Meer informatie",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Kennisophaling in/uit",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Configureer een model in metadatafiltering",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Orkestreren",
   "agentDetail.configure.preview.chatFeatures": "Chatfuncties",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Authenticatie eindgebruiker",
   "agentDetail.configure.preview.inputPlaceholder": "Bericht sturen naar {{name}}",
   "agentDetail.configure.preview.restart": "Voorbeeld opnieuw starten",
+  "agentDetail.configure.preview.sandboxNotice": "De agent wordt uitgevoerd in een Linux-sandbox.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "In Dify Community Edition draait elk van je agents in een Linux 7.0.0-10060-aws-sandboxomgeving binnen je Docker. Je wijzigingen aan de omgeving via Build Chats blijven behouden.",
   "agentDetail.configure.preview.title": "Voorbeeld",
   "agentDetail.configure.prompt.copied": "Prompt gekopieerd",
   "agentDetail.configure.prompt.copy": "Prompt kopiëren",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Meer informatie",
   "agentDetail.configure.rightPanel.modeLabel": "Agentconfiguratiemodus",
   "agentDetail.configure.rightPanel.preview": "Voorbeeld",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview is alleen beschikbaar op Dify Cloud en Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview voert de voltooide agent uit zoals je gebruikers die zien, met duidelijke antwoorden en chatfuncties.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Voorbeeld van je agent",
   "agentDetail.configure.skills.add": "Skill toevoegen",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Skill",
   "agentDetail.configure.skills.label": "Skills",
   "agentDetail.configure.skills.remove": "{{name}} verwijderen",
-  "agentDetail.configure.skills.richTip": "Voorzie uw agent van modulaire mogelijkheden. Zie <docLink>documentatie</docLink>. Verwijs er expliciet naar met / in de prompt.",
-  "agentDetail.configure.skills.tip": "Voorzie uw agent van modulaire mogelijkheden. Zie documentatie. Verwijs er expliciet naar met / in de prompt.",
+  "agentDetail.configure.skills.richTip": "Bundel de instructies, bestanden en scripts voor een terugkerende taak in een skill. Verwijs ernaar met / in de Prompt. <docLink>Meer informatie</docLink>\n\nIn de Build-modus kan de agent dit voor u instellen.",
+  "agentDetail.configure.skills.tip": "Bundel de instructies, bestanden en scripts voor een terugkerende taak in een skill. Verwijs ernaar met / in de Prompt. Meer informatie\n\nIn de Build-modus kan de agent dit voor u instellen.",
   "agentDetail.configure.skills.toggle": "Skills in/uit",
   "agentDetail.configure.skills.upload.action": "Uploaden",
   "agentDetail.configure.skills.upload.browse": "bladeren",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Meer acties voor {{name}}",
   "agentDetail.configure.tools.removeAction": "{{name}} verwijderen",
   "agentDetail.configure.tools.removeProvider": "Alle tools verwijderen",
-  "agentDetail.configure.tools.richTip": "Voorzie uw agent van Dify-plugins. Zie <pluginDocLink>documentatie</pluginDocLink>. Verwijs er expliciet naar met / in de prompt.\nU kunt de agent ook altijd via een Build-chat vragen andere tools (bijv. MCPs en CLI-tools) te installeren en te authenticeren. Let op: tools die via Build-chats zijn geconfigureerd, verschijnen hier NIET, hoewel ze voortaan nog steeds beschikbaar zijn voor uw agent. Zie <buildDocLink>documentatie</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Voorzie uw agent van Dify-plugins. Zie documentatie. Verwijs er expliciet naar met / in de prompt. U kunt de agent ook altijd via een Build-chat vragen andere tools (bijv. MCPs en CLI-tools) te installeren en te authenticeren. Let op: tools die via Build-chats zijn geconfigureerd, verschijnen hier NIET, hoewel ze voortaan nog steeds beschikbaar zijn voor uw agent. Zie documentatie.",
+  "agentDetail.configure.tools.richTip": "Voeg Dify-tools uit uw workspace toe. Verwijs ernaar met / in de Prompt. <docLink>Meer informatie</docLink>\n\nTools die de agent in zijn sandbox installeert, verschijnen hier niet, maar blijven beschikbaar.",
+  "agentDetail.configure.tools.tip": "Voeg Dify-tools uit uw workspace toe. Verwijs ernaar met / in de Prompt. Meer informatie\n\nTools die de agent in zijn sandbox installeert, verschijnen hier niet, maar blijven beschikbaar.",
   "agentDetail.configure.tools.toggle": "Tools in/uit",
   "agentDetail.configure.tools.toolTabs.all": "Alle",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugins",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Bestandssysteempad",
   "agentDetail.configure.workingDirectory.description": "Bestanden die de agent leest en schrijft tijdens het uitvoeren. Alles wat hij genereert, zoals qna_report.pdf, verschijnt hier.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Bestandssysteem",
+  "agentDetail.configure.workingDirectory.home": "home",
   "agentDetail.configure.workingDirectory.open": "Werkmap openen",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Terug naar werkmap",
   "agentDetail.configure.workingDirectory.title": "Werkmap",
   "agentDetail.configure.workingDirectory.treeLabel": "Bestanden in de werkmap",
+  "agentDetail.configure.workingDirectory.workingDirectory": "werkmap",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Volledige logs registreren de status van de applicatie tijdens het uitvoeren, inclusief gebruikersinvoer, antwoorden van de agent, planning en gebruik van tools.",
   "agentDetail.logs.empty": "Geen logs gevonden",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Naam is verplicht.",
   "roster.createForm.roleLabel": "Rol",
   "roster.createForm.rolePlaceholder": "bv. Research Assistant",
-  "roster.createForm.roleRequired": "Rol is verplicht.",
   "roster.createSuccess": "Agent aangemaakt.",
   "roster.dateTimeFormat": "DD-MM-YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Deze agent wordt uit de actieve rosterlijst verwijderd.",
@@ -354,6 +372,7 @@
   "roster.empty": "Nog geen agent",
   "roster.emptySearch": "Geen overeenkomende agents",
   "roster.filters.all": "Alle",
+  "roster.filters.createdByMe": "Door mij gemaakt",
   "roster.filters.drafts": "Concepten",
   "roster.filters.label": "Agentfilters",
   "roster.filters.published": "Gepubliceerd",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent opgeslagen in roster.",
   "roster.searchLabel": "Agents zoeken",
   "roster.searchPlaceholder": "Agents op naam zoeken…",
+  "roster.sort.earliestCreated": "Eerst aangemaakt",
+  "roster.sort.label": "Sorteren op",
+  "roster.sort.lastModified": "Laatst gewijzigd",
+  "roster.sort.optionsLabel": "Sorteeropties",
+  "roster.sort.recentlyCreated": "Recent aangemaakt",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Mens",
+  "roster.tabs.human": "Contacten",
   "roster.tabsLabel": "Rostertype",
   "roster.updateSuccess": "Agent bijgewerkt.",
   "roster.usageStatus.draft": "Concept"
diff --git a/web/i18n/nl-NL/common.json b/web/i18n/nl-NL/common.json
index afffa257c1a..272157a52b6 100644
--- a/web/i18n/nl-NL/common.json
+++ b/web/i18n/nl-NL/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Blader in de Marketplace om er een te installeren, of configureer providers in de instellingen.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Alleen compatibele modellen worden weergegeven",
   "modelProvider.selector.showIncompatibleModels": "Incompatibele modellen weergeven",
+  "modelProvider.selector.suggestion": "Suggestie",
+  "modelProvider.selector.suggestionTip": "Aanbevolen voor agenttaken",
   "modelProvider.showModels": "Show Models",
   "modelProvider.speechToTextModel.key": "Speech-to-Text Model",
   "modelProvider.speechToTextModel.tip": "Set the default model for speech-to-text input in conversation.",
diff --git a/web/i18n/nl-NL/workflow.json b/web/i18n/nl-NL/workflow.json
index 1dfad69be76..ab70abf9ad1 100644
--- a/web/i18n/nl-NL/workflow.json
+++ b/web/i18n/nl-NL/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Publish the latest workflow and ensure a connected User Input node before configuring it as a tool.",
   "common.workflowAsToolTip": "Tool reconfiguration is required after the workflow update.",
   "common.workflowProcess": "Workflow Process",
+  "common.workflowProcessFailed": "Workflowproces mislukt",
+  "common.workflowProcessPaused": "Workflowproces gepauzeerd",
+  "common.workflowProcessRunning": "Workflowproces wordt uitgevoerd",
+  "common.workflowProcessSucceeded": "Workflowproces geslaagd",
   "customWebhook": "Custom Webhook",
   "debug.copyLastRun": "Copy Last Run",
   "debug.copyLastRunError": "Failed to copy last run inputs",
diff --git a/web/i18n/pl-PL/agent-v-2.json b/web/i18n/pl-PL/agent-v-2.json
index 9e1a2686539..fe4c96745d1 100644
--- a/web/i18n/pl-PL/agent-v-2.json
+++ b/web/i18n/pl-PL/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Pokaż wartość {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Twój agent działa w środowisku Linux. Możesz wstępnie ustawić konfiguracje lub sekrety, definiując tutaj zmienne środowiskowe.\n\nAlternatywnie zawsze możesz skonfigurować je przez czat Build. Zobacz <docLink>dokumentację</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Ustaw wartości, które agent odczytuje w runtime, takie jak klucz API potrzebny narzędziu, bez umieszczania ich w Prompt.\n\nW trybie Build agent może skonfigurować je za Ciebie.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Wybierz zakres zmiennej środowiskowej",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Twój agent działa w środowisku Linux. Możesz wstępnie ustawić konfiguracje lub sekrety, definiując tutaj zmienne środowiskowe. Alternatywnie zawsze możesz skonfigurować je przez czat Build. Zobacz dokumentację.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Ustaw wartości, które agent odczytuje w runtime, takie jak klucz API potrzebny narzędziu, bez umieszczania ich w Prompt.\n\nW trybie Build agent może skonfigurować je za Ciebie.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Przełącz Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Buduj agenta przez czat",
   "agentDetail.configure.build.inputPlaceholder": "Opisz, co agent ma robić",
   "agentDetail.configure.build.startBuild": "Rozpocznij budowanie",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Bieżąca notatka tej kompilacji, której agent używa do zachowania kontekstu między sesjami. Zostanie dodana do plików po zastosowaniu.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} zmiana do zastosowania",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} zmiany do zastosowania",
   "agentDetail.configure.buildDraft.discard": "Odrzuć",
   "agentDetail.configure.buildDraft.modeBadge": "Tryb budowania",
   "agentDetail.configure.buildDraft.modeDescription": "Jesteś w trybie budowania. Dostosuj tę konfigurację przez czat po prawej, a następnie zastosuj.",
+  "agentDetail.configure.buildDraft.rewritten": "Przepisano",
   "agentDetail.configure.buildDraft.title": "Szkic budowania",
+  "agentDetail.configure.buildDraft.updated": "Zaktualizowano",
   "agentDetail.configure.chatFeatures.description": "Ukształtuj doświadczenie czatu użytkownika końcowego w aplikacji webowej i powierzchniach czatu.",
   "agentDetail.configure.chatFeatures.title": "Funkcje czatu",
   "agentDetail.configure.files.add": "Dodaj plik",
+  "agentDetail.configure.files.buildNote.generated": "Wygenerowano",
+  "agentDetail.configure.files.buildNote.richTooltip": "Zapis agenta tego, co skonfigurował w trybie Build. Odczytuje go na początku każdej rozmowy razem z Twoim Promptem. <docLink>Dowiedz się więcej</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Zapis agenta tego, co skonfigurował w trybie Build. Odczytuje go na początku każdej rozmowy razem z Twoim Promptem. Dowiedz się więcej",
   "agentDetail.configure.files.empty.description": "Prześlij dokumenty, które agent może czytać, np. specyfikacje, szablony lub wytyczne",
   "agentDetail.configure.files.empty.title": "Brak plików",
   "agentDetail.configure.files.label": "Pliki",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Nie udało się załadować podglądu.",
   "agentDetail.configure.files.preview.unsupported": "Ten plik nie obsługuje podglądu.",
   "agentDetail.configure.files.remove": "Usuń {{name}}",
-  "agentDetail.configure.files.tip": "Przekaż agentowi dowolny przykład, wytyczne lub po prostu nośnik informacji, który jest zbyt długi dla promptu. Odwołaj się do niego jawnie za pomocą / w prompcie.",
+  "agentDetail.configure.files.tip": "Dla materiałów referencyjnych zbyt długich do wklejenia w Prompt: specyfikacje, szablony, wytyczne. Odwołuj się za pomocą / w Prompt.\n\nW trybie Build agent może skonfigurować je za Ciebie.",
   "agentDetail.configure.files.toggle": "Przełącz pliki",
   "agentDetail.configure.files.treeLabel": "Pliki agenta",
   "agentDetail.configure.files.upload.action": "Prześlij",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Usuń {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Źródła wyszukiwania wiedzy, których ten agent może używać podczas wykonywania zadań.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Połącz bazy wiedzy, które agent przeszukuje podczas pracy, aby jego odpowiedzi opierały się na Twoich treściach. <docLink>Dowiedz się więcej</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Połącz bazy wiedzy, które agent przeszukuje podczas pracy, aby jego odpowiedzi opierały się na Twoich treściach. Dowiedz się więcej",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Przełącz wyszukiwanie wiedzy",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Skonfiguruj model w filtrowaniu metadanych",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Orkiestracja",
   "agentDetail.configure.preview.chatFeatures": "Funkcje czatu",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Uwierzytelnianie użytkownika końcowego",
   "agentDetail.configure.preview.inputPlaceholder": "Wyślij wiadomość do {{name}}",
   "agentDetail.configure.preview.restart": "Uruchom podgląd ponownie",
+  "agentDetail.configure.preview.sandboxNotice": "Agent działa w piaskownicy Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "W Dify Community Edition każdy z Twoich agentów działa w środowisku piaskownicy Linux 7.0.0-10060-aws w Twoim Dockerze. Zmiany środowiska wprowadzone przez Build Chats są trwałe.",
   "agentDetail.configure.preview.title": "Podgląd",
   "agentDetail.configure.prompt.copied": "Prompt skopiowany",
   "agentDetail.configure.prompt.copy": "Kopiuj prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Dowiedz się więcej",
   "agentDetail.configure.rightPanel.modeLabel": "Tryb konfiguracji agenta",
   "agentDetail.configure.rightPanel.preview": "Podgląd",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview jest dostępny tylko w Dify Cloud i Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview uruchamia gotowego agenta tak, jak zobaczą go użytkownicy, z przejrzystymi odpowiedziami i funkcjami czatu.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Podgląd agenta",
   "agentDetail.configure.skills.add": "Dodaj umiejętność",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Umiejętność",
   "agentDetail.configure.skills.label": "Umiejętności",
   "agentDetail.configure.skills.remove": "Usuń {{name}}",
-  "agentDetail.configure.skills.richTip": "Wyposaż agenta w modułowe możliwości. Zobacz <docLink>dokumentację</docLink>. Odwołaj się do nich jawnie za pomocą / w prompcie.",
-  "agentDetail.configure.skills.tip": "Wyposaż agenta w modułowe możliwości. Zobacz dokumentację. Odwołaj się do nich jawnie za pomocą / w prompcie.",
+  "agentDetail.configure.skills.richTip": "Spakuj instrukcje, pliki i skrypty dla powtarzalnego zadania w skill. Odwołuj się za pomocą / w Prompt. <docLink>Dowiedz się więcej</docLink>\n\nW trybie Build agent może skonfigurować je za Ciebie.",
+  "agentDetail.configure.skills.tip": "Spakuj instrukcje, pliki i skrypty dla powtarzalnego zadania w skill. Odwołuj się za pomocą / w Prompt. Dowiedz się więcej\n\nW trybie Build agent może skonfigurować je za Ciebie.",
   "agentDetail.configure.skills.toggle": "Przełącz umiejętności",
   "agentDetail.configure.skills.upload.action": "Prześlij",
   "agentDetail.configure.skills.upload.browse": "przeglądaj",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Więcej akcji dla {{name}}",
   "agentDetail.configure.tools.removeAction": "Usuń {{name}}",
   "agentDetail.configure.tools.removeProvider": "Usuń wszystkie narzędzia",
-  "agentDetail.configure.tools.richTip": "Wyposaż agenta we wtyczki Dify. Zobacz <pluginDocLink>dokumentację</pluginDocLink>. Odwołaj się do nich jawnie za pomocą / w prompcie.\nAlternatywnie możesz zawsze polecić agentowi instalację i uwierzytelnienie innych narzędzi (np. MCPs i narzędzi CLI) przez czat Build. Pamiętaj, że narzędzia skonfigurowane przez czaty Build NIE pojawią się tutaj, choć nadal będą dostępne dla agenta w przyszłości. Zobacz <buildDocLink>dokumentację</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Wyposaż agenta we wtyczki Dify. Zobacz dokumentację. Odwołaj się do nich jawnie za pomocą / w prompcie. Alternatywnie możesz zawsze polecić agentowi instalację i uwierzytelnienie innych narzędzi (np. MCPs i narzędzi CLI) przez czat Build. Pamiętaj, że narzędzia skonfigurowane przez czaty Build NIE pojawią się tutaj, choć nadal będą dostępne dla agenta w przyszłości. Zobacz dokumentację.",
+  "agentDetail.configure.tools.richTip": "Dodaj narzędzia Dify ze swojego workspace. Odwołuj się za pomocą / w Prompt. <docLink>Dowiedz się więcej</docLink>\n\nNarzędzia instalowane przez agenta w jego sandbox nie pojawią się tutaj, ale pozostaną dla niego dostępne.",
+  "agentDetail.configure.tools.tip": "Dodaj narzędzia Dify ze swojego workspace. Odwołuj się za pomocą / w Prompt. Dowiedz się więcej\n\nNarzędzia instalowane przez agenta w jego sandbox nie pojawią się tutaj, ale pozostaną dla niego dostępne.",
   "agentDetail.configure.tools.toggle": "Przełącz narzędzia",
   "agentDetail.configure.tools.toolTabs.all": "Wszystkie",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Integracje",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Ścieżka systemu plików",
   "agentDetail.configure.workingDirectory.description": "Pliki, które agent odczytuje i zapisuje podczas działania. Wszystko, co wygeneruje, np. qna_report.pdf, pojawi się tutaj.",
+  "agentDetail.configure.workingDirectory.fileSystem": "System plików",
+  "agentDetail.configure.workingDirectory.home": "home",
   "agentDetail.configure.workingDirectory.open": "Otwórz katalog roboczy",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Wróć do katalogu roboczego",
   "agentDetail.configure.workingDirectory.title": "Katalog roboczy",
   "agentDetail.configure.workingDirectory.treeLabel": "Pliki katalogu roboczego",
+  "agentDetail.configure.workingDirectory.workingDirectory": "katalog roboczy",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Pełne logi rejestrują stan działania aplikacji, w tym dane wejściowe użytkownika, odpowiedzi agenta, planowanie i użycie narzędzi.",
   "agentDetail.logs.empty": "Nie znaleziono logów",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Nazwa jest wymagana.",
   "roster.createForm.roleLabel": "Rola",
   "roster.createForm.rolePlaceholder": "np. Asystent badawczy",
-  "roster.createForm.roleRequired": "Rola jest wymagana.",
   "roster.createSuccess": "Agent utworzony.",
   "roster.dateTimeFormat": "DD.MM.YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Ten agent zostanie usunięty z aktywnej listy Roster.",
@@ -354,6 +372,7 @@
   "roster.empty": "Brak agentów",
   "roster.emptySearch": "Brak pasujących agentów",
   "roster.filters.all": "Wszyscy",
+  "roster.filters.createdByMe": "Utworzone przeze mnie",
   "roster.filters.drafts": "Wersje robocze",
   "roster.filters.label": "Filtry agentów",
   "roster.filters.published": "Opublikowani",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent zapisany w Roster.",
   "roster.searchLabel": "Szukaj agentów",
   "roster.searchPlaceholder": "Wyszukaj agentów po nazwie…",
+  "roster.sort.earliestCreated": "Najwcześniej utworzone",
+  "roster.sort.label": "Sortuj według",
+  "roster.sort.lastModified": "Ostatnio zmodyfikowane",
+  "roster.sort.optionsLabel": "Opcje sortowania",
+  "roster.sort.recentlyCreated": "Ostatnio utworzone",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Człowiek",
+  "roster.tabs.human": "Kontakty",
   "roster.tabsLabel": "Typ Roster",
   "roster.updateSuccess": "Agent zaktualizowany.",
   "roster.usageStatus.draft": "Wersja robocza"
diff --git a/web/i18n/pl-PL/common.json b/web/i18n/pl-PL/common.json
index 712bb5a88e6..3943bd38479 100644
--- a/web/i18n/pl-PL/common.json
+++ b/web/i18n/pl-PL/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Przeglądaj Marketplace, aby zainstalować dostawcę, lub skonfiguruj dostawców w ustawieniach.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Wyświetlane są tylko kompatybilne modele",
   "modelProvider.selector.showIncompatibleModels": "Pokaż niezgodne modele",
+  "modelProvider.selector.suggestion": "Sugestia",
+  "modelProvider.selector.suggestionTip": "Zalecane do zadań agentów",
   "modelProvider.showModels": "Pokaż modele",
   "modelProvider.speechToTextModel.key": "Model mowy na tekst",
   "modelProvider.speechToTextModel.tip": "Ustaw domyślny model do przetwarzania mowy na tekst w rozmowach.",
diff --git a/web/i18n/pl-PL/workflow.json b/web/i18n/pl-PL/workflow.json
index 3ccbad6e0d3..19b10e33d50 100644
--- a/web/i18n/pl-PL/workflow.json
+++ b/web/i18n/pl-PL/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Opublikuj najnowszy przepływ pracy i upewnij się, że węzeł Wprowadzanie danych użytkownika jest połączony przed skonfigurowaniem go jako narzędzie.",
   "common.workflowAsToolTip": "Wymagana rekonfiguracja narzędzia po aktualizacji przepływu pracy.",
   "common.workflowProcess": "Proces przepływu pracy",
+  "common.workflowProcessFailed": "Proces przepływu pracy nie powiódł się",
+  "common.workflowProcessPaused": "Proces przepływu pracy wstrzymany",
+  "common.workflowProcessRunning": "Proces przepływu pracy w toku",
+  "common.workflowProcessSucceeded": "Proces przepływu pracy zakończony powodzeniem",
   "customWebhook": "Niestandardowy webhook",
   "debug.copyLastRun": "Kopiuj ostatnie uruchomienie",
   "debug.copyLastRunError": "Nie udało się skopiować danych wejściowych z ostatniego uruchomienia",
diff --git a/web/i18n/pt-BR/agent-v-2.json b/web/i18n/pt-BR/agent-v-2.json
index f5751c45a16..39671d80b88 100644
--- a/web/i18n/pt-BR/agent-v-2.json
+++ b/web/i18n/pt-BR/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Revelar valor de {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Seu agente é executado em um ambiente Linux. Você pode predefinir configurações ou segredos definindo variáveis de ambiente aqui.\n\nComo alternativa, você sempre pode configurar tudo por meio de um chat Build. Consulte a <docLink>documentação</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Defina valores que o agente lê em runtime, como uma chave de API necessária para uma ferramenta, sem colocá-los no Prompt.\n\nNo modo Build, o agente pode configurar isso para você.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Selecionar escopo da variável de ambiente",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Seu agente é executado em um ambiente Linux. Você pode predefinir configurações ou segredos definindo variáveis de ambiente aqui. Como alternativa, você sempre pode configurar tudo por meio de um chat Build. Consulte a documentação.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Defina valores que o agente lê em runtime, como uma chave de API necessária para uma ferramenta, sem colocá-los no Prompt.\n\nNo modo Build, o agente pode configurar isso para você.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Alternar Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Crie seu agente conversando",
   "agentDetail.configure.build.inputPlaceholder": "Descreva o que seu agente deve fazer",
   "agentDetail.configure.build.startBuild": "Iniciar build",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Uma nota contínua desta build que o agente usa para manter o contexto entre sessões. Ela é adicionada aos seus arquivos ao aplicar.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} alteração para aplicar",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} alterações para aplicar",
   "agentDetail.configure.buildDraft.discard": "Descartar",
   "agentDetail.configure.buildDraft.modeBadge": "Modo build",
   "agentDetail.configure.buildDraft.modeDescription": "Você está no modo build. Ajuste esta configuração pelo chat à direita e depois aplique.",
+  "agentDetail.configure.buildDraft.rewritten": "Reescrito",
   "agentDetail.configure.buildDraft.title": "Rascunho de build",
+  "agentDetail.configure.buildDraft.updated": "Atualizado",
   "agentDetail.configure.chatFeatures.description": "Modele a experiência de chat do usuário final no seu webapp e superfícies de chat.",
   "agentDetail.configure.chatFeatures.title": "Recursos de chat",
   "agentDetail.configure.files.add": "Adicionar arquivo",
+  "agentDetail.configure.files.buildNote.generated": "Gerado",
+  "agentDetail.configure.files.buildNote.richTooltip": "O registro do agente do que ele configurou no modo Build. Ele lê isso no início de cada conversa, junto com seu Prompt. <docLink>Saiba mais</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "O registro do agente do que ele configurou no modo Build. Ele lê isso no início de cada conversa, junto com seu Prompt. Saiba mais",
   "agentDetail.configure.files.empty.description": "Envie documentos que o agente possa ler, como especificações, modelos ou diretrizes",
   "agentDetail.configure.files.empty.title": "Ainda não há arquivos",
   "agentDetail.configure.files.label": "Arquivos",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Falha ao carregar a pré-visualização.",
   "agentDetail.configure.files.preview.unsupported": "Este arquivo não oferece suporte à pré-visualização.",
   "agentDetail.configure.files.remove": "Remover {{name}}",
-  "agentDetail.configure.files.tip": "Forneça ao seu agente qualquer exemplo, diretriz ou simplesmente um portador de informações que seja longo demais para o prompt. Faça referência explícita com / no prompt.",
+  "agentDetail.configure.files.tip": "Para materiais de referência longos demais para colar no Prompt: especificações, modelos, diretrizes. Referencie com / no Prompt.\n\nNo modo Build, o agente pode configurar isso para você.",
   "agentDetail.configure.files.toggle": "Alternar arquivos",
   "agentDetail.configure.files.treeLabel": "Arquivos do agente",
   "agentDetail.configure.files.upload.action": "Enviar",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Remover {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Fontes de recuperação de conhecimento que este agente pode usar ao orquestrar tarefas.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Conecte bases de conhecimento que o agente pesquisa enquanto trabalha, para que as respostas usem seu conteúdo. <docLink>Saiba mais</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Conecte bases de conhecimento que o agente pesquisa enquanto trabalha, para que as respostas usem seu conteúdo. Saiba mais",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Alternar recuperação de conhecimento",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Configure um modelo na filtragem de metadados",
   "agentDetail.configure.model.label": "Modelo",
   "agentDetail.configure.orchestrate": "Orquestrar",
   "agentDetail.configure.preview.chatFeatures": "Recursos de chat",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Autenticação do usuário final",
   "agentDetail.configure.preview.inputPlaceholder": "Enviar mensagem para {{name}}",
   "agentDetail.configure.preview.restart": "Reiniciar pré-visualização",
+  "agentDetail.configure.preview.sandboxNotice": "O agente é executado em uma sandbox Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "No Dify Community Edition, cada um dos seus agents é executado em um ambiente sandbox Linux 7.0.0-10060-aws dentro do seu Docker. Suas edições no ambiente via Build Chats são persistentes.",
   "agentDetail.configure.preview.title": "Pré-visualização",
   "agentDetail.configure.prompt.copied": "Prompt copiado",
   "agentDetail.configure.prompt.copy": "Copiar prompt",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Saiba mais",
   "agentDetail.configure.rightPanel.modeLabel": "Modo de configuração do agente",
   "agentDetail.configure.rightPanel.preview": "Pré-visualização",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview está disponível apenas no Dify Cloud e Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview executa o agente finalizado como seus usuários o verão, com respostas claras e recursos de chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Pré-visualize seu agente",
   "agentDetail.configure.skills.add": "Adicionar habilidade",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Habilidade",
   "agentDetail.configure.skills.label": "Habilidades",
   "agentDetail.configure.skills.remove": "Remover {{name}}",
-  "agentDetail.configure.skills.richTip": "Equipe seu agente com capacidades modulares. Consulte a <docLink>documentação</docLink>. Faça referência explícita com / no prompt.",
-  "agentDetail.configure.skills.tip": "Equipe seu agente com capacidades modulares. Consulte a documentação. Faça referência explícita com / no prompt.",
+  "agentDetail.configure.skills.richTip": "Empacote as instruções, arquivos e scripts de uma tarefa recorrente em uma skill. Referencie com / no Prompt. <docLink>Saiba mais</docLink>\n\nNo modo Build, o agente pode configurar isso para você.",
+  "agentDetail.configure.skills.tip": "Empacote as instruções, arquivos e scripts de uma tarefa recorrente em uma skill. Referencie com / no Prompt. Saiba mais\n\nNo modo Build, o agente pode configurar isso para você.",
   "agentDetail.configure.skills.toggle": "Alternar habilidades",
   "agentDetail.configure.skills.upload.action": "Enviar",
   "agentDetail.configure.skills.upload.browse": "procurar",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Mais ações para {{name}}",
   "agentDetail.configure.tools.removeAction": "Remover {{name}}",
   "agentDetail.configure.tools.removeProvider": "Remover todas as ferramentas",
-  "agentDetail.configure.tools.richTip": "Equipe seu agente com os plugins do Dify. Consulte a <pluginDocLink>documentação</pluginDocLink>. Faça referência explícita com / no prompt.\nComo alternativa, você sempre pode dizer ao agente para instalar e autenticar outras ferramentas (por exemplo, MCPs e ferramentas CLI) via um chat de Build. Observe que as ferramentas configuradas por chats de Build NÃO aparecerão aqui, embora continuem disponíveis para o seu agente daqui em diante. Consulte a <buildDocLink>documentação</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Equipe seu agente com os plugins do Dify. Consulte a documentação. Faça referência explícita com / no prompt. Como alternativa, você sempre pode dizer ao agente para instalar e autenticar outras ferramentas (por exemplo, MCPs e ferramentas CLI) via um chat de Build. Observe que as ferramentas configuradas por chats de Build NÃO aparecerão aqui, embora continuem disponíveis para o seu agente daqui em diante. Consulte a documentação.",
+  "agentDetail.configure.tools.richTip": "Adicione ferramentas Dify do seu workspace. Referencie com / no Prompt. <docLink>Saiba mais</docLink>\n\nAs ferramentas que o agente instala no sandbox não aparecerão aqui, mas continuarão disponíveis para ele.",
+  "agentDetail.configure.tools.tip": "Adicione ferramentas Dify do seu workspace. Referencie com / no Prompt. Saiba mais\n\nAs ferramentas que o agente instala no sandbox não aparecerão aqui, mas continuarão disponíveis para ele.",
   "agentDetail.configure.tools.toggle": "Alternar ferramentas",
   "agentDetail.configure.tools.toolTabs.all": "Todas",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugins",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Caminho do sistema de arquivos",
   "agentDetail.configure.workingDirectory.description": "Arquivos que o agente lê e grava enquanto executa. Tudo o que ele gera, como qna_report.pdf, aparece aqui.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Sistema de arquivos",
+  "agentDetail.configure.workingDirectory.home": "início",
   "agentDetail.configure.workingDirectory.open": "Abrir diretório de trabalho",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Voltar ao diretório de trabalho",
   "agentDetail.configure.workingDirectory.title": "Diretório de trabalho",
   "agentDetail.configure.workingDirectory.treeLabel": "Arquivos do diretório de trabalho",
+  "agentDetail.configure.workingDirectory.workingDirectory": "diretório de trabalho",
   "agentDetail.documentTitle": "Agente",
   "agentDetail.logs.description": "Os logs completos registram o status de execução do aplicativo, incluindo entradas do usuário, respostas do agente, planejamento e uso de ferramentas.",
   "agentDetail.logs.empty": "Nenhum log encontrado",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Nome é obrigatório.",
   "roster.createForm.roleLabel": "Função",
   "roster.createForm.rolePlaceholder": "ex.: Assistente de pesquisa",
-  "roster.createForm.roleRequired": "Função é obrigatória.",
   "roster.createSuccess": "Agente criado.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Este agente será removido da lista do roster ativo.",
@@ -354,6 +372,7 @@
   "roster.empty": "Ainda não há agente",
   "roster.emptySearch": "Nenhum agente correspondente",
   "roster.filters.all": "Todos",
+  "roster.filters.createdByMe": "Criados por mim",
   "roster.filters.drafts": "Rascunhos",
   "roster.filters.label": "Filtros de agente",
   "roster.filters.published": "Publicados",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agente salvo no roster.",
   "roster.searchLabel": "Pesquisar agentes",
   "roster.searchPlaceholder": "Pesquisar agentes pelo nome…",
+  "roster.sort.earliestCreated": "Criados primeiro",
+  "roster.sort.label": "Ordenar por",
+  "roster.sort.lastModified": "Última modificação",
+  "roster.sort.optionsLabel": "Opções de ordenação",
+  "roster.sort.recentlyCreated": "Criados recentemente",
   "roster.tabs.agent": "Agente",
-  "roster.tabs.human": "Humano",
+  "roster.tabs.human": "Contatos",
   "roster.tabsLabel": "Tipo de roster",
   "roster.updateSuccess": "Agente atualizado.",
   "roster.usageStatus.draft": "Rascunho"
diff --git a/web/i18n/pt-BR/common.json b/web/i18n/pt-BR/common.json
index 1677b739f07..3fb45b62ccd 100644
--- a/web/i18n/pt-BR/common.json
+++ b/web/i18n/pt-BR/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Navegue pelo Marketplace para instalar um, ou configure provedores nas configurações.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Apenas modelos compatíveis são exibidos",
   "modelProvider.selector.showIncompatibleModels": "Mostrar modelos incompatíveis",
+  "modelProvider.selector.suggestion": "Sugestão",
+  "modelProvider.selector.suggestionTip": "Recomendado para tarefas de agente",
   "modelProvider.showModels": "Mostrar Modelos",
   "modelProvider.speechToTextModel.key": "Modelo de Fala para Texto",
   "modelProvider.speechToTextModel.tip": "Defina o modelo padrão para entrada de fala para texto na conversa.",
diff --git a/web/i18n/pt-BR/workflow.json b/web/i18n/pt-BR/workflow.json
index 35cbb6dd623..d35f798a02c 100644
--- a/web/i18n/pt-BR/workflow.json
+++ b/web/i18n/pt-BR/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Publique o fluxo de trabalho mais recente e garanta que haja um nó de Entrada do Usuário conectado antes de configurá-lo como uma ferramenta.",
   "common.workflowAsToolTip": "É necessária a reconfiguração da ferramenta após a atualização do fluxo de trabalho.",
   "common.workflowProcess": "Processo de fluxo de trabalho",
+  "common.workflowProcessFailed": "Processo de fluxo de trabalho falhou",
+  "common.workflowProcessPaused": "Processo de fluxo de trabalho pausado",
+  "common.workflowProcessRunning": "Processo de fluxo de trabalho em execução",
+  "common.workflowProcessSucceeded": "Processo de fluxo de trabalho bem-sucedido",
   "customWebhook": "Webhook Personalizado",
   "debug.copyLastRun": "Copiar Última Execução",
   "debug.copyLastRunError": "Falha ao copiar as entradas da última execução",
diff --git a/web/i18n/ro-RO/agent-v-2.json b/web/i18n/ro-RO/agent-v-2.json
index 5769b77387a..7390fe70517 100644
--- a/web/i18n/ro-RO/agent-v-2.json
+++ b/web/i18n/ro-RO/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Dezvăluie valoarea {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Agentul tău rulează într-un mediu Linux. Poți preseta configurații sau secrete definind aici variabile de mediu.\n\nAlternativ, poți configura oricând lucrurile printr-un chat Build. Consultă <docLink>documentația</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Setați valorile pe care agentul le citește la runtime, cum ar fi o cheie API necesară unui instrument, fără să le puneți în Prompt.\n\nÎn modul Build, agentul le poate configura pentru dvs.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Selectați domeniul variabilei de mediu",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Agentul tău rulează într-un mediu Linux. Poți preseta configurații sau secrete definind aici variabile de mediu. Alternativ, poți configura oricând lucrurile printr-un chat Build. Consultă documentația.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Setați valorile pe care agentul le citește la runtime, cum ar fi o cheie API necesară unui instrument, fără să le puneți în Prompt.\n\nÎn modul Build, agentul le poate configura pentru dvs.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Comută Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Construiește agentul prin chat",
   "agentDetail.configure.build.inputPlaceholder": "Descrie ce ar trebui să facă agentul tău",
   "agentDetail.configure.build.startBuild": "Pornește build-ul",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "O notă continuă a acestui build pe care agentul o folosește pentru a păstra contextul între sesiuni. Este adăugată în fișiere când aplici.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} modificare de aplicat",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} modificări de aplicat",
   "agentDetail.configure.buildDraft.discard": "Renunță",
   "agentDetail.configure.buildDraft.modeBadge": "Mod build",
   "agentDetail.configure.buildDraft.modeDescription": "Ești în modul build. Ajustează această configurare prin chatul din dreapta, apoi aplică.",
+  "agentDetail.configure.buildDraft.rewritten": "Rescris",
   "agentDetail.configure.buildDraft.title": "Schiță de build",
+  "agentDetail.configure.buildDraft.updated": "Actualizat",
   "agentDetail.configure.chatFeatures.description": "Modelează experiența de chat a utilizatorului final pe webapp-ul tău și pe suprafețele de chat.",
   "agentDetail.configure.chatFeatures.title": "Funcții de chat",
   "agentDetail.configure.files.add": "Adaugă fișier",
+  "agentDetail.configure.files.buildNote.generated": "Generat",
+  "agentDetail.configure.files.buildNote.richTooltip": "Înregistrarea agentului despre ce a configurat în modul Build. O citește la începutul fiecărei conversații, împreună cu Promptul dvs. <docLink>Aflați mai multe</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Înregistrarea agentului despre ce a configurat în modul Build. O citește la începutul fiecărei conversații, împreună cu Promptul dvs. Aflați mai multe",
   "agentDetail.configure.files.empty.description": "Încarcă documente pe care agentul le poate citi, precum specificații, șabloane sau ghiduri",
   "agentDetail.configure.files.empty.title": "Niciun fișier încă",
   "agentDetail.configure.files.label": "Fișiere",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Încărcarea previzualizării a eșuat.",
   "agentDetail.configure.files.preview.unsupported": "Acest fișier nu acceptă previzualizarea.",
   "agentDetail.configure.files.remove": "Elimină {{name}}",
-  "agentDetail.configure.files.tip": "Oferiți agentului orice exemplu, ghid sau pur și simplu un suport de informații prea lung pentru prompt. Faceți referire explicită cu / în prompt.",
+  "agentDetail.configure.files.tip": "Pentru materiale de referință prea lungi pentru a fi lipite în Prompt: specificații, șabloane, ghiduri. Faceți referire cu / în Prompt.\n\nÎn modul Build, agentul le poate configura pentru dvs.",
   "agentDetail.configure.files.toggle": "Comută fișierele",
   "agentDetail.configure.files.treeLabel": "Fișierele agentului",
   "agentDetail.configure.files.upload.action": "Încarcă",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Elimină {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Sursele de recuperare de cunoștințe pe care acest agent le poate folosi în timpul orchestrării sarcinilor.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Conectați baze de cunoștințe pe care agentul le caută în timp ce lucrează, astfel încât răspunsurile sale să se bazeze pe conținutul dvs. <docLink>Aflați mai multe</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Conectați baze de cunoștințe pe care agentul le caută în timp ce lucrează, astfel încât răspunsurile sale să se bazeze pe conținutul dvs. Aflați mai multe",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Comută recuperarea de cunoștințe",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Configurați un model în filtrarea metadatelor",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Orchestrare",
   "agentDetail.configure.preview.chatFeatures": "Funcții de chat",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Autentificare utilizator final",
   "agentDetail.configure.preview.inputPlaceholder": "Trimite mesaj către {{name}}",
   "agentDetail.configure.preview.restart": "Reporniți previzualizarea",
+  "agentDetail.configure.preview.sandboxNotice": "Agentul rulează într-un sandbox Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "În Dify Community Edition, fiecare dintre agenții tăi rulează într-un mediu sandbox Linux 7.0.0-10060-aws în Dockerul tău. Modificările aduse mediului prin Build Chats sunt persistente.",
   "agentDetail.configure.preview.title": "Previzualizare",
   "agentDetail.configure.prompt.copied": "Prompt copiat",
   "agentDetail.configure.prompt.copy": "Copiază prompt-ul",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Află mai multe",
   "agentDetail.configure.rightPanel.modeLabel": "Mod de configurare agent",
   "agentDetail.configure.rightPanel.preview": "Previzualizare",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview este disponibil doar pe Dify Cloud și Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview rulează agentul finalizat așa cum îl vor vedea utilizatorii, cu răspunsuri clare și funcții de chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Previzualizează agentul",
   "agentDetail.configure.skills.add": "Adaugă abilitate",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Abilitate",
   "agentDetail.configure.skills.label": "Abilități",
   "agentDetail.configure.skills.remove": "Elimină {{name}}",
-  "agentDetail.configure.skills.richTip": "Echipați agentul cu capabilități modulare. Consultați <docLink>documentația</docLink>. Faceți referire explicită cu / în prompt.",
-  "agentDetail.configure.skills.tip": "Echipați agentul cu capabilități modulare. Consultați documentația. Faceți referire explicită cu / în prompt.",
+  "agentDetail.configure.skills.richTip": "Împachetați instrucțiunile, fișierele și scripturile pentru o sarcină recurentă într-un skill. Faceți referire cu / în Prompt. <docLink>Aflați mai multe</docLink>\n\nÎn modul Build, agentul le poate configura pentru dvs.",
+  "agentDetail.configure.skills.tip": "Împachetați instrucțiunile, fișierele și scripturile pentru o sarcină recurentă într-un skill. Faceți referire cu / în Prompt. Aflați mai multe\n\nÎn modul Build, agentul le poate configura pentru dvs.",
   "agentDetail.configure.skills.toggle": "Comută abilitățile",
   "agentDetail.configure.skills.upload.action": "Încarcă",
   "agentDetail.configure.skills.upload.browse": "răsfoiește",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Mai multe acțiuni pentru {{name}}",
   "agentDetail.configure.tools.removeAction": "Eliminați {{name}}",
   "agentDetail.configure.tools.removeProvider": "Eliminați toate instrumentele",
-  "agentDetail.configure.tools.richTip": "Echipați agentul cu pluginurile Dify. Consultați <pluginDocLink>documentația</pluginDocLink>. Faceți referire explicită cu / în prompt.\nAlternativ, puteți oricând să îi spuneți agentului să instaleze și să autentifice alte instrumente (de ex. MCPs și instrumente CLI) printr-un chat Build. Rețineți că instrumentele configurate prin chat-uri Build NU vor apărea aici, deși vor rămâne disponibile pentru agentul dvs. în continuare. Consultați <buildDocLink>documentația</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Echipați agentul cu pluginurile Dify. Consultați documentația. Faceți referire explicită cu / în prompt. Alternativ, puteți oricând să îi spuneți agentului să instaleze și să autentifice alte instrumente (de ex. MCPs și instrumente CLI) printr-un chat Build. Rețineți că instrumentele configurate prin chat-uri Build NU vor apărea aici, deși vor rămâne disponibile pentru agentul dvs. în continuare. Consultați documentația.",
+  "agentDetail.configure.tools.richTip": "Adăugați instrumente Dify din workspace. Faceți referire cu / în Prompt. <docLink>Aflați mai multe</docLink>\n\nInstrumentele pe care agentul le instalează în sandbox nu vor apărea aici, dar îi vor rămâne disponibile.",
+  "agentDetail.configure.tools.tip": "Adăugați instrumente Dify din workspace. Faceți referire cu / în Prompt. Aflați mai multe\n\nInstrumentele pe care agentul le instalează în sandbox nu vor apărea aici, dar îi vor rămâne disponibile.",
   "agentDetail.configure.tools.toggle": "Comută instrumentele",
   "agentDetail.configure.tools.toolTabs.all": "Toate",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugin-uri",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Calea sistemului de fișiere",
   "agentDetail.configure.workingDirectory.description": "Fișierele pe care agentul le citește și le scrie în timpul rulării. Tot ce generează, precum qna_report.pdf, apare aici.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Sistem de fișiere",
+  "agentDetail.configure.workingDirectory.home": "acasă",
   "agentDetail.configure.workingDirectory.open": "Deschide directorul de lucru",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Înapoi la directorul de lucru",
   "agentDetail.configure.workingDirectory.title": "Director de lucru",
   "agentDetail.configure.workingDirectory.treeLabel": "Fișiere din directorul de lucru",
+  "agentDetail.configure.workingDirectory.workingDirectory": "director de lucru",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Jurnalele complete înregistrează starea de execuție a aplicației, inclusiv intrările utilizatorilor, răspunsurile agentului, planificarea și utilizarea instrumentelor.",
   "agentDetail.logs.empty": "Niciun jurnal găsit",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Numele este obligatoriu.",
   "roster.createForm.roleLabel": "Rol",
   "roster.createForm.rolePlaceholder": "de ex. Asistent de cercetare",
-  "roster.createForm.roleRequired": "Rolul este obligatoriu.",
   "roster.createSuccess": "Agent creat.",
   "roster.dateTimeFormat": "DD.MM.YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Acest agent va fi eliminat din lista activă a roster-ului.",
@@ -354,6 +372,7 @@
   "roster.empty": "Niciun agent încă",
   "roster.emptySearch": "Niciun agent corespunzător",
   "roster.filters.all": "Toți",
+  "roster.filters.createdByMe": "Create de mine",
   "roster.filters.drafts": "Ciorne",
   "roster.filters.label": "Filtre agent",
   "roster.filters.published": "Publicați",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent salvat în roster.",
   "roster.searchLabel": "Căutați agenți",
   "roster.searchPlaceholder": "Căutați agenți după nume…",
+  "roster.sort.earliestCreated": "Create mai întâi",
+  "roster.sort.label": "Sortează după",
+  "roster.sort.lastModified": "Ultima modificare",
+  "roster.sort.optionsLabel": "Opțiuni de sortare",
+  "roster.sort.recentlyCreated": "Create recent",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Uman",
+  "roster.tabs.human": "Contacte",
   "roster.tabsLabel": "Tip de roster",
   "roster.updateSuccess": "Agent actualizat.",
   "roster.usageStatus.draft": "Ciornă"
diff --git a/web/i18n/ro-RO/common.json b/web/i18n/ro-RO/common.json
index a3ab4e8b4e4..8e070517325 100644
--- a/web/i18n/ro-RO/common.json
+++ b/web/i18n/ro-RO/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Răsfoiți Marketplace pentru a instala unul sau configurați furnizorii în setări.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Sunt afișate doar modelele compatibile",
   "modelProvider.selector.showIncompatibleModels": "Afișează modelele incompatibile",
+  "modelProvider.selector.suggestion": "Sugestie",
+  "modelProvider.selector.suggestionTip": "Recomandat pentru sarcinile agentului",
   "modelProvider.showModels": "Arată modele",
   "modelProvider.speechToTextModel.key": "Model de conversie text-la-vorbire",
   "modelProvider.speechToTextModel.tip": "Setați modelul implicit pentru intrarea de conversie text-la-vorbire în conversație.",
diff --git a/web/i18n/ro-RO/workflow.json b/web/i18n/ro-RO/workflow.json
index 880d3fd365b..4ca4078118d 100644
--- a/web/i18n/ro-RO/workflow.json
+++ b/web/i18n/ro-RO/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Publicați fluxul de lucru cel mai recent și asigurați-vă că există un nod de Intrare Utilizator conectat înainte de a-l configura ca instrument.",
   "common.workflowAsToolTip": "Reconfigurarea instrumentului este necesară după actualizarea fluxului de lucru.",
   "common.workflowProcess": "Proces de flux de lucru",
+  "common.workflowProcessFailed": "Procesul de flux de lucru a eșuat",
+  "common.workflowProcessPaused": "Procesul de flux de lucru este întrerupt",
+  "common.workflowProcessRunning": "Procesul de flux de lucru rulează",
+  "common.workflowProcessSucceeded": "Procesul de flux de lucru a reușit",
   "customWebhook": "Webhook personalizat",
   "debug.copyLastRun": "Copiază ultima execuție",
   "debug.copyLastRunError": "Nu s-au putut copia ultimele intrări de rulare",
diff --git a/web/i18n/ru-RU/agent-v-2.json b/web/i18n/ru-RU/agent-v-2.json
index 8ff226ea64d..c3378ae51fd 100644
--- a/web/i18n/ru-RU/agent-v-2.json
+++ b/web/i18n/ru-RU/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Показать значение {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Ваш агент работает в среде Linux. Вы можете заранее задать конфигурации или секреты, определив здесь переменные окружения.\n\nТакже вы всегда можете настроить всё через чат Build. См. <docLink>документацию</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Задайте значения, которые агент читает во время выполнения, например API-ключ, нужный инструменту, не помещая их в Prompt.\n\nВ режиме Build агент может настроить это за вас.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Выберите область переменной окружения",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Ваш агент работает в среде Linux. Вы можете заранее задать конфигурации или секреты, определив здесь переменные окружения. Также вы всегда можете настроить всё через чат Build. См. документацию.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Задайте значения, которые агент читает во время выполнения, например API-ключ, нужный инструменту, не помещая их в Prompt.\n\nВ режиме Build агент может настроить это за вас.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Переключить Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Создайте агента в чате",
   "agentDetail.configure.build.inputPlaceholder": "Опишите, что должен делать агент",
   "agentDetail.configure.build.startBuild": "Начать сборку",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Постоянная заметка этой сборки, которую агент использует, чтобы сохранять контекст между сеансами. Она добавляется в ваши файлы при применении.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} изменение для применения",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} изменений для применения",
   "agentDetail.configure.buildDraft.discard": "Отменить",
   "agentDetail.configure.buildDraft.modeBadge": "Режим сборки",
   "agentDetail.configure.buildDraft.modeDescription": "Вы в режиме сборки. Настройте эту конфигурацию через чат справа, затем примените изменения.",
+  "agentDetail.configure.buildDraft.rewritten": "Переписано",
   "agentDetail.configure.buildDraft.title": "Черновик сборки",
+  "agentDetail.configure.buildDraft.updated": "Обновлено",
   "agentDetail.configure.chatFeatures.description": "Настройте чат-опыт конечного пользователя в вашем веб-приложении и чат-поверхностях.",
   "agentDetail.configure.chatFeatures.title": "Функции чата",
   "agentDetail.configure.files.add": "Добавить файл",
+  "agentDetail.configure.files.buildNote.generated": "Сгенерировано",
+  "agentDetail.configure.files.buildNote.richTooltip": "Запись агента о том, что он настроил в режиме Build. Он читает ее в начале каждого разговора вместе с вашим Prompt. <docLink>Подробнее</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Запись агента о том, что он настроил в режиме Build. Он читает ее в начале каждого разговора вместе с вашим Prompt. Подробнее",
   "agentDetail.configure.files.empty.description": "Загрузите документы, которые может прочитать агент, например спецификации, шаблоны или руководства",
   "agentDetail.configure.files.empty.title": "Пока нет файлов",
   "agentDetail.configure.files.label": "Файлы",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Не удалось загрузить предпросмотр.",
   "agentDetail.configure.files.preview.unsupported": "Этот файл не поддерживает предварительный просмотр.",
   "agentDetail.configure.files.remove": "Удалить {{name}}",
-  "agentDetail.configure.files.tip": "Предоставьте агенту любой пример, инструкцию или просто носитель информации, который слишком длинный для промпта. Явно ссылайтесь на него через / в промпте.",
+  "agentDetail.configure.files.tip": "Для справочных материалов, которые слишком длинные для вставки в Prompt: спецификации, шаблоны, инструкции. Ссылайтесь на них через / в Prompt.\n\nВ режиме Build агент может настроить это за вас.",
   "agentDetail.configure.files.toggle": "Переключить файлы",
   "agentDetail.configure.files.treeLabel": "Файлы агента",
   "agentDetail.configure.files.upload.action": "Загрузить",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Удалить {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Источники поиска по знаниям, которые этот агент может использовать при выполнении задач.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Подключите базы знаний, которые агент ищет во время работы, чтобы его ответы опирались на ваш контент. <docLink>Подробнее</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Подключите базы знаний, которые агент ищет во время работы, чтобы его ответы опирались на ваш контент. Подробнее",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Переключить поиск по знаниям",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Настройте модель в фильтрации метаданных",
   "agentDetail.configure.model.label": "Модель",
   "agentDetail.configure.orchestrate": "Оркестрация",
   "agentDetail.configure.preview.chatFeatures": "Функции чата",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Аутентификация конечного пользователя",
   "agentDetail.configure.preview.inputPlaceholder": "Написать {{name}}",
   "agentDetail.configure.preview.restart": "Перезапустить предпросмотр",
+  "agentDetail.configure.preview.sandboxNotice": "Agent работает в песочнице Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "В Dify Community Edition каждый ваш Agent работает в среде песочницы Linux 7.0.0-10060-aws внутри вашего Docker. Изменения среды через Build Chats сохраняются.",
   "agentDetail.configure.preview.title": "Предпросмотр",
   "agentDetail.configure.prompt.copied": "Промпт скопирован",
   "agentDetail.configure.prompt.copy": "Копировать промпт",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Подробнее",
   "agentDetail.configure.rightPanel.modeLabel": "Режим настройки агента",
   "agentDetail.configure.rightPanel.preview": "Предпросмотр",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview доступен только в Dify Cloud и Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview запускает готового агента так, как его увидят пользователи, с чистыми ответами и функциями чата.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Предпросмотр агента",
   "agentDetail.configure.skills.add": "Добавить навык",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Навык",
   "agentDetail.configure.skills.label": "Навыки",
   "agentDetail.configure.skills.remove": "Удалить {{name}}",
-  "agentDetail.configure.skills.richTip": "Оснастите агента модульными возможностями. См. <docLink>документацию</docLink>. Явно ссылайтесь на них через / в промпте.",
-  "agentDetail.configure.skills.tip": "Оснастите агента модульными возможностями. См. документацию. Явно ссылайтесь на них через / в промпте.",
+  "agentDetail.configure.skills.richTip": "Упакуйте инструкции, файлы и скрипты для повторяющейся задачи в Skill. Ссылайтесь на него через / в Prompt. <docLink>Подробнее</docLink>\n\nВ режиме Build агент может настроить это за вас.",
+  "agentDetail.configure.skills.tip": "Упакуйте инструкции, файлы и скрипты для повторяющейся задачи в Skill. Ссылайтесь на него через / в Prompt. Подробнее\n\nВ режиме Build агент может настроить это за вас.",
   "agentDetail.configure.skills.toggle": "Переключить навыки",
   "agentDetail.configure.skills.upload.action": "Загрузить",
   "agentDetail.configure.skills.upload.browse": "обзор",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Дополнительные действия для {{name}}",
   "agentDetail.configure.tools.removeAction": "Удалить {{name}}",
   "agentDetail.configure.tools.removeProvider": "Удалить все инструменты",
-  "agentDetail.configure.tools.richTip": "Оснастите агента плагинами Dify. См. <pluginDocLink>документацию</pluginDocLink>. Явно ссылайтесь на них через / в промпте.\nТакже вы всегда можете попросить агента установить и аутентифицировать другие инструменты (например, MCPs и CLI-инструменты) через Build-чат. Обратите внимание: инструменты, настроенные через Build-чаты, НЕ будут отображаться здесь, хотя и останутся доступны вашему агенту в дальнейшем. См. <buildDocLink>документацию</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Оснастите агента плагинами Dify. См. документацию. Явно ссылайтесь на них через / в промпте. Также вы всегда можете попросить агента установить и аутентифицировать другие инструменты (например, MCPs и CLI-инструменты) через Build-чат. Обратите внимание: инструменты, настроенные через Build-чаты, НЕ будут отображаться здесь, хотя и останутся доступны вашему агенту в дальнейшем. См. документацию.",
+  "agentDetail.configure.tools.richTip": "Добавьте инструменты Dify из своего workspace. Ссылайтесь на них через / в Prompt. <docLink>Подробнее</docLink>\n\nИнструменты, которые агент устанавливает в своей sandbox, не появятся здесь, но останутся ему доступны.",
+  "agentDetail.configure.tools.tip": "Добавьте инструменты Dify из своего workspace. Ссылайтесь на них через / в Prompt. Подробнее\n\nИнструменты, которые агент устанавливает в своей sandbox, не появятся здесь, но останутся ему доступны.",
   "agentDetail.configure.tools.toggle": "Переключить инструменты",
   "agentDetail.configure.tools.toolTabs.all": "Все",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Интеграции",
   "agentDetail.configure.tools.toolTabs.workflow": "Рабочий процесс",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Путь файловой системы",
   "agentDetail.configure.workingDirectory.description": "Файлы, которые агент читает и записывает во время работы. Все, что он создает, например qna_report.pdf, появится здесь.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Файловая система",
+  "agentDetail.configure.workingDirectory.home": "главная",
   "agentDetail.configure.workingDirectory.open": "Открыть рабочий каталог",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Вернуться в рабочий каталог",
   "agentDetail.configure.workingDirectory.title": "Рабочий каталог",
   "agentDetail.configure.workingDirectory.treeLabel": "Файлы рабочего каталога",
+  "agentDetail.configure.workingDirectory.workingDirectory": "рабочий каталог",
   "agentDetail.documentTitle": "Агент",
   "agentDetail.logs.description": "Полные журналы записывают статус работы приложения, включая ввод пользователя, ответы агента, планирование и использование инструментов.",
   "agentDetail.logs.empty": "Журналы не найдены",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Имя обязательно.",
   "roster.createForm.roleLabel": "Роль",
   "roster.createForm.rolePlaceholder": "напр. Помощник по исследованиям",
-  "roster.createForm.roleRequired": "Роль обязательна.",
   "roster.createSuccess": "Агент создан.",
   "roster.dateTimeFormat": "DD.MM.YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Этот агент будет удалён из активного списка Roster.",
@@ -354,6 +372,7 @@
   "roster.empty": "Агентов пока нет",
   "roster.emptySearch": "Подходящих агентов не найдено",
   "roster.filters.all": "Все",
+  "roster.filters.createdByMe": "Созданные мной",
   "roster.filters.drafts": "Черновики",
   "roster.filters.label": "Фильтры агентов",
   "roster.filters.published": "Опубликованные",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Агент сохранён в Roster.",
   "roster.searchLabel": "Поиск агентов",
   "roster.searchPlaceholder": "Поиск агентов по имени…",
+  "roster.sort.earliestCreated": "Сначала созданные",
+  "roster.sort.label": "Сортировать по",
+  "roster.sort.lastModified": "Последнее изменение",
+  "roster.sort.optionsLabel": "Параметры сортировки",
+  "roster.sort.recentlyCreated": "Недавно созданные",
   "roster.tabs.agent": "Агент",
-  "roster.tabs.human": "Человек",
+  "roster.tabs.human": "Контакты",
   "roster.tabsLabel": "Тип Roster",
   "roster.updateSuccess": "Агент обновлён.",
   "roster.usageStatus.draft": "Черновик"
diff --git a/web/i18n/ru-RU/common.json b/web/i18n/ru-RU/common.json
index 80dea8750bb..edb825318d8 100644
--- a/web/i18n/ru-RU/common.json
+++ b/web/i18n/ru-RU/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Установите из Marketplace или настройте поставщиков в настройках.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Показаны только совместимые модели",
   "modelProvider.selector.showIncompatibleModels": "Показать несовместимые модели",
+  "modelProvider.selector.suggestion": "Рекомендация",
+  "modelProvider.selector.suggestionTip": "Рекомендуется для задач агента",
   "modelProvider.showModels": "Показать модели",
   "modelProvider.speechToTextModel.key": "Модель преобразования речи в текст",
   "modelProvider.speechToTextModel.tip": "Установите модель по умолчанию для ввода речи в текст в разговоре.",
diff --git a/web/i18n/ru-RU/workflow.json b/web/i18n/ru-RU/workflow.json
index d40cf6b4dcd..7957ffc3162 100644
--- a/web/i18n/ru-RU/workflow.json
+++ b/web/i18n/ru-RU/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Опубликуйте последний рабочий процесс и убедитесь, что подключен узел ввода пользователя, прежде чем настраивать его как инструмент.",
   "common.workflowAsToolTip": "После обновления рабочего процесса требуется перенастройка инструмента.",
   "common.workflowProcess": "Процесс рабочего процесса",
+  "common.workflowProcessFailed": "Процесс рабочего процесса завершился с ошибкой",
+  "common.workflowProcessPaused": "Процесс рабочего процесса приостановлен",
+  "common.workflowProcessRunning": "Процесс рабочего процесса выполняется",
+  "common.workflowProcessSucceeded": "Процесс рабочего процесса выполнен успешно",
   "customWebhook": "Пользовательский вебхук",
   "debug.copyLastRun": "Копировать последний запуск",
   "debug.copyLastRunError": "Не удалось скопировать последние входные данные выполнения",
diff --git a/web/i18n/sl-SI/agent-v-2.json b/web/i18n/sl-SI/agent-v-2.json
index 593422d42ac..533f2813f30 100644
--- a/web/i18n/sl-SI/agent-v-2.json
+++ b/web/i18n/sl-SI/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Razkrij vrednost {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Vaš agent se izvaja v okolju Linux. Konfiguracije ali skrivnosti lahko vnaprej nastavite tako, da tukaj določite okoljske spremenljivke.\n\nDruga možnost je, da nastavitev vedno izvedete prek klepeta Build. Glejte <docLink>dokumentacijo</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Nastavite vrednosti, ki jih agent bere med izvajanjem, na primer API ključ, ki ga potrebuje orodje, ne da bi jih dali v Prompt.\n\nV načinu Build jih lahko agent nastavi namesto vas.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Izberite obseg okoljske spremenljivke",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Vaš agent se izvaja v okolju Linux. Konfiguracije ali skrivnosti lahko vnaprej nastavite tako, da tukaj določite okoljske spremenljivke. Druga možnost je, da nastavitev vedno izvedete prek klepeta Build. Glejte dokumentacijo.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Nastavite vrednosti, ki jih agent bere med izvajanjem, na primer API ključ, ki ga potrebuje orodje, ne da bi jih dali v Prompt.\n\nV načinu Build jih lahko agent nastavi namesto vas.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Preklopi Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Izdelajte agenta s klepetom",
   "agentDetail.configure.build.inputPlaceholder": "Opišite, kaj naj vaš agent počne",
   "agentDetail.configure.build.startBuild": "Začni gradnjo",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Sprotna opomba te gradnje, ki jo agent uporablja za ohranjanje konteksta med sejami. Ob uporabi se doda med vaše datoteke.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} sprememba za uveljavitev",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} spremembe za uveljavitev",
   "agentDetail.configure.buildDraft.discard": "Zavrzi",
   "agentDetail.configure.buildDraft.modeBadge": "Način gradnje",
   "agentDetail.configure.buildDraft.modeDescription": "Ste v načinu gradnje. Nastavitev oblikujte s klepetom na desni, nato jo uporabite.",
+  "agentDetail.configure.buildDraft.rewritten": "Prepisano",
   "agentDetail.configure.buildDraft.title": "Osnutek gradnje",
+  "agentDetail.configure.buildDraft.updated": "Posodobljeno",
   "agentDetail.configure.chatFeatures.description": "Oblikujte uporabniško izkušnjo klepeta v vaši spletni aplikaciji in klepetalnih površinah.",
   "agentDetail.configure.chatFeatures.title": "Funkcije klepeta",
   "agentDetail.configure.files.add": "Dodaj datoteko",
+  "agentDetail.configure.files.buildNote.generated": "Ustvarjeno",
+  "agentDetail.configure.files.buildNote.richTooltip": "Agentov zapis tega, kar je nastavil v načinu Build. Prebere ga na začetku vsakega pogovora skupaj z vašim Promptom. <docLink>Več informacij</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Agentov zapis tega, kar je nastavil v načinu Build. Prebere ga na začetku vsakega pogovora skupaj z vašim Promptom. Več informacij",
   "agentDetail.configure.files.empty.description": "Naložite dokumente, ki jih lahko agent bere, npr. specifikacije, predloge ali smernice",
   "agentDetail.configure.files.empty.title": "Še ni datotek",
   "agentDetail.configure.files.label": "Datoteke",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Predogleda ni bilo mogoče naložiti.",
   "agentDetail.configure.files.preview.unsupported": "Ta datoteka ne podpira predogleda.",
   "agentDetail.configure.files.remove": "Odstrani {{name}}",
-  "agentDetail.configure.files.tip": "Agentu zagotovite primer, smernice ali preprosto nosilec informacij, ki je predolg za poziv. V pozivu ga izrecno navedite z /.",
+  "agentDetail.configure.files.tip": "Za referenčno gradivo, ki je predolgo za lepljenje v Prompt: specifikacije, predloge, smernice. Sklicujte se z / v Promptu.\n\nV načinu Build jih lahko agent nastavi namesto vas.",
   "agentDetail.configure.files.toggle": "Preklopi datoteke",
   "agentDetail.configure.files.treeLabel": "Datoteke agenta",
   "agentDetail.configure.files.upload.action": "Naloži",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Odstrani {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Viri iskanja znanja, ki jih ta agent lahko uporablja pri izvajanju opravil.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Povežite baze znanja, po katerih agent išče med delom, da njegovi odgovori temeljijo na vaši vsebini. <docLink>Več informacij</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Povežite baze znanja, po katerih agent išče med delom, da njegovi odgovori temeljijo na vaši vsebini. Več informacij",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Preklopi iskanje znanja",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Konfigurirajte model pri filtriranju metapodatkov",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Orkestracija",
   "agentDetail.configure.preview.chatFeatures": "Funkcije klepeta",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Preverjanje pristnosti končnega uporabnika",
   "agentDetail.configure.preview.inputPlaceholder": "Pošlji sporočilo za {{name}}",
   "agentDetail.configure.preview.restart": "Znova zaženi predogled",
+  "agentDetail.configure.preview.sandboxNotice": "Agent se izvaja v peskovniku Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "V Dify Community Edition vsak vaš Agent teče v peskovniškem okolju Linux 7.0.0-10060-aws znotraj vašega Dockerja. Spremembe okolja prek Build Chats so trajne.",
   "agentDetail.configure.preview.title": "Predogled",
   "agentDetail.configure.prompt.copied": "Poziv kopiran",
   "agentDetail.configure.prompt.copy": "Kopiraj poziv",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Več o tem",
   "agentDetail.configure.rightPanel.modeLabel": "Način konfiguracije agenta",
   "agentDetail.configure.rightPanel.preview": "Predogled",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview je na voljo samo v Dify Cloud in Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview zažene dokončanega agenta tako, kot ga bodo videli uporabniki, z jasnimi odgovori in funkcijami klepeta.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Predogled agenta",
   "agentDetail.configure.skills.add": "Dodaj veščino",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Veščina",
   "agentDetail.configure.skills.label": "Veščine",
   "agentDetail.configure.skills.remove": "Odstrani {{name}}",
-  "agentDetail.configure.skills.richTip": "Opremite agenta z modularnimi zmožnostmi. Glejte <docLink>dokumentacijo</docLink>. V pozivu jih izrecno navedite z /.",
-  "agentDetail.configure.skills.tip": "Opremite agenta z modularnimi zmožnostmi. Glejte dokumentacijo. V pozivu jih izrecno navedite z /.",
+  "agentDetail.configure.skills.richTip": "Navodila, datoteke in skripte za ponavljajočo se nalogo zapakirajte v skill. Sklicujte se z / v Promptu. <docLink>Več informacij</docLink>\n\nV načinu Build jih lahko agent nastavi namesto vas.",
+  "agentDetail.configure.skills.tip": "Navodila, datoteke in skripte za ponavljajočo se nalogo zapakirajte v skill. Sklicujte se z / v Promptu. Več informacij\n\nV načinu Build jih lahko agent nastavi namesto vas.",
   "agentDetail.configure.skills.toggle": "Preklopi veščine",
   "agentDetail.configure.skills.upload.action": "Naloži",
   "agentDetail.configure.skills.upload.browse": "prebrskaj",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Več dejanj za {{name}}",
   "agentDetail.configure.tools.removeAction": "Odstrani {{name}}",
   "agentDetail.configure.tools.removeProvider": "Odstrani vsa orodja",
-  "agentDetail.configure.tools.richTip": "Opremite agenta z vtičniki Dify. Glejte <pluginDocLink>dokumentacijo</pluginDocLink>. V pozivu jih izrecno navedite z /.\nLahko pa agentu prek klepeta Build vedno naročite, naj namesti in overi druga orodja (npr. MCPs in orodja CLI). Orodja, konfigurirana prek klepetov Build, se tukaj NE bodo prikazala, vendar bodo agentu še naprej na voljo. Glejte <buildDocLink>dokumentacijo</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Opremite agenta z vtičniki Dify. Glejte dokumentacijo. V pozivu jih izrecno navedite z /. Lahko pa agentu prek klepeta Build vedno naročite, naj namesti in overi druga orodja (npr. MCPs in orodja CLI). Orodja, konfigurirana prek klepetov Build, se tukaj NE bodo prikazala, vendar bodo agentu še naprej na voljo. Glejte dokumentacijo.",
+  "agentDetail.configure.tools.richTip": "Dodajte orodja Dify iz svojega workspacea. Sklicujte se z / v Promptu. <docLink>Več informacij</docLink>\n\nOrodja, ki jih agent namesti v svoj sandbox, se tukaj ne bodo prikazala, vendar mu ostanejo na voljo.",
+  "agentDetail.configure.tools.tip": "Dodajte orodja Dify iz svojega workspacea. Sklicujte se z / v Promptu. Več informacij\n\nOrodja, ki jih agent namesti v svoj sandbox, se tukaj ne bodo prikazala, vendar mu ostanejo na voljo.",
   "agentDetail.configure.tools.toggle": "Preklopi orodja",
   "agentDetail.configure.tools.toolTabs.all": "Vse",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Integracije",
   "agentDetail.configure.tools.toolTabs.workflow": "Potek dela",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Pot datotečnega sistema",
   "agentDetail.configure.workingDirectory.description": "Datoteke, ki jih agent bere in zapisuje med izvajanjem. Vse, kar ustvari, na primer qna_report.pdf, se prikaže tukaj.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Datotečni sistem",
+  "agentDetail.configure.workingDirectory.home": "domov",
   "agentDetail.configure.workingDirectory.open": "Odpri delovni imenik",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Nazaj v delovni imenik",
   "agentDetail.configure.workingDirectory.title": "Delovni imenik",
   "agentDetail.configure.workingDirectory.treeLabel": "Datoteke delovnega imenika",
+  "agentDetail.configure.workingDirectory.workingDirectory": "delovni imenik",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "Polni dnevniki beležijo stanje izvajanja aplikacije, vključno z uporabniškimi vnosi, odgovori agenta, načrtovanjem in uporabo orodij.",
   "agentDetail.logs.empty": "Dnevnikov ni",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Ime je obvezno.",
   "roster.createForm.roleLabel": "Vloga",
   "roster.createForm.rolePlaceholder": "npr. Raziskovalni asistent",
-  "roster.createForm.roleRequired": "Vloga je obvezna.",
   "roster.createSuccess": "Agent ustvarjen.",
   "roster.dateTimeFormat": "DD. MM. YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Ta agent bo odstranjen iz aktivnega seznama Roster.",
@@ -354,6 +372,7 @@
   "roster.empty": "Še ni agentov",
   "roster.emptySearch": "Ni ujemajočih se agentov",
   "roster.filters.all": "Vsi",
+  "roster.filters.createdByMe": "Ustvaril sem",
   "roster.filters.drafts": "Osnutki",
   "roster.filters.label": "Filtri agentov",
   "roster.filters.published": "Objavljeni",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent shranjen v Roster.",
   "roster.searchLabel": "Iskanje agentov",
   "roster.searchPlaceholder": "Iskanje agentov po imenu…",
+  "roster.sort.earliestCreated": "Najprej ustvarjeno",
+  "roster.sort.label": "Razvrsti po",
+  "roster.sort.lastModified": "Nazadnje spremenjeno",
+  "roster.sort.optionsLabel": "Možnosti razvrščanja",
+  "roster.sort.recentlyCreated": "Nedavno ustvarjeno",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "Človek",
+  "roster.tabs.human": "Stiki",
   "roster.tabsLabel": "Tip Roster",
   "roster.updateSuccess": "Agent posodobljen.",
   "roster.usageStatus.draft": "Osnutek"
diff --git a/web/i18n/sl-SI/common.json b/web/i18n/sl-SI/common.json
index cfd0a94b2ec..a51dd2cd160 100644
--- a/web/i18n/sl-SI/common.json
+++ b/web/i18n/sl-SI/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Poiščite v Tržnici za namestitev ali konfigurirajte ponudnike v nastavitvah.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Prikazani so samo združljivi modeli",
   "modelProvider.selector.showIncompatibleModels": "Prikaži nezdružljive modele",
+  "modelProvider.selector.suggestion": "Predlog",
+  "modelProvider.selector.suggestionTip": "Priporočeno za opravila agenta",
   "modelProvider.showModels": "Prikaži modele",
   "modelProvider.speechToTextModel.key": "Model za pretvorbo govora v besedilo",
   "modelProvider.speechToTextModel.tip": "Nastavite privzeti model za vnos govora v besedilo v pogovoru.",
diff --git a/web/i18n/sl-SI/workflow.json b/web/i18n/sl-SI/workflow.json
index 1c31c87d45d..054b88657ac 100644
--- a/web/i18n/sl-SI/workflow.json
+++ b/web/i18n/sl-SI/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Objavite najnovejše delovne tokove in zagotovite, da je pred konfiguracijo kot orodje povezan vozel za vnos uporabnika.",
   "common.workflowAsToolTip": "Zaradi posodobitve delovnega poteka je potrebna ponovna konfiguracija orodja.",
   "common.workflowProcess": "Delovni postopek",
+  "common.workflowProcessFailed": "Delovni postopek ni uspel",
+  "common.workflowProcessPaused": "Delovni postopek je začasno ustavljen",
+  "common.workflowProcessRunning": "Delovni postopek se izvaja",
+  "common.workflowProcessSucceeded": "Delovni postopek je uspel",
   "customWebhook": "Prilagojeni spletni kavelj",
   "debug.copyLastRun": "Kopiraj zadnji zagon",
   "debug.copyLastRunError": "Kopiranje vhodov zadnjega zagona ni uspelo",
diff --git a/web/i18n/th-TH/agent-v-2.json b/web/i18n/th-TH/agent-v-2.json
index ac3d2e0e752..1683bb55939 100644
--- a/web/i18n/th-TH/agent-v-2.json
+++ b/web/i18n/th-TH/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "แสดงค่า {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "เอเจนต์ของคุณทำงานในสภาพแวดล้อม Linux คุณสามารถตั้งค่าคอนฟิกหรือความลับไว้ล่วงหน้าได้โดยกำหนดตัวแปรสภาพแวดล้อมที่นี่\n\nหรือจะให้ตั้งค่าผ่านแชท Build ก็ได้ ดู<docLink>เอกสารประกอบ</docLink>",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "ตั้งค่าที่ agent อ่านตอน runtime เช่น API key ที่เครื่องมือต้องใช้ โดยไม่ต้องใส่ไว้ใน Prompt\n\nในโหมด Build agent สามารถตั้งค่าสิ่งเหล่านี้ให้คุณได้",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "เลือกขอบเขตตัวแปรสภาพแวดล้อม",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "เอเจนต์ของคุณทำงานในสภาพแวดล้อม Linux คุณสามารถตั้งค่าคอนฟิกหรือความลับไว้ล่วงหน้าได้โดยกำหนดตัวแปรสภาพแวดล้อมที่นี่ หรือจะให้ตั้งค่าผ่านแชท Build ก็ได้ ดูเอกสารประกอบ",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "ตั้งค่าที่ agent อ่านตอน runtime เช่น API key ที่เครื่องมือต้องใช้ โดยไม่ต้องใส่ไว้ใน Prompt\n\nในโหมด Build agent สามารถตั้งค่าสิ่งเหล่านี้ให้คุณได้",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "สลับ Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "สร้างเอเจนต์ของคุณด้วยการแชท",
   "agentDetail.configure.build.inputPlaceholder": "อธิบายว่าเอเจนต์ของคุณควรทำอะไร",
   "agentDetail.configure.build.startBuild": "เริ่มสร้าง",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "บันทึกต่อเนื่องของบิลด์นี้ที่เอเจนต์ใช้เพื่อเก็บบริบทข้ามเซสชัน ระบบจะเพิ่มลงในไฟล์ของคุณเมื่อคุณนำไปใช้",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} การเปลี่ยนแปลงที่ต้องนำไปใช้",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} การเปลี่ยนแปลงที่ต้องนำไปใช้",
   "agentDetail.configure.buildDraft.discard": "ทิ้ง",
   "agentDetail.configure.buildDraft.modeBadge": "โหมด Build",
   "agentDetail.configure.buildDraft.modeDescription": "คุณอยู่ในโหมด Build ปรับแต่งการตั้งค่านี้ผ่านแชททางขวา แล้วกดนำไปใช้",
+  "agentDetail.configure.buildDraft.rewritten": "เขียนใหม่แล้ว",
   "agentDetail.configure.buildDraft.title": "ฉบับร่าง Build",
+  "agentDetail.configure.buildDraft.updated": "อัปเดตแล้ว",
   "agentDetail.configure.chatFeatures.description": "กำหนดประสบการณ์การแชทของผู้ใช้ปลายทางบน Web app และหน้าจอแชท",
   "agentDetail.configure.chatFeatures.title": "ฟีเจอร์แชท",
   "agentDetail.configure.files.add": "เพิ่มไฟล์",
+  "agentDetail.configure.files.buildNote.generated": "สร้างแล้ว",
+  "agentDetail.configure.files.buildNote.richTooltip": "บันทึกของ agent เกี่ยวกับสิ่งที่ตั้งค่าไว้ในโหมด Build โดยจะอ่านสิ่งนี้ตอนเริ่มทุกบทสนทนา พร้อมกับ Prompt ของคุณ <docLink>เรียนรู้เพิ่มเติม</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "บันทึกของ agent เกี่ยวกับสิ่งที่ตั้งค่าไว้ในโหมด Build โดยจะอ่านสิ่งนี้ตอนเริ่มทุกบทสนทนา พร้อมกับ Prompt ของคุณ เรียนรู้เพิ่มเติม",
   "agentDetail.configure.files.empty.description": "อัปโหลดเอกสารที่ตัวแทนสามารถอ่านได้ เช่น ข้อกำหนด เทมเพลต หรือแนวทาง",
   "agentDetail.configure.files.empty.title": "ยังไม่มีไฟล์",
   "agentDetail.configure.files.label": "ไฟล์",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "โหลดตัวอย่างไม่สำเร็จ",
   "agentDetail.configure.files.preview.unsupported": "ไฟล์นี้ไม่รองรับการแสดงตัวอย่าง",
   "agentDetail.configure.files.remove": "ลบ {{name}}",
-  "agentDetail.configure.files.tip": "มอบตัวอย่าง แนวทาง หรือข้อมูลที่ยาวเกินกว่าจะใส่ในพรอมป์ให้ตัวแทนของคุณ อ้างอิงอย่างชัดเจนด้วย / ในพรอมป์",
+  "agentDetail.configure.files.tip": "สำหรับเอกสารอ้างอิงที่ยาวเกินกว่าจะวางใน Prompt: สเปก เทมเพลต แนวทาง อ้างอิงด้วย / ใน Prompt\n\nในโหมด Build agent สามารถตั้งค่าสิ่งเหล่านี้ให้คุณได้",
   "agentDetail.configure.files.toggle": "สลับไฟล์",
   "agentDetail.configure.files.treeLabel": "ไฟล์ตัวแทน",
   "agentDetail.configure.files.upload.action": "อัปโหลด",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "ลบ {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "แหล่งดึงความรู้ที่ตัวแทนนี้สามารถใช้ได้ขณะจัดการงาน",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "เชื่อมต่อฐานความรู้ที่ agent ค้นหาระหว่างทำงาน เพื่อให้คำตอบอ้างอิงจากเนื้อหาของคุณ <docLink>เรียนรู้เพิ่มเติม</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "เชื่อมต่อฐานความรู้ที่ agent ค้นหาระหว่างทำงาน เพื่อให้คำตอบอ้างอิงจากเนื้อหาของคุณ เรียนรู้เพิ่มเติม",
   "agentDetail.configure.knowledgeRetrieval.toggle": "สลับการดึงความรู้",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "โปรดกำหนดค่าโมเดลในการกรองเมตาดาต้า",
   "agentDetail.configure.model.label": "โมเดล",
   "agentDetail.configure.orchestrate": "การจัดการ",
   "agentDetail.configure.preview.chatFeatures": "ฟีเจอร์แชท",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "การยืนยันตัวตนของผู้ใช้ปลายทาง",
   "agentDetail.configure.preview.inputPlaceholder": "ส่งข้อความถึง {{name}}",
   "agentDetail.configure.preview.restart": "เริ่มการแสดงตัวอย่างใหม่",
+  "agentDetail.configure.preview.sandboxNotice": "Agent ทำงานในแซนด์บ็อกซ์ Linux",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "สำหรับ Dify Community Edition Agent แต่ละตัวของคุณจะทำงานในสภาพแวดล้อมแซนด์บ็อกซ์ Linux 7.0.0-10060-aws ภายใน Docker ของคุณ การแก้ไขสภาพแวดล้อมผ่าน Build Chats จะคงอยู่ถาวร",
   "agentDetail.configure.preview.title": "แสดงตัวอย่าง",
   "agentDetail.configure.prompt.copied": "คัดลอกพรอมต์แล้ว",
   "agentDetail.configure.prompt.copy": "คัดลอกพรอมต์",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "เรียนรู้เพิ่มเติม",
   "agentDetail.configure.rightPanel.modeLabel": "โหมดการกำหนดค่าเอเจนต์",
   "agentDetail.configure.rightPanel.preview": "แสดงตัวอย่าง",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview พร้อมใช้งานเฉพาะบน Dify Cloud และ Enterprise เท่านั้น",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview เรียกใช้เอเจนต์ที่เสร็จแล้วในแบบที่ผู้ใช้จะเห็น พร้อมคำตอบที่ชัดเจนและฟีเจอร์แชท",
   "agentDetail.configure.rightPanel.previewTipTitle": "ดูตัวอย่างเอเจนต์",
   "agentDetail.configure.skills.add": "เพิ่มทักษะ",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "ทักษะ",
   "agentDetail.configure.skills.label": "ทักษะ",
   "agentDetail.configure.skills.remove": "ลบ {{name}}",
-  "agentDetail.configure.skills.richTip": "เพิ่มความสามารถแบบโมดูลาร์ให้ตัวแทนของคุณ ดู<docLink>เอกสาร</docLink> อ้างอิงอย่างชัดเจนด้วย / ในพรอมป์",
-  "agentDetail.configure.skills.tip": "เพิ่มความสามารถแบบโมดูลาร์ให้ตัวแทนของคุณ ดูเอกสาร อ้างอิงอย่างชัดเจนด้วย / ในพรอมป์",
+  "agentDetail.configure.skills.richTip": "แพ็กเกจคำสั่ง ไฟล์ และสคริปต์สำหรับงานที่ทำซ้ำเป็น Skill อ้างอิงด้วย / ใน Prompt <docLink>เรียนรู้เพิ่มเติม</docLink>\n\nในโหมด Build agent สามารถตั้งค่าสิ่งเหล่านี้ให้คุณได้",
+  "agentDetail.configure.skills.tip": "แพ็กเกจคำสั่ง ไฟล์ และสคริปต์สำหรับงานที่ทำซ้ำเป็น Skill อ้างอิงด้วย / ใน Prompt เรียนรู้เพิ่มเติม\n\nในโหมด Build agent สามารถตั้งค่าสิ่งเหล่านี้ให้คุณได้",
   "agentDetail.configure.skills.toggle": "สลับทักษะ",
   "agentDetail.configure.skills.upload.action": "อัปโหลด",
   "agentDetail.configure.skills.upload.browse": "เรียกดู",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "การกระทำเพิ่มเติมสำหรับ {{name}}",
   "agentDetail.configure.tools.removeAction": "ลบ {{name}}",
   "agentDetail.configure.tools.removeProvider": "ลบเครื่องมือทั้งหมด",
-  "agentDetail.configure.tools.richTip": "เพิ่มปลั๊กอินของ Dify ให้ตัวแทนของคุณ ดู<pluginDocLink>เอกสาร</pluginDocLink> อ้างอิงอย่างชัดเจนด้วย / ในพรอมป์\nอีกทางหนึ่ง คุณสามารถบอกให้ตัวแทนติดตั้งและยืนยันตัวตนเครื่องมืออื่น ๆ (เช่น MCPs และเครื่องมือ CLI) ผ่านแชต Build ได้เสมอ โปรดทราบว่าเครื่องมือที่กำหนดค่าผ่านแชต Build จะไม่แสดงที่นี่ แม้ว่าจะยังพร้อมใช้งานสำหรับตัวแทนของคุณต่อไป ดู<buildDocLink>เอกสาร</buildDocLink>",
-  "agentDetail.configure.tools.tip": "เพิ่มปลั๊กอินของ Dify ให้ตัวแทนของคุณ ดูเอกสาร อ้างอิงอย่างชัดเจนด้วย / ในพรอมป์ อีกทางหนึ่ง คุณสามารถบอกให้ตัวแทนติดตั้งและยืนยันตัวตนเครื่องมืออื่น ๆ (เช่น MCPs และเครื่องมือ CLI) ผ่านแชต Build ได้เสมอ โปรดทราบว่าเครื่องมือที่กำหนดค่าผ่านแชต Build จะไม่แสดงที่นี่ แม้ว่าจะยังพร้อมใช้งานสำหรับตัวแทนของคุณต่อไป ดูเอกสาร",
+  "agentDetail.configure.tools.richTip": "เพิ่มเครื่องมือ Dify จาก workspace ของคุณ อ้างอิงด้วย / ใน Prompt <docLink>เรียนรู้เพิ่มเติม</docLink>\n\nเครื่องมือที่ agent ติดตั้งใน sandbox จะไม่แสดงที่นี่ แต่ยังคงพร้อมใช้งานสำหรับ agent",
+  "agentDetail.configure.tools.tip": "เพิ่มเครื่องมือ Dify จาก workspace ของคุณ อ้างอิงด้วย / ใน Prompt เรียนรู้เพิ่มเติม\n\nเครื่องมือที่ agent ติดตั้งใน sandbox จะไม่แสดงที่นี่ แต่ยังคงพร้อมใช้งานสำหรับ agent",
   "agentDetail.configure.tools.toggle": "สลับเครื่องมือ",
   "agentDetail.configure.tools.toolTabs.all": "ทั้งหมด",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "ปลั๊กอิน",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "เส้นทางระบบไฟล์",
   "agentDetail.configure.workingDirectory.description": "ไฟล์ที่ Agent อ่านและเขียนระหว่างทำงาน สิ่งที่สร้างขึ้น เช่น qna_report.pdf จะแสดงที่นี่",
+  "agentDetail.configure.workingDirectory.fileSystem": "ระบบไฟล์",
+  "agentDetail.configure.workingDirectory.home": "หน้าแรก",
   "agentDetail.configure.workingDirectory.open": "เปิดไดเรกทอรีทำงาน",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "กลับไปยังไดเรกทอรีทำงาน",
   "agentDetail.configure.workingDirectory.title": "ไดเรกทอรีทำงาน",
   "agentDetail.configure.workingDirectory.treeLabel": "ไฟล์ในไดเรกทอรีทำงาน",
+  "agentDetail.configure.workingDirectory.workingDirectory": "ไดเรกทอรีทำงาน",
   "agentDetail.documentTitle": "ตัวแทน",
   "agentDetail.logs.description": "บันทึกแบบเต็มจะบันทึกสถานะการทำงานของแอปพลิเคชัน รวมถึงอินพุตของผู้ใช้ การตอบกลับของตัวแทน การวางแผน และการใช้เครื่องมือ",
   "agentDetail.logs.empty": "ไม่พบบันทึก",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "ต้องระบุชื่อ",
   "roster.createForm.roleLabel": "บทบาท",
   "roster.createForm.rolePlaceholder": "เช่น ผู้ช่วยวิจัย",
-  "roster.createForm.roleRequired": "ต้องระบุบทบาท",
   "roster.createSuccess": "สร้างตัวแทนแล้ว",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "ตัวแทนนี้จะถูกลบออกจากรายการ Roster ที่ใช้งานอยู่",
@@ -354,6 +372,7 @@
   "roster.empty": "ยังไม่มีตัวแทน",
   "roster.emptySearch": "ไม่มีตัวแทนที่ตรงกัน",
   "roster.filters.all": "ทั้งหมด",
+  "roster.filters.createdByMe": "สร้างโดยฉัน",
   "roster.filters.drafts": "ฉบับร่าง",
   "roster.filters.label": "ตัวกรองตัวแทน",
   "roster.filters.published": "เผยแพร่แล้ว",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "บันทึกตัวแทนไปยัง Roster แล้ว",
   "roster.searchLabel": "ค้นหาตัวแทน",
   "roster.searchPlaceholder": "ค้นหาตัวแทนตามชื่อ…",
+  "roster.sort.earliestCreated": "สร้างแรกสุด",
+  "roster.sort.label": "เรียงตาม",
+  "roster.sort.lastModified": "แก้ไขล่าสุด",
+  "roster.sort.optionsLabel": "ตัวเลือกการเรียง",
+  "roster.sort.recentlyCreated": "สร้างล่าสุด",
   "roster.tabs.agent": "ตัวแทน",
-  "roster.tabs.human": "บุคคล",
+  "roster.tabs.human": "ผู้ติดต่อ",
   "roster.tabsLabel": "ประเภท Roster",
   "roster.updateSuccess": "อัปเดตตัวแทนแล้ว",
   "roster.usageStatus.draft": "ฉบับร่าง"
diff --git a/web/i18n/th-TH/common.json b/web/i18n/th-TH/common.json
index b60054ab78d..12c99d7a0b7 100644
--- a/web/i18n/th-TH/common.json
+++ b/web/i18n/th-TH/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "เรียกดู Marketplace เพื่อติดตั้ง หรือกำหนดค่าผู้ให้บริการในการตั้งค่า",
   "modelProvider.selector.onlyCompatibleModelsShown": "แสดงเฉพาะโมเดลที่รองรับเท่านั้น",
   "modelProvider.selector.showIncompatibleModels": "แสดงโมเดลที่ไม่รองรับ",
+  "modelProvider.selector.suggestion": "คำแนะนำ",
+  "modelProvider.selector.suggestionTip": "แนะนำสำหรับงานของเอเจนต์",
   "modelProvider.showModels": "แสดงโมเดล",
   "modelProvider.speechToTextModel.key": "โมเดลคําพูดเป็นข้อความ",
   "modelProvider.speechToTextModel.tip": "ตั้งค่าโมเดลเริ่มต้นสําหรับการป้อนข้อมูลคําพูดเป็นข้อความในการสนทนา",
diff --git a/web/i18n/th-TH/workflow.json b/web/i18n/th-TH/workflow.json
index 947feceff19..bdea9c0414e 100644
--- a/web/i18n/th-TH/workflow.json
+++ b/web/i18n/th-TH/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "เผยแพร่เวิร์กโฟลว์ล่าสุดและตรวจสอบให้แน่ใจว่ามีโหนดป้อนข้อมูลผู้ใช้เชื่อมต่อก่อนที่จะกำหนดค่าเป็นเครื่องมือ",
   "common.workflowAsToolTip": "จําเป็นต้องมีการกําหนดค่าเครื่องมือใหม่หลังจากการอัปเดตเวิร์กโฟลว์",
   "common.workflowProcess": "กระบวนการเวิร์กโฟลว์",
+  "common.workflowProcessFailed": "กระบวนการเวิร์กโฟลว์ล้มเหลว",
+  "common.workflowProcessPaused": "กระบวนการเวิร์กโฟลว์หยุดชั่วคราว",
+  "common.workflowProcessRunning": "กระบวนการเวิร์กโฟลว์กำลังทำงาน",
+  "common.workflowProcessSucceeded": "กระบวนการเวิร์กโฟลว์สำเร็จ",
   "customWebhook": "เว็บฮุกที่กำหนดเอง",
   "debug.copyLastRun": "คัดลอกการทำงานล่าสุด",
   "debug.copyLastRunError": "ไม่สามารถคัดลอกข้อมูลการทำงานครั้งสุดท้ายได้",
diff --git a/web/i18n/tr-TR/agent-v-2.json b/web/i18n/tr-TR/agent-v-2.json
index 7ee0136a1e2..d15a40df8fc 100644
--- a/web/i18n/tr-TR/agent-v-2.json
+++ b/web/i18n/tr-TR/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "{{key}} değerini göster",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Ajanınız bir Linux ortamında çalışır. Burada ortam değişkenleri tanımlayarak yapılandırmaları veya gizli değerleri önceden ayarlayabilirsiniz.\n\nAlternatif olarak, kurulumları her zaman bir Build sohbeti üzerinden yaptırabilirsiniz. <docLink>Dokümanlara</docLink> bakın.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Agent'ın runtime sırasında okuduğu değerleri, örneğin bir aracın ihtiyaç duyduğu API anahtarını, Prompt'a koymadan ayarlayın.\n\nBuild mode'da agent bunları sizin için ayarlayabilir.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Çevre değişkeni kapsamını seç",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Ajanınız bir Linux ortamında çalışır. Burada ortam değişkenleri tanımlayarak yapılandırmaları veya gizli değerleri önceden ayarlayabilirsiniz. Alternatif olarak, kurulumları her zaman bir Build sohbeti üzerinden yaptırabilirsiniz. Dokümanlara bakın.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Agent'ın runtime sırasında okuduğu değerleri, örneğin bir aracın ihtiyaç duyduğu API anahtarını, Prompt'a koymadan ayarlayın.\n\nBuild mode'da agent bunları sizin için ayarlayabilir.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Env Editor'ü değiştir",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Aracınızı sohbet ederek oluşturun",
   "agentDetail.configure.build.inputPlaceholder": "Aracınızın ne yapması gerektiğini açıklayın",
   "agentDetail.configure.build.startBuild": "Build’i başlat",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Aracının oturumlar arasında bağlamı korumak için kullandığı bu derlemeye ait sürekli bir not. Uyguladığınızda dosyalarınıza eklenir.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "Uygulanacak {{count}} değişiklik",
+  "agentDetail.configure.buildDraft.changesToApply_other": "Uygulanacak {{count}} değişiklik",
   "agentDetail.configure.buildDraft.discard": "Vazgeç",
   "agentDetail.configure.buildDraft.modeBadge": "Build modu",
   "agentDetail.configure.buildDraft.modeDescription": "Build modundasınız. Bu kurulumu sağdaki sohbetle şekillendirin, ardından uygulayın.",
+  "agentDetail.configure.buildDraft.rewritten": "Yeniden yazıldı",
   "agentDetail.configure.buildDraft.title": "Build taslağı",
+  "agentDetail.configure.buildDraft.updated": "Güncellendi",
   "agentDetail.configure.chatFeatures.description": "Web app ve sohbet yüzeylerinizde son kullanıcı sohbet deneyimini şekillendirin.",
   "agentDetail.configure.chatFeatures.title": "Sohbet Özellikleri",
   "agentDetail.configure.files.add": "Dosya ekle",
+  "agentDetail.configure.files.buildNote.generated": "Oluşturuldu",
+  "agentDetail.configure.files.buildNote.richTooltip": "Agent'ın Build mode'da kurduğu şeylerin kaydı. Her konuşmanın başında bunu Prompt'unuzla birlikte okur. <docLink>Daha fazla bilgi</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Agent'ın Build mode'da kurduğu şeylerin kaydı. Her konuşmanın başında bunu Prompt'unuzla birlikte okur. Daha fazla bilgi",
   "agentDetail.configure.files.empty.description": "Ajanın okuyabileceği belgeleri yükleyin, örneğin spesifikasyonlar, şablonlar veya yönergeler",
   "agentDetail.configure.files.empty.title": "Henüz dosya yok",
   "agentDetail.configure.files.label": "Dosyalar",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Önizleme yüklenemedi.",
   "agentDetail.configure.files.preview.unsupported": "Bu dosya önizlemeyi desteklemiyor.",
   "agentDetail.configure.files.remove": "{{name}} kaldır",
-  "agentDetail.configure.files.tip": "Aracıya, istem için çok uzun olan herhangi bir örnek, yönerge veya bilgi taşıyıcısı sağlayın. İstemde / ile açıkça referans verin.",
+  "agentDetail.configure.files.tip": "Prompt içine yapıştırmak için fazla uzun olan referans materyalleri için: teknik özellikler, şablonlar, yönergeler. Prompt içinde / ile referans verin.\n\nBuild mode'da agent bunları sizin için ayarlayabilir.",
   "agentDetail.configure.files.toggle": "Dosyaları değiştir",
   "agentDetail.configure.files.treeLabel": "Ajan dosyaları",
   "agentDetail.configure.files.upload.action": "Yükle",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "{{name}} kaldır",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Bu ajanın görevleri koordine ederken kullanabileceği bilgi geri alım kaynakları.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Agent çalışırken arayacağı bilgi tabanlarını bağlayın, böylece yanıtları içeriğinize dayanır. <docLink>Daha fazla bilgi</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Agent çalışırken arayacağı bilgi tabanlarını bağlayın, böylece yanıtları içeriğinize dayanır. Daha fazla bilgi",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Bilgi geri alımını değiştir",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Meta veri filtrelemede bir model yapılandırın",
   "agentDetail.configure.model.label": "Model",
   "agentDetail.configure.orchestrate": "Koordine Et",
   "agentDetail.configure.preview.chatFeatures": "Sohbet Özellikleri",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Son kullanıcı kimlik doğrulaması",
   "agentDetail.configure.preview.inputPlaceholder": "{{name}} ile mesajlaş",
   "agentDetail.configure.preview.restart": "Önizlemeyi yeniden başlat",
+  "agentDetail.configure.preview.sandboxNotice": "Agent bir Linux korumalı alanında çalışır.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Dify Community Edition’da, agent’larınızın her biri Docker’ınız içinde Linux 7.0.0-10060-aws korumalı alan ortamında çalışır. Build Chats üzerinden ortamda yaptığınız düzenlemeler kalıcıdır.",
   "agentDetail.configure.preview.title": "Önizleme",
   "agentDetail.configure.prompt.copied": "İstem kopyalandı",
   "agentDetail.configure.prompt.copy": "İstemi kopyala",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Daha fazla bilgi",
   "agentDetail.configure.rightPanel.modeLabel": "Aracı yapılandırma modu",
   "agentDetail.configure.rightPanel.preview": "Önizleme",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview yalnızca Dify Cloud ve Enterprise üzerinde kullanılabilir.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview, tamamlanan ajanı kullanıcılarınızın göreceği şekilde, temiz yanıtlar ve sohbet özellikleriyle çalıştırır.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Ajanınızı önizleyin",
   "agentDetail.configure.skills.add": "Beceri ekle",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Beceri",
   "agentDetail.configure.skills.label": "Beceriler",
   "agentDetail.configure.skills.remove": "{{name}} kaldır",
-  "agentDetail.configure.skills.richTip": "Aracınızı modüler yeteneklerle donatın. <docLink>Belgelere</docLink> bakın. İstemde / ile açıkça referans verin.",
-  "agentDetail.configure.skills.tip": "Aracınızı modüler yeteneklerle donatın. Belgelere bakın. İstemde / ile açıkça referans verin.",
+  "agentDetail.configure.skills.richTip": "Tekrarlanan bir görev için gereken talimatları, dosyaları ve betikleri bir Skill olarak paketleyin. Prompt içinde / ile referans verin. <docLink>Daha fazla bilgi</docLink>\n\nBuild mode'da agent bunları sizin için ayarlayabilir.",
+  "agentDetail.configure.skills.tip": "Tekrarlanan bir görev için gereken talimatları, dosyaları ve betikleri bir Skill olarak paketleyin. Prompt içinde / ile referans verin. Daha fazla bilgi\n\nBuild mode'da agent bunları sizin için ayarlayabilir.",
   "agentDetail.configure.skills.toggle": "Becerileri değiştir",
   "agentDetail.configure.skills.upload.action": "Yükle",
   "agentDetail.configure.skills.upload.browse": "göz at",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "{{name}} için daha fazla eylem",
   "agentDetail.configure.tools.removeAction": "{{name}} kaldır",
   "agentDetail.configure.tools.removeProvider": "Tüm araçları kaldır",
-  "agentDetail.configure.tools.richTip": "Aracınızı Dify eklentileriyle donatın. <pluginDocLink>Belgelere</pluginDocLink> bakın. İstemde / ile açıkça referans verin.\nAlternatif olarak, aracınıza bir Build sohbeti üzerinden başka araçları (örn. MCPs ve CLI araçları) yüklemesini ve kimlik doğrulamasını yapmasını her zaman söyleyebilirsiniz. Build sohbetleriyle yapılandırılan araçların burada görünmeyeceğini, ancak ileride aracınız için kullanılabilir olmaya devam edeceğini unutmayın. <buildDocLink>Belgelere</buildDocLink> bakın.",
-  "agentDetail.configure.tools.tip": "Aracınızı Dify eklentileriyle donatın. Belgelere bakın. İstemde / ile açıkça referans verin. Alternatif olarak, aracınıza bir Build sohbeti üzerinden başka araçları (örn. MCPs ve CLI araçları) yüklemesini ve kimlik doğrulamasını yapmasını her zaman söyleyebilirsiniz. Build sohbetleriyle yapılandırılan araçların burada görünmeyeceğini, ancak ileride aracınız için kullanılabilir olmaya devam edeceğini unutmayın. Belgelere bakın.",
+  "agentDetail.configure.tools.richTip": "Workspace'inizden Dify araçları ekleyin. Prompt içinde / ile referans verin. <docLink>Daha fazla bilgi</docLink>\n\nAgent'ın sandbox içinde yüklediği araçlar burada görünmez, ancak kullanıma açık kalır.",
+  "agentDetail.configure.tools.tip": "Workspace'inizden Dify araçları ekleyin. Prompt içinde / ile referans verin. Daha fazla bilgi\n\nAgent'ın sandbox içinde yüklediği araçlar burada görünmez, ancak kullanıma açık kalır.",
   "agentDetail.configure.tools.toggle": "Araçları değiştir",
   "agentDetail.configure.tools.toolTabs.all": "Tümü",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Eklentiler",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Dosya sistemi yolu",
   "agentDetail.configure.workingDirectory.description": "Agent çalışırken okuduğu ve yazdığı dosyalar. qna_report.pdf gibi ürettiği her şey burada görünür.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Dosya sistemi",
+  "agentDetail.configure.workingDirectory.home": "ana sayfa",
   "agentDetail.configure.workingDirectory.open": "Çalışma dizinini aç",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Çalışma dizinine dön",
   "agentDetail.configure.workingDirectory.title": "Çalışma dizini",
   "agentDetail.configure.workingDirectory.treeLabel": "Çalışma dizini dosyaları",
+  "agentDetail.configure.workingDirectory.workingDirectory": "çalışma dizini",
   "agentDetail.documentTitle": "Ajan",
   "agentDetail.logs.description": "Tam günlükler, kullanıcı girişleri, ajan yanıtları, planlama ve araç kullanımları dahil olmak üzere uygulamanın çalışma durumunu kaydeder.",
   "agentDetail.logs.empty": "Günlük bulunamadı",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Ad gereklidir.",
   "roster.createForm.roleLabel": "Rol",
   "roster.createForm.rolePlaceholder": "örn. Araştırma Asistanı",
-  "roster.createForm.roleRequired": "Rol gereklidir.",
   "roster.createSuccess": "Ajan oluşturuldu.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Bu ajan, aktif Roster listesinden kaldırılacak.",
@@ -354,6 +372,7 @@
   "roster.empty": "Henüz ajan yok",
   "roster.emptySearch": "Eşleşen ajan yok",
   "roster.filters.all": "Tümü",
+  "roster.filters.createdByMe": "Benim oluşturduklarım",
   "roster.filters.drafts": "Taslaklar",
   "roster.filters.label": "Ajan filtreleri",
   "roster.filters.published": "Yayınlandı",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Ajan Roster'a kaydedildi.",
   "roster.searchLabel": "Ajanları ara",
   "roster.searchPlaceholder": "Ajanları ada göre ara…",
+  "roster.sort.earliestCreated": "İlk oluşturulanlar",
+  "roster.sort.label": "Sıralama ölçütü",
+  "roster.sort.lastModified": "Son değiştirilme",
+  "roster.sort.optionsLabel": "Sıralama seçenekleri",
+  "roster.sort.recentlyCreated": "Yeni oluşturulanlar",
   "roster.tabs.agent": "Ajan",
-  "roster.tabs.human": "İnsan",
+  "roster.tabs.human": "Kişiler",
   "roster.tabsLabel": "Roster türü",
   "roster.updateSuccess": "Ajan güncellendi.",
   "roster.usageStatus.draft": "Taslak"
diff --git a/web/i18n/tr-TR/common.json b/web/i18n/tr-TR/common.json
index 0bcefc17f4c..f661f3fd0fd 100644
--- a/web/i18n/tr-TR/common.json
+++ b/web/i18n/tr-TR/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Yüklemek için Pazar Yeri'ne göz atın veya ayarlardan sağlayıcıları yapılandırın.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Yalnızca uyumlu modeller gösterilir",
   "modelProvider.selector.showIncompatibleModels": "Uyumsuz modelleri göster",
+  "modelProvider.selector.suggestion": "Öneri",
+  "modelProvider.selector.suggestionTip": "Aracı görevleri için önerilir",
   "modelProvider.showModels": "Modelleri Göster",
   "modelProvider.speechToTextModel.key": "Konuşmadan Metne Modeli",
   "modelProvider.speechToTextModel.tip": "Konuşmada konuşmadan metne giriş için varsayılan modeli ayarlayın.",
diff --git a/web/i18n/tr-TR/workflow.json b/web/i18n/tr-TR/workflow.json
index c9bb1d563d2..7580b32613c 100644
--- a/web/i18n/tr-TR/workflow.json
+++ b/web/i18n/tr-TR/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "En son iş akışını yayınlayın ve bunu bir araç olarak yapılandırmadan önce bağlı bir Kullanıcı Girdisi düğümünün olduğundan emin olun.",
   "common.workflowAsToolTip": "İş Akışı güncellemesinden sonra araç yeniden yapılandırması gereklidir.",
   "common.workflowProcess": "İş Akışı Süreci",
+  "common.workflowProcessFailed": "İş Akışı Süreci başarısız oldu",
+  "common.workflowProcessPaused": "İş Akışı Süreci duraklatıldı",
+  "common.workflowProcessRunning": "İş Akışı Süreci çalışıyor",
+  "common.workflowProcessSucceeded": "İş Akışı Süreci başarılı oldu",
   "customWebhook": "Özel Webhook",
   "debug.copyLastRun": "Son Çalışmayı Kopyala",
   "debug.copyLastRunError": "Son çalışma girdilerini kopyalamak başarısız oldu.",
diff --git a/web/i18n/uk-UA/agent-v-2.json b/web/i18n/uk-UA/agent-v-2.json
index b7c1e35782b..717923d30a2 100644
--- a/web/i18n/uk-UA/agent-v-2.json
+++ b/web/i18n/uk-UA/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Показати значення {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Ваш агент працює в середовищі Linux. Ви можете попередньо задати конфігурації або секрети, визначивши тут змінні середовища.\n\nТакож ви завжди можете налаштувати все через чат Build. Див. <docLink>документацію</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Задайте значення, які агент читає під час виконання, наприклад API-ключ, потрібний інструменту, не додаючи їх у Prompt.\n\nУ режимі Build агент може налаштувати це за вас.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Виберіть область змінної середовища",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Ваш агент працює в середовищі Linux. Ви можете попередньо задати конфігурації або секрети, визначивши тут змінні середовища. Також ви завжди можете налаштувати все через чат Build. Див. документацію.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Задайте значення, які агент читає під час виконання, наприклад API-ключ, потрібний інструменту, не додаючи їх у Prompt.\n\nУ режимі Build агент може налаштувати це за вас.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Перемкнути Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Створіть агента через чат",
   "agentDetail.configure.build.inputPlaceholder": "Опишіть, що має робити ваш агент",
   "agentDetail.configure.build.startBuild": "Почати збірку",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Поточна нотатка цієї збірки, яку агент використовує, щоб зберігати контекст між сеансами. Вона додається до ваших файлів після застосування.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} зміна для застосування",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} змін для застосування",
   "agentDetail.configure.buildDraft.discard": "Відхилити",
   "agentDetail.configure.buildDraft.modeBadge": "Режим збірки",
   "agentDetail.configure.buildDraft.modeDescription": "Ви в режимі збірки. Налаштуйте цю конфігурацію через чат праворуч, а потім застосуйте.",
+  "agentDetail.configure.buildDraft.rewritten": "Переписано",
   "agentDetail.configure.buildDraft.title": "Чернетка збірки",
+  "agentDetail.configure.buildDraft.updated": "Оновлено",
   "agentDetail.configure.chatFeatures.description": "Налаштуйте чат-досвід кінцевого користувача у вашому веб-застосунку та чат-поверхнях.",
   "agentDetail.configure.chatFeatures.title": "Функції чату",
   "agentDetail.configure.files.add": "Додати файл",
+  "agentDetail.configure.files.buildNote.generated": "Згенеровано",
+  "agentDetail.configure.files.buildNote.richTooltip": "Запис агента про те, що він налаштував у режимі Build. Він читає його на початку кожної розмови разом із вашим Prompt. <docLink>Докладніше</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Запис агента про те, що він налаштував у режимі Build. Він читає його на початку кожної розмови разом із вашим Prompt. Докладніше",
   "agentDetail.configure.files.empty.description": "Завантажте документи, які може читати агент, наприклад специфікації, шаблони чи інструкції",
   "agentDetail.configure.files.empty.title": "Файлів ще немає",
   "agentDetail.configure.files.label": "Файли",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Не вдалося завантажити перегляд.",
   "agentDetail.configure.files.preview.unsupported": "Попередній перегляд цього файлу не підтримується.",
   "agentDetail.configure.files.remove": "Видалити {{name}}",
-  "agentDetail.configure.files.tip": "Надайте агенту будь-який приклад, інструкцію або просто носій інформації, який занадто довгий для промпта. Явно посилайтеся на нього через / у промпті.",
+  "agentDetail.configure.files.tip": "Для довідкових матеріалів, які занадто довгі для вставлення в Prompt: специфікації, шаблони, інструкції. Посилайтеся на них через / у Prompt.\n\nУ режимі Build агент може налаштувати це за вас.",
   "agentDetail.configure.files.toggle": "Перемкнути файли",
   "agentDetail.configure.files.treeLabel": "Файли агента",
   "agentDetail.configure.files.upload.action": "Завантажити",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Видалити {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Джерела пошуку по знаннях, які цей агент може використовувати під час оркестрації задач.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Підключіть бази знань, які агент шукає під час роботи, щоб його відповіді спиралися на ваш контент. <docLink>Докладніше</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Підключіть бази знань, які агент шукає під час роботи, щоб його відповіді спиралися на ваш контент. Докладніше",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Перемкнути пошук по знаннях",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Налаштуйте модель у фільтрації метаданих",
   "agentDetail.configure.model.label": "Модель",
   "agentDetail.configure.orchestrate": "Оркестрація",
   "agentDetail.configure.preview.chatFeatures": "Функції чату",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Автентифікація кінцевого користувача",
   "agentDetail.configure.preview.inputPlaceholder": "Надіслати повідомлення {{name}}",
   "agentDetail.configure.preview.restart": "Перезапустити перегляд",
+  "agentDetail.configure.preview.sandboxNotice": "Agent працює в пісочниці Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "У Dify Community Edition кожен ваш Agent працює в середовищі пісочниці Linux 7.0.0-10060-aws у вашому Docker. Зміни середовища через Build Chats зберігаються.",
   "agentDetail.configure.preview.title": "Перегляд",
   "agentDetail.configure.prompt.copied": "Промпт скопійовано",
   "agentDetail.configure.prompt.copy": "Скопіювати промпт",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Дізнатися більше",
   "agentDetail.configure.rightPanel.modeLabel": "Режим налаштування агента",
   "agentDetail.configure.rightPanel.preview": "Перегляд",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview доступний лише в Dify Cloud та Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview запускає готового агента так, як його побачать користувачі, з чистими відповідями та функціями чату.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Попередній перегляд агента",
   "agentDetail.configure.skills.add": "Додати навичку",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Навичка",
   "agentDetail.configure.skills.label": "Навички",
   "agentDetail.configure.skills.remove": "Видалити {{name}}",
-  "agentDetail.configure.skills.richTip": "Оснастіть агента модульними можливостями. Див. <docLink>документацію</docLink>. Явно посилайтеся на них через / у промпті.",
-  "agentDetail.configure.skills.tip": "Оснастіть агента модульними можливостями. Див. документацію. Явно посилайтеся на них через / у промпті.",
+  "agentDetail.configure.skills.richTip": "Запакуйте інструкції, файли та скрипти для повторюваного завдання в Skill. Посилайтеся на нього через / у Prompt. <docLink>Докладніше</docLink>\n\nУ режимі Build агент може налаштувати це за вас.",
+  "agentDetail.configure.skills.tip": "Запакуйте інструкції, файли та скрипти для повторюваного завдання в Skill. Посилайтеся на нього через / у Prompt. Докладніше\n\nУ режимі Build агент може налаштувати це за вас.",
   "agentDetail.configure.skills.toggle": "Перемкнути навички",
   "agentDetail.configure.skills.upload.action": "Завантажити",
   "agentDetail.configure.skills.upload.browse": "огляд",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Більше дій для {{name}}",
   "agentDetail.configure.tools.removeAction": "Видалити {{name}}",
   "agentDetail.configure.tools.removeProvider": "Видалити всі інструменти",
-  "agentDetail.configure.tools.richTip": "Оснастіть агента плагінами Dify. Див. <pluginDocLink>документацію</pluginDocLink>. Явно посилайтеся на них через / у промпті.\nКрім того, ви завжди можете сказати агенту встановити й автентифікувати інші інструменти (наприклад, MCPs та CLI-інструменти) через Build-чат. Зверніть увагу: інструменти, налаштовані через Build-чати, НЕ відображатимуться тут, хоча й надалі будуть доступні вашому агенту. Див. <buildDocLink>документацію</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Оснастіть агента плагінами Dify. Див. документацію. Явно посилайтеся на них через / у промпті. Крім того, ви завжди можете сказати агенту встановити й автентифікувати інші інструменти (наприклад, MCPs та CLI-інструменти) через Build-чат. Зверніть увагу: інструменти, налаштовані через Build-чати, НЕ відображатимуться тут, хоча й надалі будуть доступні вашому агенту. Див. документацію.",
+  "agentDetail.configure.tools.richTip": "Додайте інструменти Dify зі свого workspace. Посилайтеся на них через / у Prompt. <docLink>Докладніше</docLink>\n\nІнструменти, які агент встановлює у своїй sandbox, не з'являться тут, але залишаться йому доступними.",
+  "agentDetail.configure.tools.tip": "Додайте інструменти Dify зі свого workspace. Посилайтеся на них через / у Prompt. Докладніше\n\nІнструменти, які агент встановлює у своїй sandbox, не з'являться тут, але залишаться йому доступними.",
   "agentDetail.configure.tools.toggle": "Перемкнути інструменти",
   "agentDetail.configure.tools.toolTabs.all": "Усі",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Інтеграції",
   "agentDetail.configure.tools.toolTabs.workflow": "Робочий процес",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Шлях файлової системи",
   "agentDetail.configure.workingDirectory.description": "Файли, які агент читає і записує під час роботи. Усе, що він створює, наприклад qna_report.pdf, з’являється тут.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Файлова система",
+  "agentDetail.configure.workingDirectory.home": "головна",
   "agentDetail.configure.workingDirectory.open": "Відкрити робочий каталог",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Повернутися до робочого каталогу",
   "agentDetail.configure.workingDirectory.title": "Робочий каталог",
   "agentDetail.configure.workingDirectory.treeLabel": "Файли робочого каталогу",
+  "agentDetail.configure.workingDirectory.workingDirectory": "робочий каталог",
   "agentDetail.documentTitle": "Агент",
   "agentDetail.logs.description": "Повні журнали записують стан роботи застосунка, включаючи введення користувача, відповіді агента, планування та використання інструментів.",
   "agentDetail.logs.empty": "Журнали не знайдено",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Ім'я обов'язкове.",
   "roster.createForm.roleLabel": "Роль",
   "roster.createForm.rolePlaceholder": "напр. Помічник з досліджень",
-  "roster.createForm.roleRequired": "Роль обов'язкова.",
   "roster.createSuccess": "Агента створено.",
   "roster.dateTimeFormat": "DD.MM.YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Цього агента буде видалено з активного списку Roster.",
@@ -354,6 +372,7 @@
   "roster.empty": "Поки що немає агентів",
   "roster.emptySearch": "Немає відповідних агентів",
   "roster.filters.all": "Усі",
+  "roster.filters.createdByMe": "Створені мною",
   "roster.filters.drafts": "Чернетки",
   "roster.filters.label": "Фільтри агентів",
   "roster.filters.published": "Опубліковані",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Агента збережено в Roster.",
   "roster.searchLabel": "Пошук агентів",
   "roster.searchPlaceholder": "Пошук агентів за ім'ям…",
+  "roster.sort.earliestCreated": "Спочатку створені",
+  "roster.sort.label": "Сортувати за",
+  "roster.sort.lastModified": "Остання зміна",
+  "roster.sort.optionsLabel": "Параметри сортування",
+  "roster.sort.recentlyCreated": "Нещодавно створені",
   "roster.tabs.agent": "Агент",
-  "roster.tabs.human": "Людина",
+  "roster.tabs.human": "Контакти",
   "roster.tabsLabel": "Тип Roster",
   "roster.updateSuccess": "Агента оновлено.",
   "roster.usageStatus.draft": "Чернетка"
diff --git a/web/i18n/uk-UA/common.json b/web/i18n/uk-UA/common.json
index f0277a7468e..19db8af712e 100644
--- a/web/i18n/uk-UA/common.json
+++ b/web/i18n/uk-UA/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Перегляньте Marketplace для встановлення або налаштуйте постачальників у параметрах.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Показано лише сумісні моделі",
   "modelProvider.selector.showIncompatibleModels": "Показати несумісні моделі",
+  "modelProvider.selector.suggestion": "Рекомендація",
+  "modelProvider.selector.suggestionTip": "Рекомендовано для завдань агента",
   "modelProvider.showModels": "Показати моделі",
   "modelProvider.speechToTextModel.key": "Модель перетворення мовлення в текст",
   "modelProvider.speechToTextModel.tip": "Встановіть модель за замовчуванням для введення мовлення в текст під час розмови.",
diff --git a/web/i18n/uk-UA/workflow.json b/web/i18n/uk-UA/workflow.json
index 6d96671e5ab..d927cd9689d 100644
--- a/web/i18n/uk-UA/workflow.json
+++ b/web/i18n/uk-UA/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Опублікуйте останній робочий процес і переконайтеся, що перед налаштуванням його як інструменту є підключений вузол введення користувача.",
   "common.workflowAsToolTip": "Після оновлення робочого потоку необхідна переконфігурація інструменту.",
   "common.workflowProcess": "Процес робочого потоку",
+  "common.workflowProcessFailed": "Процес робочого потоку завершився помилкою",
+  "common.workflowProcessPaused": "Процес робочого потоку призупинено",
+  "common.workflowProcessRunning": "Процес робочого потоку виконується",
+  "common.workflowProcessSucceeded": "Процес робочого потоку успішно завершено",
   "customWebhook": "Користувацький вебхук",
   "debug.copyLastRun": "Копіювати останній запуск",
   "debug.copyLastRunError": "Не вдалося скопіювати вхідні дані останнього виконання",
diff --git a/web/i18n/vi-VN/agent-v-2.json b/web/i18n/vi-VN/agent-v-2.json
index 2f1769fbd91..8a15e581f67 100644
--- a/web/i18n/vi-VN/agent-v-2.json
+++ b/web/i18n/vi-VN/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "Hiện giá trị của {{key}}",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "Agent của bạn chạy trong môi trường Linux. Bạn có thể thiết lập sẵn cấu hình hoặc bí mật bằng cách định nghĩa biến môi trường tại đây.\n\nNgoài ra, bạn luôn có thể thiết lập thông qua chat Build. Xem <docLink>tài liệu</docLink>.",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "Đặt các giá trị mà tác nhân đọc lúc runtime, như API key mà một công cụ cần, mà không đưa chúng vào Prompt.\n\nỞ chế độ Build, tác nhân có thể thiết lập những mục này cho bạn.",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "Chọn phạm vi biến môi trường",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "Agent của bạn chạy trong môi trường Linux. Bạn có thể thiết lập sẵn cấu hình hoặc bí mật bằng cách định nghĩa biến môi trường tại đây. Ngoài ra, bạn luôn có thể thiết lập thông qua chat Build. Xem tài liệu.",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "Đặt các giá trị mà tác nhân đọc lúc runtime, như API key mà một công cụ cần, mà không đưa chúng vào Prompt.\n\nỞ chế độ Build, tác nhân có thể thiết lập những mục này cho bạn.",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "Bật/tắt Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "Xây dựng tác nhân bằng trò chuyện",
   "agentDetail.configure.build.inputPlaceholder": "Mô tả tác nhân của bạn nên làm gì",
   "agentDetail.configure.build.startBuild": "Bắt đầu build",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "Ghi chú liên tục của bản dựng này mà agent dùng để giữ ngữ cảnh giữa các phiên. Ghi chú sẽ được thêm vào tệp của bạn khi áp dụng.",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} thay đổi để áp dụng",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} thay đổi để áp dụng",
   "agentDetail.configure.buildDraft.discard": "Hủy bỏ",
   "agentDetail.configure.buildDraft.modeBadge": "Chế độ build",
   "agentDetail.configure.buildDraft.modeDescription": "Bạn đang ở chế độ build. Điều chỉnh thiết lập này qua khung chat bên phải, rồi Áp dụng.",
+  "agentDetail.configure.buildDraft.rewritten": "Đã viết lại",
   "agentDetail.configure.buildDraft.title": "Bản nháp build",
+  "agentDetail.configure.buildDraft.updated": "Đã cập nhật",
   "agentDetail.configure.chatFeatures.description": "Định hình trải nghiệm trò chuyện cho người dùng cuối trên Web app và các bề mặt trò chuyện.",
   "agentDetail.configure.chatFeatures.title": "Tính năng trò chuyện",
   "agentDetail.configure.files.add": "Thêm tệp",
+  "agentDetail.configure.files.buildNote.generated": "Đã tạo",
+  "agentDetail.configure.files.buildNote.richTooltip": "Bản ghi của tác nhân về những gì nó đã thiết lập trong chế độ Build. Nó đọc bản ghi này ở đầu mỗi cuộc trò chuyện, cùng với Prompt của bạn. <docLink>Tìm hiểu thêm</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Bản ghi của tác nhân về những gì nó đã thiết lập trong chế độ Build. Nó đọc bản ghi này ở đầu mỗi cuộc trò chuyện, cùng với Prompt của bạn. Tìm hiểu thêm",
   "agentDetail.configure.files.empty.description": "Tải lên tài liệu mà tác nhân có thể đọc, như đặc tả, mẫu hoặc hướng dẫn",
   "agentDetail.configure.files.empty.title": "Chưa có tệp nào",
   "agentDetail.configure.files.label": "Tệp",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "Tải xem trước thất bại.",
   "agentDetail.configure.files.preview.unsupported": "Tệp này không hỗ trợ xem trước.",
   "agentDetail.configure.files.remove": "Xóa {{name}}",
-  "agentDetail.configure.files.tip": "Cung cấp cho tác nhân của bạn bất kỳ ví dụ, hướng dẫn hoặc đơn giản là một nguồn thông tin quá dài để đưa vào prompt. Tham chiếu rõ ràng bằng / trong Prompt.",
+  "agentDetail.configure.files.tip": "Dành cho tài liệu tham khảo quá dài để dán vào Prompt: thông số, mẫu, hướng dẫn. Tham chiếu bằng / trong Prompt.\n\nỞ chế độ Build, tác nhân có thể thiết lập những mục này cho bạn.",
   "agentDetail.configure.files.toggle": "Bật/tắt tệp",
   "agentDetail.configure.files.treeLabel": "Tệp tác nhân",
   "agentDetail.configure.files.upload.action": "Tải lên",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "Xóa {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "Các nguồn truy xuất kiến thức mà tác nhân này có thể sử dụng khi điều phối tác vụ.",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "Kết nối các kho kiến thức mà tác nhân tìm kiếm khi làm việc, để câu trả lời dựa trên nội dung của bạn. <docLink>Tìm hiểu thêm</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "Kết nối các kho kiến thức mà tác nhân tìm kiếm khi làm việc, để câu trả lời dựa trên nội dung của bạn. Tìm hiểu thêm",
   "agentDetail.configure.knowledgeRetrieval.toggle": "Bật/tắt truy xuất kiến thức",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "Vui lòng cấu hình một mô hình trong bộ lọc siêu dữ liệu",
   "agentDetail.configure.model.label": "Mô hình",
   "agentDetail.configure.orchestrate": "Điều phối",
   "agentDetail.configure.preview.chatFeatures": "Tính năng trò chuyện",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "Xác thực người dùng cuối",
   "agentDetail.configure.preview.inputPlaceholder": "Nhắn cho {{name}}",
   "agentDetail.configure.preview.restart": "Khởi động lại bản xem trước",
+  "agentDetail.configure.preview.sandboxNotice": "Agent chạy trong sandbox Linux.",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "Trong Dify Community Edition, mỗi Agent của bạn chạy trong môi trường sandbox Linux 7.0.0-10060-aws bên trong Docker của bạn. Các chỉnh sửa môi trường qua Build Chats sẽ được lưu giữ.",
   "agentDetail.configure.preview.title": "Xem trước",
   "agentDetail.configure.prompt.copied": "Đã sao chép lời nhắc",
   "agentDetail.configure.prompt.copy": "Sao chép lời nhắc",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "Tìm hiểu thêm",
   "agentDetail.configure.rightPanel.modeLabel": "Chế độ cấu hình tác nhân",
   "agentDetail.configure.rightPanel.preview": "Xem trước",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "Preview chỉ khả dụng trên Dify Cloud và Enterprise.",
   "agentDetail.configure.rightPanel.previewTipBody": "Preview chạy agent hoàn chỉnh theo cách người dùng sẽ thấy, với câu trả lời rõ ràng và các tính năng chat.",
   "agentDetail.configure.rightPanel.previewTipTitle": "Xem trước agent",
   "agentDetail.configure.skills.add": "Thêm kỹ năng",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Kỹ năng",
   "agentDetail.configure.skills.label": "Kỹ năng",
   "agentDetail.configure.skills.remove": "Xóa {{name}}",
-  "agentDetail.configure.skills.richTip": "Trang bị cho tác nhân các năng lực dạng mô-đun. Tham khảo <docLink>tài liệu</docLink>. Tham chiếu rõ ràng bằng / trong Prompt.",
-  "agentDetail.configure.skills.tip": "Trang bị cho tác nhân các năng lực dạng mô-đun. Tham khảo tài liệu. Tham chiếu rõ ràng bằng / trong Prompt.",
+  "agentDetail.configure.skills.richTip": "Đóng gói hướng dẫn, tệp và script cho một tác vụ lặp lại thành skill. Tham chiếu bằng / trong Prompt. <docLink>Tìm hiểu thêm</docLink>\n\nỞ chế độ Build, tác nhân có thể thiết lập những mục này cho bạn.",
+  "agentDetail.configure.skills.tip": "Đóng gói hướng dẫn, tệp và script cho một tác vụ lặp lại thành skill. Tham chiếu bằng / trong Prompt. Tìm hiểu thêm\n\nỞ chế độ Build, tác nhân có thể thiết lập những mục này cho bạn.",
   "agentDetail.configure.skills.toggle": "Bật/tắt kỹ năng",
   "agentDetail.configure.skills.upload.action": "Tải lên",
   "agentDetail.configure.skills.upload.browse": "duyệt",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "Thêm hành động cho {{name}}",
   "agentDetail.configure.tools.removeAction": "Xóa {{name}}",
   "agentDetail.configure.tools.removeProvider": "Xóa tất cả công cụ",
-  "agentDetail.configure.tools.richTip": "Trang bị cho tác nhân các plugin của Dify. Tham khảo <pluginDocLink>tài liệu</pluginDocLink>. Tham chiếu rõ ràng bằng / trong Prompt.\nNgoài ra, bạn luôn có thể yêu cầu tác nhân cài đặt và xác thực các công cụ khác (ví dụ: MCPs và công cụ CLI) thông qua Build chat. Lưu ý rằng các công cụ được cấu hình qua Build chats sẽ KHÔNG hiển thị ở đây, mặc dù chúng vẫn sẽ khả dụng cho tác nhân của bạn về sau. Tham khảo <buildDocLink>tài liệu</buildDocLink>.",
-  "agentDetail.configure.tools.tip": "Trang bị cho tác nhân các plugin của Dify. Tham khảo tài liệu. Tham chiếu rõ ràng bằng / trong Prompt. Ngoài ra, bạn luôn có thể yêu cầu tác nhân cài đặt và xác thực các công cụ khác (ví dụ: MCPs và công cụ CLI) thông qua Build chat. Lưu ý rằng các công cụ được cấu hình qua Build chats sẽ KHÔNG hiển thị ở đây, mặc dù chúng vẫn sẽ khả dụng cho tác nhân của bạn về sau. Tham khảo tài liệu.",
+  "agentDetail.configure.tools.richTip": "Thêm công cụ Dify từ workspace của bạn. Tham chiếu bằng / trong Prompt. <docLink>Tìm hiểu thêm</docLink>\n\nNhững công cụ tác nhân cài đặt trong sandbox sẽ không xuất hiện ở đây nhưng vẫn khả dụng cho tác nhân.",
+  "agentDetail.configure.tools.tip": "Thêm công cụ Dify từ workspace của bạn. Tham chiếu bằng / trong Prompt. Tìm hiểu thêm\n\nNhững công cụ tác nhân cài đặt trong sandbox sẽ không xuất hiện ở đây nhưng vẫn khả dụng cho tác nhân.",
   "agentDetail.configure.tools.toggle": "Bật/tắt công cụ",
   "agentDetail.configure.tools.toolTabs.all": "Tất cả",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "Plugin",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "Đường dẫn hệ thống tệp",
   "agentDetail.configure.workingDirectory.description": "Các tệp mà agent đọc và ghi khi chạy. Mọi thứ nó tạo ra, như qna_report.pdf, sẽ xuất hiện ở đây.",
+  "agentDetail.configure.workingDirectory.fileSystem": "Hệ thống tệp",
+  "agentDetail.configure.workingDirectory.home": "trang chủ",
   "agentDetail.configure.workingDirectory.open": "Mở thư mục làm việc",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "Quay lại thư mục làm việc",
   "agentDetail.configure.workingDirectory.title": "Thư mục làm việc",
   "agentDetail.configure.workingDirectory.treeLabel": "Tệp trong thư mục làm việc",
+  "agentDetail.configure.workingDirectory.workingDirectory": "thư mục làm việc",
   "agentDetail.documentTitle": "Tác nhân",
   "agentDetail.logs.description": "Nhật ký đầy đủ ghi lại trạng thái chạy của ứng dụng, bao gồm đầu vào của người dùng, phản hồi của tác nhân, lập kế hoạch và sử dụng công cụ.",
   "agentDetail.logs.empty": "Không tìm thấy nhật ký",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "Cần có tên.",
   "roster.createForm.roleLabel": "Vai trò",
   "roster.createForm.rolePlaceholder": "ví dụ: Trợ lý nghiên cứu",
-  "roster.createForm.roleRequired": "Cần có vai trò.",
   "roster.createSuccess": "Đã tạo tác nhân.",
   "roster.dateTimeFormat": "DD/MM/YYYY HH:mm:ss",
   "roster.deleteDialog.description": "Tác nhân này sẽ bị xóa khỏi danh sách Roster đang hoạt động.",
@@ -354,6 +372,7 @@
   "roster.empty": "Chưa có tác nhân nào",
   "roster.emptySearch": "Không có tác nhân phù hợp",
   "roster.filters.all": "Tất cả",
+  "roster.filters.createdByMe": "Do tôi tạo",
   "roster.filters.drafts": "Bản nháp",
   "roster.filters.label": "Bộ lọc tác nhân",
   "roster.filters.published": "Đã xuất bản",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Đã lưu tác nhân vào Roster.",
   "roster.searchLabel": "Tìm kiếm tác nhân",
   "roster.searchPlaceholder": "Tìm kiếm tác nhân theo tên…",
+  "roster.sort.earliestCreated": "Tạo sớm nhất",
+  "roster.sort.label": "Sắp xếp theo",
+  "roster.sort.lastModified": "Sửa đổi gần nhất",
+  "roster.sort.optionsLabel": "Tùy chọn sắp xếp",
+  "roster.sort.recentlyCreated": "Mới tạo gần đây",
   "roster.tabs.agent": "Tác nhân",
-  "roster.tabs.human": "Con người",
+  "roster.tabs.human": "Liên hệ",
   "roster.tabsLabel": "Loại Roster",
   "roster.updateSuccess": "Đã cập nhật tác nhân.",
   "roster.usageStatus.draft": "Bản nháp"
diff --git a/web/i18n/vi-VN/common.json b/web/i18n/vi-VN/common.json
index f51c833b26c..f705c057dcc 100644
--- a/web/i18n/vi-VN/common.json
+++ b/web/i18n/vi-VN/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "Duyệt Marketplace để cài đặt hoặc cấu hình nhà cung cấp trong phần cài đặt.",
   "modelProvider.selector.onlyCompatibleModelsShown": "Chỉ hiển thị các mô hình tương thích",
   "modelProvider.selector.showIncompatibleModels": "Hiển thị mô hình không tương thích",
+  "modelProvider.selector.suggestion": "Gợi ý",
+  "modelProvider.selector.suggestionTip": "Được đề xuất cho tác vụ tác nhân",
   "modelProvider.showModels": "Hiện Mô hình",
   "modelProvider.speechToTextModel.key": "Mô hình Chuyển đổi Văn bản thành Tiếng nói",
   "modelProvider.speechToTextModel.tip": "Thiết lập mô hình mặc định cho đầu vào chuyển đổi tiếng nói thành văn bản trong cuộc trò chuyện.",
diff --git a/web/i18n/vi-VN/workflow.json b/web/i18n/vi-VN/workflow.json
index de4623135ae..45e8b81e2fd 100644
--- a/web/i18n/vi-VN/workflow.json
+++ b/web/i18n/vi-VN/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "Xuất bản quy trình làm việc mới nhất và đảm bảo một nút Nhập liệu Người dùng kết nối trước khi cấu hình nó như một công cụ.",
   "common.workflowAsToolTip": "Cần cấu hình lại công cụ sau khi cập nhật quy trình làm việc.",
   "common.workflowProcess": "Quy trình làm việc",
+  "common.workflowProcessFailed": "Quy trình làm việc thất bại",
+  "common.workflowProcessPaused": "Quy trình làm việc đã tạm dừng",
+  "common.workflowProcessRunning": "Quy trình làm việc đang chạy",
+  "common.workflowProcessSucceeded": "Quy trình làm việc thành công",
   "customWebhook": "Webhook Tùy Chỉnh",
   "debug.copyLastRun": "Sao chép lần chạy cuối",
   "debug.copyLastRunError": "Không thể sao chép đầu vào của lần chạy trước",
diff --git a/web/i18n/zh-Hans/agent-v-2.json b/web/i18n/zh-Hans/agent-v-2.json
index ed1071f2b3b..bff704ebff3 100644
--- a/web/i18n/zh-Hans/agent-v-2.json
+++ b/web/i18n/zh-Hans/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "显示 {{key}} 的值",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "你的 Agent 运行在 Linux 环境中。你可以在这里通过定义环境变量预设配置或密钥。\n\n或者，也可以随时通过 Build 对话完成设置。参考 <docLink>文档</docLink>。",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "设置 Agent 运行时读取的值，例如工具所需的 API 密钥，而无需写进提示词。\n\n在构建模式中，Agent 可帮你完成这些设置。",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "选择环境变量作用域",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "你的 Agent 运行在 Linux 环境中。你可以在这里通过定义环境变量预设配置或密钥。或者，也可以随时通过 Build 对话完成设置。参考文档。",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "设置 Agent 运行时读取的值，例如工具所需的 API 密钥，而无需写进提示词。\n\n在构建模式中，Agent 可帮你完成这些设置。",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "展开或收起 Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "通过对话构建 Agent",
   "agentDetail.configure.build.inputPlaceholder": "描述你的 Agent 应该做什么",
   "agentDetail.configure.build.startBuild": "开始构建",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "这是一份此构建的持续记录，Agent 会用它在多次会话之间保留上下文。应用后，它会添加到你的文件中。",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} 项变更待应用",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} 项变更待应用",
   "agentDetail.configure.buildDraft.discard": "放弃",
   "agentDetail.configure.buildDraft.modeBadge": "构建模式",
   "agentDetail.configure.buildDraft.modeDescription": "你正在使用构建模式。通过右侧聊天调整此配置，然后应用。",
+  "agentDetail.configure.buildDraft.rewritten": "已重写",
   "agentDetail.configure.buildDraft.title": "Build 草稿",
+  "agentDetail.configure.buildDraft.updated": "已更新",
   "agentDetail.configure.chatFeatures.description": "配置 Web app 和聊天界面的终端用户聊天体验。",
   "agentDetail.configure.chatFeatures.title": "Chat 功能",
   "agentDetail.configure.files.add": "添加文件",
+  "agentDetail.configure.files.buildNote.generated": "已生成",
+  "agentDetail.configure.files.buildNote.richTooltip": "Agent 在构建模式中完成的设置都记录在这里。每次对话开始时，它会连同提示词一起读取这份记录。<docLink>了解更多</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Agent 在构建模式中完成的设置都记录在这里。每次对话开始时，它会连同提示词一起读取这份记录。了解更多",
   "agentDetail.configure.files.empty.description": "上传 Agent 可读取的文档，例如规格、模板或指南",
   "agentDetail.configure.files.empty.title": "暂无文件",
   "agentDetail.configure.files.label": "文件",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "预览加载失败。",
   "agentDetail.configure.files.preview.unsupported": "该文件不支持预览。",
   "agentDetail.configure.files.remove": "移除 {{name}}",
-  "agentDetail.configure.files.tip": "为 Agent 提供示例、指南，或任何过长而不适合放入提示词的信息载体。请在提示词中使用 / 显式引用。",
+  "agentDetail.configure.files.tip": "上传过长、不适合放入提示词的参考资料：规格、模板、指南。在提示词中用 / 引用。\n\n在构建模式中，Agent 可帮你完成这些设置。",
   "agentDetail.configure.files.toggle": "展开或收起文件",
   "agentDetail.configure.files.treeLabel": "Agent 文件",
   "agentDetail.configure.files.upload.action": "上传",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "删除 {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "此 Agent 在编排任务时可使用的知识检索来源。",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "连接 Agent 工作时可检索的知识库，以基于其内容生成回复。<docLink>了解更多</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "连接 Agent 工作时可检索的知识库，以基于其内容生成回复。了解更多",
   "agentDetail.configure.knowledgeRetrieval.toggle": "展开或收起知识检索",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "请在元数据过滤中配置模型",
   "agentDetail.configure.model.label": "模型",
   "agentDetail.configure.orchestrate": "编排",
   "agentDetail.configure.preview.chatFeatures": "Chat 功能",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "终端用户认证",
   "agentDetail.configure.preview.inputPlaceholder": "向 {{name}} 发送消息",
   "agentDetail.configure.preview.restart": "重新开始预览",
+  "agentDetail.configure.preview.sandboxNotice": "Agent 运行在 Linux 沙盒中。",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "在 Dify Community Edition 中，每个 Agent 都运行在你 Docker 内的 Linux 7.0.0-10060-aws 沙盒环境中。你通过 Build Chats 对环境所做的编辑会持久保留。",
   "agentDetail.configure.preview.title": "预览",
   "agentDetail.configure.prompt.copied": "提示词已复制",
   "agentDetail.configure.prompt.copy": "复制提示词",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "了解更多",
   "agentDetail.configure.rightPanel.modeLabel": "Agent 配置模式",
   "agentDetail.configure.rightPanel.preview": "预览",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "预览仅在 Dify Cloud 和 Dify Enterprise 中可用。",
   "agentDetail.configure.rightPanel.previewTipBody": "像用户一样测试它。你的交互不会影响 Agent 后续的行为。",
   "agentDetail.configure.rightPanel.previewTipTitle": "预览你的 Agent",
   "agentDetail.configure.skills.add": "添加 Skill",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Skill",
   "agentDetail.configure.skills.label": "Skill",
   "agentDetail.configure.skills.remove": "移除 {{name}}",
-  "agentDetail.configure.skills.richTip": "为 Agent 配备模块化能力。请参考<docLink>文档</docLink>。请在提示词中使用 / 显式引用。",
-  "agentDetail.configure.skills.tip": "为 Agent 配备模块化能力。请参考文档。请在提示词中使用 / 显式引用。",
+  "agentDetail.configure.skills.richTip": "将重复任务所需的指令、文件和脚本打包成 Skill。在提示词中用 / 引用。<docLink>了解更多</docLink>\n\n在构建模式中，Agent 可帮你完成这些设置。",
+  "agentDetail.configure.skills.tip": "将重复任务所需的指令、文件和脚本打包成 Skill。在提示词中用 / 引用。了解更多\n\n在构建模式中，Agent 可帮你完成这些设置。",
   "agentDetail.configure.skills.toggle": "展开或收起 Skill",
   "agentDetail.configure.skills.upload.action": "上传",
   "agentDetail.configure.skills.upload.browse": "选择",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "{{name}} 的更多操作",
   "agentDetail.configure.tools.removeAction": "移除 {{name}}",
   "agentDetail.configure.tools.removeProvider": "移除所有工具",
-  "agentDetail.configure.tools.richTip": "为 Agent 配备 Dify 插件。请参考<pluginDocLink>文档</pluginDocLink>。请在提示词中使用 / 显式引用。\n或者，你也可以随时通过 Build 对话让 Agent 安装并认证其他工具（例如 MCP 和 CLI 工具）。注意，通过 Build 对话配置的工具不会显示在这里，但后续仍可供你的 Agent 使用。请参考<buildDocLink>文档</buildDocLink>。",
-  "agentDetail.configure.tools.tip": "为 Agent 配备 Dify 插件。请参考文档。请在提示词中使用 / 显式引用。或者，你也可以随时通过 Build 对话让 Agent 安装并认证其他工具（例如 MCP 和 CLI 工具）。注意，通过 Build 对话配置的工具不会显示在这里，但后续仍可供你的 Agent 使用。请参考文档。",
+  "agentDetail.configure.tools.richTip": "添加工作空间中的 Dify 工具。在提示词中用 / 引用。<docLink>了解更多</docLink>\n\nAgent 在沙箱中自行安装的工具不会显示在这里，但仍然可用。",
+  "agentDetail.configure.tools.tip": "添加工作空间中的 Dify 工具。在提示词中用 / 引用。了解更多\n\nAgent 在沙箱中自行安装的工具不会显示在这里，但仍然可用。",
   "agentDetail.configure.tools.toggle": "展开或收起工具",
   "agentDetail.configure.tools.toolTabs.all": "全部",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "插件",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "文件系统路径",
   "agentDetail.configure.workingDirectory.description": "Agent 运行时读取和写入的文件。它生成的任何内容，例如 qna_report.pdf，都会显示在这里。",
+  "agentDetail.configure.workingDirectory.fileSystem": "文件系统",
+  "agentDetail.configure.workingDirectory.home": "home",
   "agentDetail.configure.workingDirectory.open": "打开工作目录",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "返回工作目录",
   "agentDetail.configure.workingDirectory.title": "工作目录",
   "agentDetail.configure.workingDirectory.treeLabel": "工作目录文件",
+  "agentDetail.configure.workingDirectory.workingDirectory": "工作目录",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "完整日志记录应用运行状态，包括用户输入、Agent 回复、规划和工具使用。",
   "agentDetail.logs.empty": "暂无日志",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "请输入名称。",
   "roster.createForm.roleLabel": "角色",
   "roster.createForm.rolePlaceholder": "例如：研究助理",
-  "roster.createForm.roleRequired": "请输入角色。",
   "roster.createSuccess": "Agent 已创建。",
   "roster.dateTimeFormat": "YYYY-MM-DD HH:mm:ss",
   "roster.deleteDialog.description": "此 Agent 将从当前活跃 Roster 列表中移除。",
@@ -354,6 +372,7 @@
   "roster.empty": "暂无 Agent",
   "roster.emptySearch": "没有匹配的 Agent",
   "roster.filters.all": "全部",
+  "roster.filters.createdByMe": "我创建的",
   "roster.filters.drafts": "草稿",
   "roster.filters.label": "Agent 筛选",
   "roster.filters.published": "已发布",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent 已保存到 Roster。",
   "roster.searchLabel": "搜索 Agent",
   "roster.searchPlaceholder": "按名称搜索 Agent…",
+  "roster.sort.earliestCreated": "最早创建",
+  "roster.sort.label": "排序方式",
+  "roster.sort.lastModified": "最后修改",
+  "roster.sort.optionsLabel": "排序选项",
+  "roster.sort.recentlyCreated": "最近创建",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "人工成员",
+  "roster.tabs.human": "联系人",
   "roster.tabsLabel": "Roster 类型",
   "roster.updateSuccess": "Agent 已更新。",
   "roster.usageStatus.draft": "草稿"
diff --git a/web/i18n/zh-Hans/common.json b/web/i18n/zh-Hans/common.json
index f4f479b0a2b..83e309caa57 100644
--- a/web/i18n/zh-Hans/common.json
+++ b/web/i18n/zh-Hans/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "前往 Marketplace 安装，或在设置中配置提供商。",
   "modelProvider.selector.onlyCompatibleModelsShown": "仅显示兼容的模型",
   "modelProvider.selector.showIncompatibleModels": "显示不兼容模型",
+  "modelProvider.selector.suggestion": "推荐",
+  "modelProvider.selector.suggestionTip": "推荐用于 Agent 任务",
   "modelProvider.showModels": "显示模型",
   "modelProvider.speechToTextModel.key": "语音转文本模型",
   "modelProvider.speechToTextModel.tip": "设置对话中语音转文字输入的默认使用模型。",
diff --git a/web/i18n/zh-Hans/workflow.json b/web/i18n/zh-Hans/workflow.json
index 828f6480570..ef4b3ca3f12 100644
--- a/web/i18n/zh-Hans/workflow.json
+++ b/web/i18n/zh-Hans/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "请先发布最新的工作流，并确保已连接的 User Input 节点后再配置为工具。",
   "common.workflowAsToolTip": "工作流更新后需要重新配置工具参数",
   "common.workflowProcess": "工作流",
+  "common.workflowProcessFailed": "工作流运行失败",
+  "common.workflowProcessPaused": "工作流运行已暂停",
+  "common.workflowProcessRunning": "工作流运行中",
+  "common.workflowProcessSucceeded": "工作流运行成功",
   "customWebhook": "自定义 Webhook",
   "debug.copyLastRun": "复制上次运行值",
   "debug.copyLastRunError": "复制上次运行输入失败",
diff --git a/web/i18n/zh-Hant/agent-v-2.json b/web/i18n/zh-Hant/agent-v-2.json
index 18bfc8016ad..4c0daade372 100644
--- a/web/i18n/zh-Hant/agent-v-2.json
+++ b/web/i18n/zh-Hant/agent-v-2.json
@@ -51,12 +51,13 @@
   "agentDetail.configure.advancedSettings.envEditor.keyPlaceholder": "Key",
   "agentDetail.configure.advancedSettings.envEditor.label": "Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.revealValue": "顯示 {{key}} 的值",
-  "agentDetail.configure.advancedSettings.envEditor.richTip": "你的 Agent 執行於 Linux 環境中。你可以在這裡透過定義環境變數預設設定或密鑰。\n\n或者，也可以隨時透過 Build 對話完成設定。參考 <docLink>文件</docLink>。",
+  "agentDetail.configure.advancedSettings.envEditor.richTip": "設定 Agent 執行時讀取的值，例如工具所需的 API 金鑰，而無需寫進提示詞。\n\n在建置模式中，Agent 可幫你完成這些設定。",
   "agentDetail.configure.advancedSettings.envEditor.scopeColumn": "Scope",
   "agentDetail.configure.advancedSettings.envEditor.scopePlain": "Plain",
   "agentDetail.configure.advancedSettings.envEditor.scopeSecret": "Secret",
   "agentDetail.configure.advancedSettings.envEditor.scopeSelector": "選擇環境變數作用域",
-  "agentDetail.configure.advancedSettings.envEditor.tip": "你的 Agent 執行於 Linux 環境中。你可以在這裡透過定義環境變數預設設定或密鑰。或者，也可以隨時透過 Build 對話完成設定。參考文件。",
+  "agentDetail.configure.advancedSettings.envEditor.shortLabel": "Env",
+  "agentDetail.configure.advancedSettings.envEditor.tip": "設定 Agent 執行時讀取的值，例如工具所需的 API 金鑰，而無需寫進提示詞。\n\n在建置模式中，Agent 可幫你完成這些設定。",
   "agentDetail.configure.advancedSettings.envEditor.toggle": "展開或收合 Env Editor",
   "agentDetail.configure.advancedSettings.envEditor.valueColumn": "Value",
   "agentDetail.configure.advancedSettings.envEditor.valuePlaceholder": "Value",
@@ -66,13 +67,21 @@
   "agentDetail.configure.build.empty.title": "透過對話建置 Agent",
   "agentDetail.configure.build.inputPlaceholder": "描述你的 Agent 應該做什麼",
   "agentDetail.configure.build.startBuild": "開始構建",
+  "agentDetail.configure.buildDraft.buildNoteDescription": "這是一份此構建的持續記錄，Agent 會用它在多次工作階段之間保留上下文。套用後，它會新增到你的檔案中。",
+  "agentDetail.configure.buildDraft.changesToApply_one": "{{count}} 項變更待套用",
+  "agentDetail.configure.buildDraft.changesToApply_other": "{{count}} 項變更待套用",
   "agentDetail.configure.buildDraft.discard": "放棄",
   "agentDetail.configure.buildDraft.modeBadge": "建置模式",
   "agentDetail.configure.buildDraft.modeDescription": "你正在使用建置模式。透過右側聊天調整此設定，然後套用。",
+  "agentDetail.configure.buildDraft.rewritten": "已重寫",
   "agentDetail.configure.buildDraft.title": "Build 草稿",
+  "agentDetail.configure.buildDraft.updated": "已更新",
   "agentDetail.configure.chatFeatures.description": "配置 Web app 和聊天介面的終端使用者聊天體驗。",
   "agentDetail.configure.chatFeatures.title": "Chat 功能",
   "agentDetail.configure.files.add": "新增檔案",
+  "agentDetail.configure.files.buildNote.generated": "已生成",
+  "agentDetail.configure.files.buildNote.richTooltip": "Agent 在建置模式中完成的設定都記錄在這裡。每次對話開始時，它會連同提示詞一起讀取這份記錄。<docLink>了解更多</docLink>",
+  "agentDetail.configure.files.buildNote.tooltip": "Agent 在建置模式中完成的設定都記錄在這裡。每次對話開始時，它會連同提示詞一起讀取這份記錄。了解更多",
   "agentDetail.configure.files.empty.description": "上傳 Agent 可讀取的文件，例如規格、範本或指南",
   "agentDetail.configure.files.empty.title": "暫無檔案",
   "agentDetail.configure.files.label": "檔案",
@@ -80,7 +89,7 @@
   "agentDetail.configure.files.preview.failed": "預覽載入失敗。",
   "agentDetail.configure.files.preview.unsupported": "此檔案不支援預覽。",
   "agentDetail.configure.files.remove": "移除 {{name}}",
-  "agentDetail.configure.files.tip": "為 Agent 提供範例、指南，或任何過長而不適合放入提示詞的資訊載體。請在提示詞中使用 / 明確引用。",
+  "agentDetail.configure.files.tip": "上傳過長、不適合放入提示詞的參考資料：規格、範本、指南。在提示詞中用 / 引用。\n\n在建置模式中，Agent 可幫你完成這些設定。",
   "agentDetail.configure.files.toggle": "展開或收合檔案",
   "agentDetail.configure.files.treeLabel": "Agent 檔案",
   "agentDetail.configure.files.upload.action": "上傳",
@@ -121,8 +130,10 @@
   "agentDetail.configure.knowledgeRetrieval.remove": "刪除 {{name}}",
   "agentDetail.configure.knowledgeRetrieval.retrievalOne": "Retrieval 1",
   "agentDetail.configure.knowledgeRetrieval.retrievalTwo": "Retrieval 2",
-  "agentDetail.configure.knowledgeRetrieval.tip": "此 Agent 在編排任務時可使用的知識檢索來源。",
+  "agentDetail.configure.knowledgeRetrieval.richTip": "連接 Agent 工作時可檢索的知識庫，以基於其內容生成回覆。<docLink>了解更多</docLink>",
+  "agentDetail.configure.knowledgeRetrieval.tip": "連接 Agent 工作時可檢索的知識庫，以基於其內容生成回覆。了解更多",
   "agentDetail.configure.knowledgeRetrieval.toggle": "展開或收合知識檢索",
+  "agentDetail.configure.knowledgeRetrieval.validation.metadataModelRequired": "請在中繼資料篩選中設定模型",
   "agentDetail.configure.model.label": "模型",
   "agentDetail.configure.orchestrate": "編排",
   "agentDetail.configure.preview.chatFeatures": "Chat 功能",
@@ -133,6 +144,8 @@
   "agentDetail.configure.preview.endUserAuth": "終端使用者驗證",
   "agentDetail.configure.preview.inputPlaceholder": "傳訊息給 {{name}}",
   "agentDetail.configure.preview.restart": "重新開始預覽",
+  "agentDetail.configure.preview.sandboxNotice": "Agent 執行於 Linux 沙盒中。",
+  "agentDetail.configure.preview.sandboxNoticeTooltip": "在 Dify Community Edition 中，每個 Agent 都執行於你 Docker 內的 Linux 7.0.0-10060-aws 沙盒環境。你透過 Build Chats 對環境所做的編輯會持久保留。",
   "agentDetail.configure.preview.title": "預覽",
   "agentDetail.configure.prompt.copied": "提示詞已複製",
   "agentDetail.configure.prompt.copy": "複製提示詞",
@@ -171,6 +184,7 @@
   "agentDetail.configure.rightPanel.learnMore": "了解更多",
   "agentDetail.configure.rightPanel.modeLabel": "Agent 配置模式",
   "agentDetail.configure.rightPanel.preview": "預覽",
+  "agentDetail.configure.rightPanel.previewDisabledTip": "預覽僅在 Dify Cloud 和 Dify Enterprise 中可用。",
   "agentDetail.configure.rightPanel.previewTipBody": "像使用者一樣測試它。你的互動不會影響 Agent 後續的行為。",
   "agentDetail.configure.rightPanel.previewTipTitle": "預覽你的 Agent",
   "agentDetail.configure.skills.add": "新增 Skill",
@@ -182,8 +196,8 @@
   "agentDetail.configure.skills.itemType": "Skill",
   "agentDetail.configure.skills.label": "Skill",
   "agentDetail.configure.skills.remove": "移除 {{name}}",
-  "agentDetail.configure.skills.richTip": "為 Agent 配備模組化能力。請參考<docLink>文件</docLink>。請在提示詞中使用 / 明確引用。",
-  "agentDetail.configure.skills.tip": "為 Agent 配備模組化能力。請參考文件。請在提示詞中使用 / 明確引用。",
+  "agentDetail.configure.skills.richTip": "將重複任務所需的指令、檔案和指令碼打包成 Skill。在提示詞中用 / 引用。<docLink>了解更多</docLink>\n\n在建置模式中，Agent 可幫你完成這些設定。",
+  "agentDetail.configure.skills.tip": "將重複任務所需的指令、檔案和指令碼打包成 Skill。在提示詞中用 / 引用。了解更多\n\n在建置模式中，Agent 可幫你完成這些設定。",
   "agentDetail.configure.skills.toggle": "展開或收合 Skill",
   "agentDetail.configure.skills.upload.action": "上傳",
   "agentDetail.configure.skills.upload.browse": "選擇",
@@ -225,8 +239,8 @@
   "agentDetail.configure.tools.moreActions": "{{name}} 的更多操作",
   "agentDetail.configure.tools.removeAction": "移除 {{name}}",
   "agentDetail.configure.tools.removeProvider": "移除所有工具",
-  "agentDetail.configure.tools.richTip": "為 Agent 配備 Dify 外掛。請參考<pluginDocLink>文件</pluginDocLink>。請在提示詞中使用 / 明確引用。\n或者，你也可以隨時透過 Build 對話讓 Agent 安裝並驗證其他工具（例如 MCP 和 CLI 工具）。注意，透過 Build 對話設定的工具不會顯示在這裡，但後續仍可供你的 Agent 使用。請參考<buildDocLink>文件</buildDocLink>。",
-  "agentDetail.configure.tools.tip": "為 Agent 配備 Dify 外掛。請參考文件。請在提示詞中使用 / 明確引用。或者，你也可以隨時透過 Build 對話讓 Agent 安裝並驗證其他工具（例如 MCP 和 CLI 工具）。注意，透過 Build 對話設定的工具不會顯示在這裡，但後續仍可供你的 Agent 使用。請參考文件。",
+  "agentDetail.configure.tools.richTip": "新增工作區中的 Dify 工具。在提示詞中用 / 引用。<docLink>了解更多</docLink>\n\nAgent 在沙箱中自行安裝的工具不會顯示在這裡，但仍然可用。",
+  "agentDetail.configure.tools.tip": "新增工作區中的 Dify 工具。在提示詞中用 / 引用。了解更多\n\nAgent 在沙箱中自行安裝的工具不會顯示在這裡，但仍然可用。",
   "agentDetail.configure.tools.toggle": "展開或收合工具",
   "agentDetail.configure.tools.toolTabs.all": "全部",
   "agentDetail.configure.tools.toolTabs.cli": "CLI",
@@ -234,10 +248,15 @@
   "agentDetail.configure.tools.toolTabs.mcp": "MCP",
   "agentDetail.configure.tools.toolTabs.plugins": "外掛",
   "agentDetail.configure.tools.toolTabs.workflow": "Workflow",
+  "agentDetail.configure.workingDirectory.breadcrumbLabel": "檔案系統路徑",
   "agentDetail.configure.workingDirectory.description": "Agent 執行時讀取和寫入的檔案。它產生的任何內容，例如 qna_report.pdf，都會顯示在這裡。",
+  "agentDetail.configure.workingDirectory.fileSystem": "檔案系統",
+  "agentDetail.configure.workingDirectory.home": "home",
   "agentDetail.configure.workingDirectory.open": "開啟工作目錄",
+  "agentDetail.configure.workingDirectory.returnToWorkspace": "返回工作目錄",
   "agentDetail.configure.workingDirectory.title": "工作目錄",
   "agentDetail.configure.workingDirectory.treeLabel": "工作目錄檔案",
+  "agentDetail.configure.workingDirectory.workingDirectory": "工作目錄",
   "agentDetail.documentTitle": "Agent",
   "agentDetail.logs.description": "完整日誌記錄應用程式執行狀態，包括使用者輸入、Agent 回覆、規劃和工具使用。",
   "agentDetail.logs.empty": "暫無日誌",
@@ -336,7 +355,6 @@
   "roster.createForm.nameRequired": "請輸入名稱。",
   "roster.createForm.roleLabel": "角色",
   "roster.createForm.rolePlaceholder": "例如：研究助理",
-  "roster.createForm.roleRequired": "請輸入角色。",
   "roster.createSuccess": "Agent 已建立。",
   "roster.dateTimeFormat": "YYYY-MM-DD HH:mm:ss",
   "roster.deleteDialog.description": "此 Agent 將從目前活躍 Roster 清單中移除。",
@@ -354,6 +372,7 @@
   "roster.empty": "暫無 Agent",
   "roster.emptySearch": "沒有符合的 Agent",
   "roster.filters.all": "全部",
+  "roster.filters.createdByMe": "我建立的",
   "roster.filters.drafts": "草稿",
   "roster.filters.label": "Agent 篩選",
   "roster.filters.published": "已發佈",
@@ -376,8 +395,13 @@
   "roster.saveToRosterSuccess": "Agent 已儲存到 Roster。",
   "roster.searchLabel": "搜尋 Agent",
   "roster.searchPlaceholder": "依名稱搜尋 Agent…",
+  "roster.sort.earliestCreated": "最早建立",
+  "roster.sort.label": "排序方式",
+  "roster.sort.lastModified": "最後修改",
+  "roster.sort.optionsLabel": "排序選項",
+  "roster.sort.recentlyCreated": "最近建立",
   "roster.tabs.agent": "Agent",
-  "roster.tabs.human": "人工成員",
+  "roster.tabs.human": "聯絡人",
   "roster.tabsLabel": "Roster 類型",
   "roster.updateSuccess": "Agent 已更新。",
   "roster.usageStatus.draft": "草稿"
diff --git a/web/i18n/zh-Hant/common.json b/web/i18n/zh-Hant/common.json
index 5c9624bfb39..f84b7fde1d0 100644
--- a/web/i18n/zh-Hant/common.json
+++ b/web/i18n/zh-Hant/common.json
@@ -407,6 +407,8 @@
   "modelProvider.selector.noProviderConfiguredDesc": "前往 Marketplace 安裝，或在設定中配置供應商。",
   "modelProvider.selector.onlyCompatibleModelsShown": "僅顯示相容的模型",
   "modelProvider.selector.showIncompatibleModels": "顯示不相容模型",
+  "modelProvider.selector.suggestion": "推薦",
+  "modelProvider.selector.suggestionTip": "推薦用於 Agent 任務",
   "modelProvider.showModels": "顯示模型",
   "modelProvider.speechToTextModel.key": "語音轉文字模型",
   "modelProvider.speechToTextModel.tip": "設定對話中語音轉文字輸入的預設使用模型。",
diff --git a/web/i18n/zh-Hant/workflow.json b/web/i18n/zh-Hant/workflow.json
index 0b0fae9717d..478795b44df 100644
--- a/web/i18n/zh-Hant/workflow.json
+++ b/web/i18n/zh-Hant/workflow.json
@@ -275,6 +275,10 @@
   "common.workflowAsToolDisabledHint": "發布最新的工作流程，並確保在將其配置為工具之前有一個已連接的使用者輸入節點。",
   "common.workflowAsToolTip": "工作流更新後需要重新配置工具參數",
   "common.workflowProcess": "工作流",
+  "common.workflowProcessFailed": "工作流運行失敗",
+  "common.workflowProcessPaused": "工作流運行已暫停",
+  "common.workflowProcessRunning": "工作流運行中",
+  "common.workflowProcessSucceeded": "工作流運行成功",
   "customWebhook": "自訂 Webhook",
   "debug.copyLastRun": "複製上一次運行",
   "debug.copyLastRunError": "未能複製上一次運行的輸入",
diff --git a/web/service/base.ts b/web/service/base.ts
index 68f839cab6f..69a4c2b6980 100644
--- a/web/service/base.ts
+++ b/web/service/base.ts
@@ -55,6 +55,13 @@ type IOnMessageEnd = (messageEnd: MessageEnd) => void
 export type IOnMessageReplace = (messageReplace: MessageReplace) => void
 export type IOnCompleted = (hasError?: boolean, errorMessage?: string) => void
 export type IOnError = (msg: string, code?: string) => void
+type UnhandledEventError = {
+  conversationId?: string
+  errorCode?: string
+  errorMessage: string
+  messageId?: string
+}
+type IOnUnhandledEvent = (event: Record<string, unknown>) => UnhandledEventError | void
 
 type IOnWorkflowStarted = (workflowStarted: WorkflowStartedResponse) => void
 type IOnWorkflowFinished = (workflowFinished: WorkflowFinishedResponse) => void
@@ -103,6 +110,7 @@ export type IOtherOptions = {
   onMessageEnd?: IOnMessageEnd
   onMessageReplace?: IOnMessageReplace
   onError?: IOnError
+  onUnhandledEvent?: IOnUnhandledEvent
   onCompleted?: IOnCompleted // for stream
   getAbortController?: (abortController: AbortController) => void
 
@@ -227,6 +235,7 @@ export const handleStream = (
   onDataSourceNodeCompleted?: IOnDataSourceNodeCompleted,
   onDataSourceNodeError?: IOnDataSourceNodeError,
   onReasoning?: IOnReasoning,
+  onUnhandledEvent?: IOnUnhandledEvent,
 ) => {
   if (!response.ok)
     throw new Error('Network response was not ok')
@@ -391,6 +400,18 @@ export const handleStream = (
               onDataSourceNodeError?.(bufferObj as DataSourceNodeErrorResponse)
             }
             else {
+              const unhandledEventError = onUnhandledEvent?.(bufferObj)
+              if (unhandledEventError) {
+                onData('', false, {
+                  conversationId: unhandledEventError.conversationId,
+                  messageId: unhandledEventError.messageId ?? '',
+                  errorMessage: unhandledEventError.errorMessage,
+                  errorCode: unhandledEventError.errorCode,
+                })
+                hasError = true
+                onCompleted?.(true, unhandledEventError.errorMessage)
+                return
+              }
               console.warn(`Unknown event: ${bufferObj.event}`, bufferObj)
             }
           }
@@ -513,6 +534,7 @@ export const ssePost = async (
     onDataSourceNodeProcessing,
     onDataSourceNodeCompleted,
     onDataSourceNodeError,
+    onUnhandledEvent,
   } = otherOptions
   const abortController = new AbortController()
 
@@ -622,6 +644,7 @@ export const ssePost = async (
         onDataSourceNodeCompleted,
         onDataSourceNodeError,
         onReasoning,
+        onUnhandledEvent,
       )
     })
     .catch((e) => {
@@ -673,6 +696,7 @@ export const sseGet = async (
     onDataSourceNodeProcessing,
     onDataSourceNodeCompleted,
     onDataSourceNodeError,
+    onUnhandledEvent,
   } = otherOptions
   const abortController = new AbortController()
 
@@ -775,6 +799,7 @@ export const sseGet = async (
         onDataSourceNodeCompleted,
         onDataSourceNodeError,
         onReasoning,
+        onUnhandledEvent,
       )
     })
     .catch((e) => {
diff --git a/web/types/doc-paths.ts b/web/types/doc-paths.ts
index 303f3008171..42e8f048d4b 100644
--- a/web/types/doc-paths.ts
+++ b/web/types/doc-paths.ts
@@ -111,6 +111,7 @@ type UseDifyPath =
   | '/use-dify/build/agent'
   | '/use-dify/build/chatbot'
   | '/use-dify/build/mcp'
+  | '/use-dify/build/new-agent/build'
   | '/use-dify/build/orchestrate-node'
   | '/use-dify/build/predefined-error-handling-logic'
   | '/use-dify/build/shortcut-key'
@@ -299,6 +300,7 @@ type SelfHostPath =
   | '/self-host/use-dify/build/agent'
   | '/self-host/use-dify/build/chatbot'
   | '/self-host/use-dify/build/mcp'
+  | '/self-host/use-dify/build/new-agent/build'
   | '/self-host/use-dify/build/orchestrate-node'
   | '/self-host/use-dify/build/predefined-error-handling-logic'
   | '/self-host/use-dify/build/shortcut-key'
@@ -531,6 +533,7 @@ export const docPathProductAvailability: Record<string, readonly DocsProduct[]>
   '/use-dify/build/agent': ['cloud', 'self-host'],
   '/use-dify/build/chatbot': ['cloud', 'self-host'],
   '/use-dify/build/mcp': ['cloud', 'self-host'],
+  '/use-dify/build/new-agent/build': ['self-host'],
   '/use-dify/build/orchestrate-node': ['cloud', 'self-host'],
   '/use-dify/build/predefined-error-handling-logic': ['cloud', 'self-host'],
   '/use-dify/build/shortcut-key': ['cloud', 'self-host'],