Merge branch 'refs/heads/main' into feat/workflow-parallel-support

# Conflicts:
#	api/core/workflow/entities/variable_pool.py
#	api/core/workflow/nodes/iteration/iteration_node.py
#	api/core/workflow/workflow_engine_manager.py

.devcontainer/post_create_command.sh

@@ -3,8 +3,8 @@
 cd web && npm install
 pipx install poetry
 
-echo 'alias start-api="cd /workspaces/dify/api && flask run --host 0.0.0.0 --port=5001 --debug"' >> ~/.bashrc
-echo 'alias start-worker="cd /workspaces/dify/api && celery -A app.celery worker -P gevent -c 1 --loglevel INFO -Q dataset,generation,mail,ops_trace,app_deletion"' >> ~/.bashrc
+echo 'alias start-api="cd /workspaces/dify/api && poetry run python -m flask run --host 0.0.0.0 --port=5001 --debug"' >> ~/.bashrc
+echo 'alias start-worker="cd /workspaces/dify/api && poetry run python -m celery -A app.celery worker -P gevent -c 1 --loglevel INFO -Q dataset,generation,mail,ops_trace,app_deletion"' >> ~/.bashrc
 echo 'alias start-web="cd /workspaces/dify/web && npm run dev"' >> ~/.bashrc
 echo 'alias start-containers="cd /workspaces/dify/docker && docker-compose -f docker-compose.middleware.yaml -p dify up -d"' >> ~/.bashrc
 

.github/DISCUSSION_TEMPLATE/general.yml

@@ -9,7 +9,7 @@ body:
           required: true
         - label: I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true

.github/DISCUSSION_TEMPLATE/help.yml

@@ -9,7 +9,7 @@ body:
           required: true
         - label: I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true

.github/DISCUSSION_TEMPLATE/suggestion.yml

@@ -9,7 +9,7 @@ body:
           required: true
         - label: I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -14,7 +14,7 @@ body:
           required: true
         - label: I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true
@@ -22,7 +22,6 @@ body:
   - type: input
     attributes:
       label: Dify version
-      placeholder: 0.6.11
       description: See about section in Dify console
     validations:
       required: true

.github/ISSUE_TEMPLATE/document_issue.yml

@@ -12,7 +12,7 @@ body:
           required: true
         - label: I confirm that I am using English to submit report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -12,7 +12,7 @@ body:
           required: true
         - label: I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true

.github/ISSUE_TEMPLATE/translation_issue.yml

@@ -12,14 +12,13 @@ body:
           required: true
         - label: I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)).
           required: true
-        - label: "请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
+        - label: "[FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）"
           required: true
         - label: "Please do not modify this template :) and fill in all the required fields."
           required: true
   - type: input
     attributes:
       label: Dify version
-      placeholder: 0.3.21
       description: Hover over system tray icon or look at Settings
     validations:
       required: true

CONTRIBUTING_JA.md

@@ -1,7 +1,7 @@
 Dify にコントリビュートしたいとお考えなのですね。それは素晴らしいことです。
 私たちは、LLM アプリケーションの構築と管理のための最も直感的なワークフローを設計するという壮大な野望を持っています。人数も資金も限られている新興企業として、コミュニティからの支援は本当に重要です。
 
-私たちは現状を鑑み、機敏かつ迅速に開発をする必要がありますが、同時にあなたのようなコントリビューターの方々に、可能な限りスムーズな貢献体験をしていただきたいと思っています。そのためにこのコントリビュートガイドを作成しました。
+私たちは現状を鑑み、機敏かつ迅速に開発をする必要がありますが、同時にあなた様のようなコントリビューターの方々に、可能な限りスムーズな貢献体験をしていただきたいと思っています。そのためにこのコントリビュートガイドを作成しました。
 コードベースやコントリビュータの方々と私たちがどのように仕事をしているのかに慣れていただき、楽しいパートにすぐに飛び込めるようにすることが目的です。
 
 このガイドは Dify そのものと同様に、継続的に改善されています。実際のプロジェクトに遅れをとることがあるかもしれませんが、ご理解のほどよろしくお願いいたします。
@@ -14,13 +14,13 @@ Dify にコントリビュートしたいとお考えなのですね。それは
 
 ### 機能リクエスト
 
-* 新しい機能要望を出す場合は、提案する機能が何を実現するものなのかを説明し、可能な限り多くのコンテキストを含めてください。[@perzeusss](https://github.com/perzeuss)は、あなたの要望を書き出すのに役立つ [Feature Request Copilot](https://udify.app/chat/MK2kVSnw1gakVwMX) を作ってくれました。気軽に試してみてください。
+* 新しい機能要望を出す場合は、提案する機能が何を実現するものなのかを説明し、可能な限り多くのコンテキストを含めてください。[@perzeusss](https://github.com/perzeuss)は、あなた様の要望を書き出すのに役立つ [Feature Request Copilot](https://udify.app/chat/MK2kVSnw1gakVwMX) を作ってくれました。気軽に試してみてください。
 
 * 既存の課題から 1 つ選びたい場合は、その下にコメントを書いてください。
 
-  関連する方向で作業しているチームメンバーが参加します。すべてが良好であれば、コーディングを開始する許可が与えられます。私たちが変更を提案した場合にあなたの作業が無駄になることがないよう、それまでこの機能の作業を控えていただくようお願いいたします。
+  関連する方向で作業しているチームメンバーが参加します。すべてが良好であれば、コーディングを開始する許可が与えられます。私たちが変更を提案した場合にあなた様の作業が無駄になることがないよう、それまでこの機能の作業を控えていただくようお願いいたします。
 
-  提案された機能がどの分野に属するかによって、あなたは異なるチーム・メンバーと話をするかもしれません。以下は、各チームメンバーが現在取り組んでいる分野の概要です。
+  提案された機能がどの分野に属するかによって、あなた様は異なるチーム・メンバーと話をするかもしれません。以下は、各チームメンバーが現在取り組んでいる分野の概要です。
 
 | Member                                                                                  | Scope                                |
 | --------------------------------------------------------------------------------------- | ------------------------------------ |
@@ -153,7 +153,7 @@ Dify のバックエンドは[Flask](https://flask.palletsprojects.com/en/3.0.x/
 いよいよ、私たちのリポジトリにプルリクエスト (PR) を提出する時が来ました。主要な機能については、まず `deploy/dev` ブランチにマージしてテストしてから `main` ブランチにマージします。
 マージ競合などの問題が発生した場合、またはプル リクエストを開く方法がわからない場合は、[GitHub's pull request tutorial](https://docs.github.com/en/pull-requests/collaborating-with-pull-requests) をチェックしてみてください。
 
-これで完了です！あなたの PR がマージされると、[README](https://github.com/langgenius/dify/blob/main/README.md) にコントリビューターとして紹介されます。
+これで完了です！あなた様の PR がマージされると、[README](https://github.com/langgenius/dify/blob/main/README.md) にコントリビューターとして紹介されます。
 
 ## ヘルプを得る
 

api/.env.example

@@ -183,6 +183,7 @@ UPLOAD_IMAGE_FILE_SIZE_LIMIT=10
 
 # Model Configuration
 MULTIMODAL_SEND_IMAGE_FORMAT=base64
+PROMPT_GENERATION_MAX_TOKENS=512
 
 # Mail configuration, support: resend, smtp
 MAIL_TYPE=
@@ -216,6 +217,7 @@ UNSTRUCTURED_API_KEY=
 
 SSRF_PROXY_HTTP_URL=
 SSRF_PROXY_HTTPS_URL=
+SSRF_DEFAULT_MAX_RETRIES=3
 
 BATCH_UPLOAD_LIMIT=10
 KEYWORD_DATA_SOURCE_TYPE=database

api/app.py

@@ -261,6 +261,7 @@ def after_request(response):
 @app.route('/health')
 def health():
     return Response(json.dumps({
+        'pid': os.getpid(),
         'status': 'ok',
         'version': app.config['CURRENT_VERSION']
     }), status=200, content_type="application/json")
@@ -284,6 +285,7 @@ def threads():
         })
 
     return {
+        'pid': os.getpid(),
         'thread_num': num_threads,
         'threads': thread_list
     }
@@ -293,6 +295,7 @@ def threads():
 def pool_stat():
     engine = db.engine
     return {
+        'pid': os.getpid(),
         'pool_size': engine.pool.size(),
         'checked_in_connections': engine.pool.checkedin(),
         'checked_out_connections': engine.pool.checkedout(),

api/commands.py

@@ -249,8 +249,7 @@ def migrate_knowledge_vector_database():
     create_count = 0
     skipped_count = 0
     total_count = 0
-    config = current_app.config
-    vector_type = config.get('VECTOR_STORE')
+    vector_type = dify_config.VECTOR_STORE
     page = 1
     while True:
         try:
@@ -484,8 +483,7 @@ def convert_to_agent_apps():
 @click.option('--field', default='metadata.doc_id', prompt=False, help='index field , default is metadata.doc_id.')
 def add_qdrant_doc_id_index(field: str):
     click.echo(click.style('Start add qdrant doc_id index.', fg='green'))
-    config = current_app.config
-    vector_type = config.get('VECTOR_STORE')
+    vector_type = dify_config.VECTOR_STORE
     if vector_type != "qdrant":
         click.echo(click.style('Sorry, only support qdrant vector store.', fg='red'))
         return
@@ -502,13 +500,15 @@ def add_qdrant_doc_id_index(field: str):
 
         from core.rag.datasource.vdb.qdrant.qdrant_vector import QdrantConfig
         for binding in bindings:
+            if dify_config.QDRANT_URL is None:
+                raise ValueError('Qdrant url is required.')
             qdrant_config = QdrantConfig(
-                endpoint=config.get('QDRANT_URL'),
-                api_key=config.get('QDRANT_API_KEY'),
+                endpoint=dify_config.QDRANT_URL,
+                api_key=dify_config.QDRANT_API_KEY,
                 root_path=current_app.root_path,
-                timeout=config.get('QDRANT_CLIENT_TIMEOUT'),
-                grpc_port=config.get('QDRANT_GRPC_PORT'),
-                prefer_grpc=config.get('QDRANT_GRPC_ENABLED')
+                timeout=dify_config.QDRANT_CLIENT_TIMEOUT,
+                grpc_port=dify_config.QDRANT_GRPC_PORT,
+                prefer_grpc=dify_config.QDRANT_GRPC_ENABLED
             )
             try:
                 client = qdrant_client.QdrantClient(**qdrant_config.to_qdrant_params())

api/configs/app_config.py

@@ -64,4 +64,6 @@ class DifyConfig(
         return f'{self.HTTP_REQUEST_NODE_MAX_TEXT_SIZE / 1024 / 1024:.2f}MB'
 
     SSRF_PROXY_HTTP_URL: str | None = None
-    SSRF_PROXY_HTTPS_URL: str | None = None
+    SSRF_PROXY_HTTPS_URL: str | None = None
+
+    MODERATION_BUFFER_SIZE: int = Field(default=300, description='The buffer size for moderation.')

api/configs/middleware/__init__.py

@@ -1,4 +1,5 @@
 from typing import Any, Optional
+from urllib.parse import quote_plus
 
 from pydantic import Field, NonNegativeInt, PositiveInt, computed_field
 from pydantic_settings import BaseSettings
@@ -104,7 +105,7 @@ class DatabaseConfig:
         ).strip("&")
         db_extras = f"?{db_extras}" if db_extras else ""
         return (f"{self.SQLALCHEMY_DATABASE_URI_SCHEME}://"
-                f"{self.DB_USERNAME}:{self.DB_PASSWORD}@{self.DB_HOST}:{self.DB_PORT}/{self.DB_DATABASE}"
+                f"{quote_plus(self.DB_USERNAME)}:{quote_plus(self.DB_PASSWORD)}@{self.DB_HOST}:{self.DB_PORT}/{self.DB_DATABASE}" 
                 f"{db_extras}")
 
     SQLALCHEMY_POOL_SIZE: NonNegativeInt = Field(

api/controllers/console/app/generator.py

@@ -1,3 +1,5 @@
+import os
+
 from flask_login import current_user
 from flask_restful import Resource, reqparse
 
@@ -28,13 +30,15 @@ class RuleGenerateApi(Resource):
         args = parser.parse_args()
 
         account = current_user
+        PROMPT_GENERATION_MAX_TOKENS = int(os.getenv('PROMPT_GENERATION_MAX_TOKENS', '512'))
 
         try:
             rules = LLMGenerator.generate_rule_config(
                 tenant_id=account.current_tenant_id,
                 instruction=args['instruction'],
                 model_config=args['model_config'],
-                no_variable=args['no_variable']
+                no_variable=args['no_variable'],
+                rule_config_max_tokens=PROMPT_GENERATION_MAX_TOKENS
             )
         except ProviderTokenNotInitError as ex:
             raise ProviderNotInitializeError(ex.description)

api/controllers/console/auth/login.py

@@ -71,7 +71,7 @@ class ResetPasswordApi(Resource):
         # AccountService.update_password(account, new_password)
 
         # todo: Send email
-        # MAILCHIMP_API_KEY = current_app.config['MAILCHIMP_TRANSACTIONAL_API_KEY']
+        # MAILCHIMP_API_KEY = dify_config.MAILCHIMP_TRANSACTIONAL_API_KEY
         # mailchimp = MailchimpTransactional(MAILCHIMP_API_KEY)
 
         # message = {
@@ -92,7 +92,7 @@ class ResetPasswordApi(Resource):
         #     'message': message,
         #     # required for transactional email
         #     ' settings': {
-        #         'sandbox_mode': current_app.config['MAILCHIMP_SANDBOX_MODE'],
+        #         'sandbox_mode': dify_config.MAILCHIMP_SANDBOX_MODE,
         #     },
         # })
 

api/controllers/service_api/app/workflow.py

@@ -29,22 +29,21 @@ from services.app_generate_service import AppGenerateService
 
 logger = logging.getLogger(__name__)
 
+workflow_run_fields = {
+    'id': fields.String,
+    'workflow_id': fields.String,
+    'status': fields.String,
+    'inputs': fields.Raw,
+    'outputs': fields.Raw,
+    'error': fields.String,
+    'total_steps': fields.Integer,
+    'total_tokens': fields.Integer,
+    'created_at': fields.DateTime,
+    'finished_at': fields.DateTime,
+    'elapsed_time': fields.Float,
+}
 
-class WorkflowRunApi(Resource):
-    workflow_run_fields = {
-        'id': fields.String,
-        'workflow_id': fields.String,
-        'status': fields.String,
-        'inputs': fields.Raw,
-        'outputs': fields.Raw,
-        'error': fields.String,
-        'total_steps': fields.Integer,
-        'total_tokens': fields.Integer,
-        'created_at': fields.DateTime,
-        'finished_at': fields.DateTime,
-        'elapsed_time': fields.Float,
-    }
-
+class WorkflowRunDetailApi(Resource):
     @validate_app_token
     @marshal_with(workflow_run_fields)
     def get(self, app_model: App, workflow_id: str):
@@ -57,7 +56,7 @@ class WorkflowRunApi(Resource):
 
         workflow_run = db.session.query(WorkflowRun).filter(WorkflowRun.id == workflow_id).first()
         return workflow_run
-
+class WorkflowRunApi(Resource):
     @validate_app_token(fetch_user_arg=FetchUserArg(fetch_from=WhereisUserArg.JSON, required=True))
     def post(self, app_model: App, end_user: EndUser):
         """
@@ -117,5 +116,6 @@ class WorkflowTaskStopApi(Resource):
         }
 
 
-api.add_resource(WorkflowRunApi, '/workflows/run/<string:workflow_id>', '/workflows/run')
+api.add_resource(WorkflowRunApi, '/workflows/run')
+api.add_resource(WorkflowRunDetailApi, '/workflows/run/<string:workflow_id>')
 api.add_resource(WorkflowTaskStopApi, '/workflows/tasks/<string:task_id>/stop')

api/core/app/apps/base_app_queue_manager.py

@@ -5,9 +5,9 @@ from collections.abc import Generator
 from enum import Enum
 from typing import Any
 
-from flask import current_app
 from sqlalchemy.orm import DeclarativeMeta
 
+from configs import dify_config
 from core.app.entities.app_invoke_entities import InvokeFrom
 from core.app.entities.queue_entities import (
     AppQueueEvent,
@@ -48,7 +48,7 @@ class AppQueueManager:
         :return:
         """
         # wait for APP_MAX_EXECUTION_TIME seconds to stop listen
-        listen_timeout = current_app.config.get("APP_MAX_EXECUTION_TIME")
+        listen_timeout = dify_config.APP_MAX_EXECUTION_TIME
         start_time = time.time()
         last_ping_time = 0
         while True:

api/core/app/segments/__init__.py

@@ -1,8 +1,21 @@
 from .segment_group import SegmentGroup
-from .segments import NoneSegment, Segment
+from .segments import (
+    ArrayAnySegment,
+    FileSegment,
+    FloatSegment,
+    IntegerSegment,
+    NoneSegment,
+    ObjectSegment,
+    Segment,
+    StringSegment,
+)
 from .types import SegmentType
 from .variables import (
-    ArrayVariable,
+    ArrayAnyVariable,
+    ArrayFileVariable,
+    ArrayNumberVariable,
+    ArrayObjectVariable,
+    ArrayStringVariable,
     FileVariable,
     FloatVariable,
     IntegerVariable,
@@ -20,11 +33,21 @@ __all__ = [
     'SecretVariable',
     'FileVariable',
     'StringVariable',
-    'ArrayVariable',
+    'ArrayAnyVariable',
     'Variable',
     'SegmentType',
     'SegmentGroup',
     'Segment',
     'NoneSegment',
     'NoneVariable',
+    'IntegerSegment',
+    'FloatSegment',
+    'ObjectSegment',
+    'ArrayAnySegment',
+    'FileSegment',
+    'StringSegment',
+    'ArrayStringVariable',
+    'ArrayNumberVariable',
+    'ArrayObjectVariable',
+    'ArrayFileVariable',
 ]

api/core/app/segments/factory.py

@@ -3,14 +3,25 @@ from typing import Any
 
 from core.file.file_obj import FileVar
 
-from .segments import Segment, StringSegment
+from .segments import (
+    ArrayAnySegment,
+    FileSegment,
+    FloatSegment,
+    IntegerSegment,
+    NoneSegment,
+    ObjectSegment,
+    Segment,
+    StringSegment,
+)
 from .types import SegmentType
 from .variables import (
-    ArrayVariable,
+    ArrayFileVariable,
+    ArrayNumberVariable,
+    ArrayObjectVariable,
+    ArrayStringVariable,
     FileVariable,
     FloatVariable,
     IntegerVariable,
-    NoneVariable,
     ObjectVariable,
     SecretVariable,
     StringVariable,
@@ -28,40 +39,48 @@ def build_variable_from_mapping(m: Mapping[str, Any], /) -> Variable:
     match value_type:
         case SegmentType.STRING:
             return StringVariable.model_validate(m)
+        case SegmentType.SECRET:
+            return SecretVariable.model_validate(m)
         case SegmentType.NUMBER if isinstance(value, int):
             return IntegerVariable.model_validate(m)
         case SegmentType.NUMBER if isinstance(value, float):
             return FloatVariable.model_validate(m)
-        case SegmentType.SECRET:
-            return SecretVariable.model_validate(m)
         case SegmentType.NUMBER if not isinstance(value, float | int):
             raise ValueError(f'invalid number value {value}')
+        case SegmentType.FILE:
+            return FileVariable.model_validate(m)
+        case SegmentType.OBJECT if isinstance(value, dict):
+            return ObjectVariable.model_validate(
+                {**m, 'value': {k: build_variable_from_mapping(v) for k, v in value.items()}}
+            )
+        case SegmentType.ARRAY_STRING if isinstance(value, list):
+            return ArrayStringVariable.model_validate({**m, 'value': [build_variable_from_mapping(v) for v in value]})
+        case SegmentType.ARRAY_NUMBER if isinstance(value, list):
+            return ArrayNumberVariable.model_validate({**m, 'value': [build_variable_from_mapping(v) for v in value]})
+        case SegmentType.ARRAY_OBJECT if isinstance(value, list):
+            return ArrayObjectVariable.model_validate({**m, 'value': [build_variable_from_mapping(v) for v in value]})
+        case SegmentType.ARRAY_FILE if isinstance(value, list):
+            return ArrayFileVariable.model_validate({**m, 'value': [build_variable_from_mapping(v) for v in value]})
     raise ValueError(f'not supported value type {value_type}')
 
 
-def build_anonymous_variable(value: Any, /) -> Variable:
-    if value is None:
-        return NoneVariable(name='anonymous')
-    if isinstance(value, str):
-        return StringVariable(name='anonymous', value=value)
-    if isinstance(value, int):
-        return IntegerVariable(name='anonymous', value=value)
-    if isinstance(value, float):
-        return FloatVariable(name='anonymous', value=value)
-    if isinstance(value, dict):
-        # TODO: Limit the depth of the object
-        obj = {k: build_anonymous_variable(v) for k, v in value.items()}
-        return ObjectVariable(name='anonymous', value=obj)
-    if isinstance(value, list):
-        # TODO: Limit the depth of the array
-        elements = [build_anonymous_variable(v) for v in value]
-        return ArrayVariable(name='anonymous', value=elements)
-    if isinstance(value, FileVar):
-        return FileVariable(name='anonymous', value=value)
-    raise ValueError(f'not supported value {value}')
-
-
 def build_segment(value: Any, /) -> Segment:
+    if value is None:
+        return NoneSegment()
     if isinstance(value, str):
         return StringSegment(value=value)
+    if isinstance(value, int):
+        return IntegerSegment(value=value)
+    if isinstance(value, float):
+        return FloatSegment(value=value)
+    if isinstance(value, dict):
+        # TODO: Limit the depth of the object
+        obj = {k: build_segment(v) for k, v in value.items()}
+        return ObjectSegment(value=obj)
+    if isinstance(value, list):
+        # TODO: Limit the depth of the array
+        elements = [build_segment(v) for v in value]
+        return ArrayAnySegment(value=elements)
+    if isinstance(value, FileVar):
+        return FileSegment(value=value)
     raise ValueError(f'not supported value {value}')

api/core/app/segments/parser.py

@@ -1,17 +1,18 @@
 import re
 
-from core.app.segments import SegmentGroup, factory
 from core.workflow.entities.variable_pool import VariablePool
 
+from . import SegmentGroup, factory
+
 VARIABLE_PATTERN = re.compile(r'\{\{#([a-zA-Z0-9_]{1,50}(?:\.[a-zA-Z_][a-zA-Z0-9_]{0,29}){1,10})#\}\}')
 
 
 def convert_template(*, template: str, variable_pool: VariablePool):
     parts = re.split(VARIABLE_PATTERN, template)
     segments = []
-    for part in parts:
+    for part in filter(lambda x: x, parts):
         if '.' in part and (value := variable_pool.get(part.split('.'))):
             segments.append(value)
         else:
             segments.append(factory.build_segment(part))
-    return SegmentGroup(segments=segments)
+    return SegmentGroup(value=segments)

api/core/app/segments/segment_group.py

@@ -1,19 +1,22 @@
-from pydantic import BaseModel
-
 from .segments import Segment
+from .types import SegmentType
 
 
-class SegmentGroup(BaseModel):
-    segments: list[Segment]
+class SegmentGroup(Segment):
+    value_type: SegmentType = SegmentType.GROUP
+    value: list[Segment]
 
     @property
     def text(self):
-        return ''.join([segment.text for segment in self.segments])
+        return ''.join([segment.text for segment in self.value])
 
     @property
     def log(self):
-        return ''.join([segment.log for segment in self.segments])
+        return ''.join([segment.log for segment in self.value])
 
     @property
     def markdown(self):
-        return ''.join([segment.markdown for segment in self.segments])
+        return ''.join([segment.markdown for segment in self.value])
+
+    def to_object(self):
+        return [segment.to_object() for segment in self.value]

api/core/app/segments/segments.py

@@ -1,7 +1,11 @@
+import json
+from collections.abc import Mapping, Sequence
 from typing import Any
 
 from pydantic import BaseModel, ConfigDict, field_validator
 
+from core.file.file_obj import FileVar
+
 from .types import SegmentType
 
 
@@ -34,12 +38,6 @@ class Segment(BaseModel):
         return str(self.value)
 
     def to_object(self) -> Any:
-        if isinstance(self.value, Segment):
-            return self.value.to_object()
-        if isinstance(self.value, list):
-            return [v.to_object() for v in self.value]
-        if isinstance(self.value, dict):
-            return {k: v.to_object() for k, v in self.value.items()}
         return self.value
 
 
@@ -63,3 +61,80 @@ class NoneSegment(Segment):
 class StringSegment(Segment):
     value_type: SegmentType = SegmentType.STRING
     value: str
+
+
+class FloatSegment(Segment):
+    value_type: SegmentType = SegmentType.NUMBER
+    value: float
+
+
+class IntegerSegment(Segment):
+    value_type: SegmentType = SegmentType.NUMBER
+    value: int
+
+
+class FileSegment(Segment):
+    value_type: SegmentType = SegmentType.FILE
+    # TODO: embed FileVar in this model.
+    value: FileVar
+
+    @property
+    def markdown(self) -> str:
+        return self.value.to_markdown()
+
+
+class ObjectSegment(Segment):
+    value_type: SegmentType = SegmentType.OBJECT
+    value: Mapping[str, Segment]
+
+    @property
+    def text(self) -> str:
+        # TODO: Process variables.
+        return json.dumps(self.model_dump()['value'], ensure_ascii=False)
+
+    @property
+    def log(self) -> str:
+        # TODO: Process variables.
+        return json.dumps(self.model_dump()['value'], ensure_ascii=False, indent=2)
+
+    @property
+    def markdown(self) -> str:
+        # TODO: Use markdown code block
+        return json.dumps(self.model_dump()['value'], ensure_ascii=False, indent=2)
+
+    def to_object(self):
+        return {k: v.to_object() for k, v in self.value.items()}
+
+
+class ArraySegment(Segment):
+    @property
+    def markdown(self) -> str:
+        return '\n'.join(['- ' + item.markdown for item in self.value])
+
+    def to_object(self):
+        return [v.to_object() for v in self.value]
+
+
+class ArrayAnySegment(ArraySegment):
+    value_type: SegmentType = SegmentType.ARRAY_ANY
+    value: Sequence[Segment]
+
+
+class ArrayStringSegment(ArraySegment):
+    value_type: SegmentType = SegmentType.ARRAY_STRING
+    value: Sequence[StringSegment]
+
+
+class ArrayNumberSegment(ArraySegment):
+    value_type: SegmentType = SegmentType.ARRAY_NUMBER
+    value: Sequence[FloatSegment | IntegerSegment]
+
+
+class ArrayObjectSegment(ArraySegment):
+    value_type: SegmentType = SegmentType.ARRAY_OBJECT
+    value: Sequence[ObjectSegment]
+
+
+class ArrayFileSegment(ArraySegment):
+    value_type: SegmentType = SegmentType.ARRAY_FILE
+    value: Sequence[FileSegment]

api/core/app/segments/types.py

@@ -6,6 +6,12 @@ class SegmentType(str, Enum):
     NUMBER = 'number'
     STRING = 'string'
     SECRET = 'secret'
-    ARRAY = 'array'
+    ARRAY_ANY = 'array[any]'
+    ARRAY_STRING = 'array[string]'
+    ARRAY_NUMBER = 'array[number]'
+    ARRAY_OBJECT = 'array[object]'
+    ARRAY_FILE = 'array[file]'
     OBJECT = 'object'
     FILE = 'file'
+
+    GROUP = 'group'

api/core/app/segments/variables.py

@@ -1,12 +1,21 @@
-import json
-from collections.abc import Mapping, Sequence
-
 from pydantic import Field
 
-from core.file.file_obj import FileVar
 from core.helper import encrypter
 
-from .segments import NoneSegment, Segment, StringSegment
+from .segments import (
+    ArrayAnySegment,
+    ArrayFileSegment,
+    ArrayNumberSegment,
+    ArrayObjectSegment,
+    ArrayStringSegment,
+    FileSegment,
+    FloatSegment,
+    IntegerSegment,
+    NoneSegment,
+    ObjectSegment,
+    Segment,
+    StringSegment,
+)
 from .types import SegmentType
 
 
@@ -27,53 +36,40 @@ class StringVariable(StringSegment, Variable):
     pass
 
 
-class FloatVariable(Variable):
-    value_type: SegmentType = SegmentType.NUMBER
-    value: float
+class FloatVariable(FloatSegment, Variable):
+    pass
 
 
-class IntegerVariable(Variable):
-    value_type: SegmentType = SegmentType.NUMBER
-    value: int
+class IntegerVariable(IntegerSegment, Variable):
+    pass
 
 
-class ObjectVariable(Variable):
-    value_type: SegmentType = SegmentType.OBJECT
-    value: Mapping[str, Variable]
-
-    @property
-    def text(self) -> str:
-        # TODO: Process variables.
-        return json.dumps(self.model_dump()['value'], ensure_ascii=False)
-
-    @property
-    def log(self) -> str:
-        # TODO: Process variables.
-        return json.dumps(self.model_dump()['value'], ensure_ascii=False, indent=2)
-
-    @property
-    def markdown(self) -> str:
-        # TODO: Use markdown code block
-        return json.dumps(self.model_dump()['value'], ensure_ascii=False, indent=2)
+class FileVariable(FileSegment, Variable):
+    pass
 
 
-class ArrayVariable(Variable):
-    value_type: SegmentType = SegmentType.ARRAY
-    value: Sequence[Variable]
-
-    @property
-    def markdown(self) -> str:
-        return '\n'.join(['- ' + item.markdown for item in self.value])
+class ObjectVariable(ObjectSegment, Variable):
+    pass
 
 
-class FileVariable(Variable):
-    value_type: SegmentType = SegmentType.FILE
-    # TODO: embed FileVar in this model.
-    value: FileVar
+class ArrayAnyVariable(ArrayAnySegment, Variable):
+    pass
 
-    @property
-    def markdown(self) -> str:
-        return self.value.to_markdown()
+
+class ArrayStringVariable(ArrayStringSegment, Variable):
+    pass
+
+
+class ArrayNumberVariable(ArrayNumberSegment, Variable):
+    pass
+
+
+class ArrayObjectVariable(ArrayObjectSegment, Variable):
+    pass
+
+
+class ArrayFileVariable(ArrayFileSegment, Variable):
+    pass
 
 
 class SecretVariable(StringVariable):

api/core/file/upload_file_parser.py

@@ -6,8 +6,7 @@ import os
 import time
 from typing import Optional
 
-from flask import current_app
-
+from configs import dify_config
 from extensions.ext_storage import storage
 
 IMAGE_EXTENSIONS = ['jpg', 'jpeg', 'png', 'webp', 'gif', 'svg']
@@ -23,7 +22,7 @@ class UploadFileParser:
         if upload_file.extension not in IMAGE_EXTENSIONS:
             return None
 
-        if current_app.config['MULTIMODAL_SEND_IMAGE_FORMAT'] == 'url' or force_url:
+        if dify_config.MULTIMODAL_SEND_IMAGE_FORMAT == 'url' or force_url:
             return cls.get_signed_temp_image_url(upload_file.id)
         else:
             # get image file base64
@@ -44,13 +43,13 @@ class UploadFileParser:
         :param upload_file: UploadFile object
         :return:
         """
-        base_url = current_app.config.get('FILES_URL')
+        base_url = dify_config.FILES_URL
         image_preview_url = f'{base_url}/files/{upload_file_id}/image-preview'
 
         timestamp = str(int(time.time()))
         nonce = os.urandom(16).hex()
         data_to_sign = f"image-preview|{upload_file_id}|{timestamp}|{nonce}"
-        secret_key = current_app.config['SECRET_KEY'].encode()
+        secret_key = dify_config.SECRET_KEY.encode()
         sign = hmac.new(secret_key, data_to_sign.encode(), hashlib.sha256).digest()
         encoded_sign = base64.urlsafe_b64encode(sign).decode()
 
@@ -68,7 +67,7 @@ class UploadFileParser:
         :return:
         """
         data_to_sign = f"image-preview|{upload_file_id}|{timestamp}|{nonce}"
-        secret_key = current_app.config['SECRET_KEY'].encode()
+        secret_key = dify_config.SECRET_KEY.encode()
         recalculated_sign = hmac.new(secret_key, data_to_sign.encode(), hashlib.sha256).digest()
         recalculated_encoded_sign = base64.urlsafe_b64encode(recalculated_sign).decode()
 
@@ -77,4 +76,4 @@ class UploadFileParser:
             return False
 
         current_time = int(time.time())
-        return current_time - int(timestamp) <= current_app.config.get('FILES_ACCESS_TIMEOUT')
+        return current_time - int(timestamp) <= dify_config.FILES_ACCESS_TIMEOUT

api/core/helper/position_helper.py

@@ -13,18 +13,10 @@ def get_position_map(folder_path: str, *, file_name: str = "_position.yaml") ->
     :param file_name: the YAML file name, default to '_position.yaml'
     :return: a dict with name as key and index as value
     """
-    position_file_name = os.path.join(folder_path, file_name)
-    if not position_file_name or not os.path.exists(position_file_name):
-        return {}
-    
-    positions = load_yaml_file(position_file_name, ignore_error=True)
-    position_map = {}
-    index = 0
-    for _, name in enumerate(positions):
-        if name and isinstance(name, str):
-            position_map[name.strip()] = index
-            index += 1
-    return position_map
+    position_file_path = os.path.join(folder_path, file_name)
+    yaml_content = load_yaml_file(file_path=position_file_path, default_value=[])
+    positions = [item.strip() for item in yaml_content if item and isinstance(item, str) and item.strip()]
+    return {name: index for index, name in enumerate(positions)}
 
 
 def sort_by_position_map(

api/core/helper/ssrf_proxy.py

@@ -1,48 +1,75 @@
 """
 Proxy requests to avoid SSRF
 """
+import logging
 import os
+import time
 
 import httpx
 
 SSRF_PROXY_ALL_URL = os.getenv('SSRF_PROXY_ALL_URL', '')
 SSRF_PROXY_HTTP_URL = os.getenv('SSRF_PROXY_HTTP_URL', '')
 SSRF_PROXY_HTTPS_URL = os.getenv('SSRF_PROXY_HTTPS_URL', '')
+SSRF_DEFAULT_MAX_RETRIES = int(os.getenv('SSRF_DEFAULT_MAX_RETRIES', '3'))
 
 proxies = {
     'http://': SSRF_PROXY_HTTP_URL,
     'https://': SSRF_PROXY_HTTPS_URL
 } if SSRF_PROXY_HTTP_URL and SSRF_PROXY_HTTPS_URL else None
 
+BACKOFF_FACTOR = 0.5
+STATUS_FORCELIST = [429, 500, 502, 503, 504]
 
-def make_request(method, url, **kwargs):
-    if SSRF_PROXY_ALL_URL:
-        return httpx.request(method=method, url=url, proxy=SSRF_PROXY_ALL_URL, **kwargs)
-    elif proxies:
-        return httpx.request(method=method, url=url, proxies=proxies, **kwargs)
-    else:
-        return httpx.request(method=method, url=url, **kwargs)
+def make_request(method, url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    if "allow_redirects" in kwargs:
+        allow_redirects = kwargs.pop("allow_redirects")
+        if "follow_redirects" not in kwargs:
+            kwargs["follow_redirects"] = allow_redirects
+    
+    retries = 0
+    while retries <= max_retries:
+        try:
+            if SSRF_PROXY_ALL_URL:
+                response = httpx.request(method=method, url=url, proxy=SSRF_PROXY_ALL_URL, **kwargs)
+            elif proxies:
+                response = httpx.request(method=method, url=url, proxies=proxies, **kwargs)
+            else:
+                response = httpx.request(method=method, url=url, **kwargs)
+
+            if response.status_code not in STATUS_FORCELIST:
+                return response
+            else:
+                logging.warning(f"Received status code {response.status_code} for URL {url} which is in the force list")
+
+        except httpx.RequestError as e:
+            logging.warning(f"Request to URL {url} failed on attempt {retries + 1}: {e}")
+
+        retries += 1
+        if retries <= max_retries:
+            time.sleep(BACKOFF_FACTOR * (2 ** (retries - 1)))
+
+    raise Exception(f"Reached maximum retries ({max_retries}) for URL {url}")
 
 
-def get(url, **kwargs):
-    return make_request('GET', url, **kwargs)
+def get(url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    return make_request('GET', url, max_retries=max_retries, **kwargs)
 
 
-def post(url, **kwargs):
-    return make_request('POST', url, **kwargs)
+def post(url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    return make_request('POST', url, max_retries=max_retries, **kwargs)
 
 
-def put(url, **kwargs):
-    return make_request('PUT', url, **kwargs)
+def put(url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    return make_request('PUT', url, max_retries=max_retries, **kwargs)
 
 
-def patch(url, **kwargs):
-    return make_request('PATCH', url, **kwargs)
+def patch(url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    return make_request('PATCH', url, max_retries=max_retries, **kwargs)
 
 
-def delete(url, **kwargs):
-    return make_request('DELETE', url, **kwargs)
+def delete(url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    return make_request('DELETE', url, max_retries=max_retries, **kwargs)
 
 
-def head(url, **kwargs):
-    return make_request('HEAD', url, **kwargs)
+def head(url, max_retries=SSRF_DEFAULT_MAX_RETRIES, **kwargs):
+    return make_request('HEAD', url, max_retries=max_retries, **kwargs)

api/core/indexing_runner.py

@@ -12,6 +12,7 @@ from flask import Flask, current_app
 from flask_login import current_user
 from sqlalchemy.orm.exc import ObjectDeletedError
 
+from configs import dify_config
 from core.errors.error import ProviderTokenNotInitError
 from core.llm_generator.llm_generator import LLMGenerator
 from core.model_manager import ModelInstance, ModelManager
@@ -224,7 +225,7 @@ class IndexingRunner:
         features = FeatureService.get_features(tenant_id)
         if features.billing.enabled:
             count = len(extract_settings)
-            batch_upload_limit = int(current_app.config['BATCH_UPLOAD_LIMIT'])
+            batch_upload_limit = dify_config.BATCH_UPLOAD_LIMIT
             if count > batch_upload_limit:
                 raise ValueError(f"You have reached the batch upload limit of {batch_upload_limit}.")
 
@@ -427,7 +428,7 @@ class IndexingRunner:
             # The user-defined segmentation rule
             rules = json.loads(processing_rule.rules)
             segmentation = rules["segmentation"]
-            max_segmentation_tokens_length = int(current_app.config['INDEXING_MAX_SEGMENTATION_TOKENS_LENGTH'])
+            max_segmentation_tokens_length = dify_config.INDEXING_MAX_SEGMENTATION_TOKENS_LENGTH
             if segmentation["max_tokens"] < 50 or segmentation["max_tokens"] > max_segmentation_tokens_length:
                 raise ValueError(f"Custom segment length should be between 50 and {max_segmentation_tokens_length}.")
 

api/core/llm_generator/llm_generator.py

@@ -118,7 +118,7 @@ class LLMGenerator:
         return questions
 
     @classmethod
-    def generate_rule_config(cls, tenant_id: str, instruction: str, model_config: dict, no_variable: bool) -> dict:
+    def generate_rule_config(cls, tenant_id: str, instruction: str, model_config: dict, no_variable: bool, rule_config_max_tokens: int = 512) -> dict:
         output_parser = RuleConfigGeneratorOutputParser()
 
         error = ""
@@ -130,7 +130,7 @@ class LLMGenerator:
             "error": ""
         }
         model_parameters = {
-            "max_tokens": 512,
+            "max_tokens": rule_config_max_tokens,
             "temperature": 0.01
         }
 

api/core/model_runtime/model_providers/__base/ai_model.py

@@ -162,7 +162,7 @@ class AIModel(ABC):
         # traverse all model_schema_yaml_paths
         for model_schema_yaml_path in model_schema_yaml_paths:
             # read yaml data from yaml file
-            yaml_data = load_yaml_file(model_schema_yaml_path, ignore_error=True)
+            yaml_data = load_yaml_file(model_schema_yaml_path)
 
             new_parameter_rules = []
             for parameter_rule in yaml_data.get('parameter_rules', []):

api/core/model_runtime/model_providers/__base/model_provider.py

@@ -44,7 +44,7 @@ class ModelProvider(ABC):
     
         # read provider schema from yaml file
         yaml_path = os.path.join(current_path, f'{provider_name}.yaml')
-        yaml_data = load_yaml_file(yaml_path, ignore_error=True)
+        yaml_data = load_yaml_file(yaml_path)
     
         try:
             # yaml_data to entity

api/core/model_runtime/model_providers/azure_openai/llm/llm.py

@@ -375,6 +375,10 @@ class AzureOpenAILargeLanguageModel(_CommonAzureOpenAI, LargeLanguageModel):
                 continue
 
             delta = chunk.choices[0]
+            # NOTE: For fix https://github.com/langgenius/dify/issues/5790
+            if delta.delta is None:
+                continue
+
 
             # extract tool calls from response
             self._update_tool_calls(tool_calls=tool_calls, tool_calls_response=delta.delta.tool_calls)

api/core/model_runtime/model_providers/bedrock/llm/_position.yaml

@@ -10,10 +10,14 @@
 - cohere.command-text-v14
 - cohere.command-r-plus-v1.0
 - cohere.command-r-v1.0
+- meta.llama3-1-8b-instruct-v1:0
+- meta.llama3-1-70b-instruct-v1:0
+- meta.llama3-1-405b-instruct-v1:0
 - meta.llama3-8b-instruct-v1:0
 - meta.llama3-70b-instruct-v1:0
 - meta.llama2-13b-chat-v1
 - meta.llama2-70b-chat-v1
+- mistral.mistral-large-2407-v1:0
 - mistral.mistral-small-2402-v1:0
 - mistral.mistral-large-2402-v1:0
 - mistral.mixtral-8x7b-instruct-v0:1

api/core/model_runtime/model_providers/bedrock/llm/cohere.command-r-plus-v1.0.yaml

@@ -3,8 +3,7 @@ label:
   en_US: Command R+
 model_type: llm
 features:
-  #- multi-tool-call
-  - agent-thought
+  - tool-call
   #- stream-tool-call
 model_properties:
   mode: chat

api/core/model_runtime/model_providers/bedrock/llm/cohere.command-r-v1.0.yaml

@@ -3,9 +3,7 @@ label:
   en_US: Command R
 model_type: llm
 features:
-  #- multi-tool-call
-  - agent-thought
-  #- stream-tool-call
+  - tool-call
 model_properties:
   mode: chat
   context_size: 128000

api/core/model_runtime/model_providers/bedrock/llm/llm.py

@@ -17,7 +17,6 @@ from botocore.exceptions import (
     ServiceNotInRegionError,
     UnknownServiceError,
 )
-from cohere import ChatMessage
 
 # local import
 from core.model_runtime.entities.llm_entities import LLMResult, LLMResultChunk, LLMResultChunkDelta
@@ -42,7 +41,6 @@ from core.model_runtime.errors.invoke import (
 )
 from core.model_runtime.errors.validate import CredentialsValidateFailedError
 from core.model_runtime.model_providers.__base.large_language_model import LargeLanguageModel
-from core.model_runtime.model_providers.cohere.llm.llm import CohereLargeLanguageModel
 
 logger = logging.getLogger(__name__)
 
@@ -59,6 +57,7 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
         {'prefix': 'mistral.mixtral-8x7b-instruct', 'support_system_prompts': False, 'support_tool_use': False},
         {'prefix': 'mistral.mistral-large', 'support_system_prompts': True, 'support_tool_use': True},
         {'prefix': 'mistral.mistral-small', 'support_system_prompts': True, 'support_tool_use': True},
+        {'prefix': 'cohere.command-r', 'support_system_prompts': True, 'support_tool_use': True},
         {'prefix': 'amazon.titan', 'support_system_prompts': False, 'support_tool_use': False}
     ]
 
@@ -94,86 +93,8 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
             model_info['model'] = model
             # invoke models via boto3 converse API
             return self._generate_with_converse(model_info, credentials, prompt_messages, model_parameters, stop, stream, user, tools)
-        # invoke Cohere models via boto3 client
-        if "cohere.command-r" in model:
-            return self._generate_cohere_chat(model, credentials, prompt_messages, model_parameters, stop, stream, user, tools)
         # invoke other models via boto3 client
         return self._generate(model, credentials, prompt_messages, model_parameters, stop, stream, user)
-    
-    def _generate_cohere_chat(
-            self, model: str, credentials: dict, prompt_messages: list[PromptMessage], model_parameters: dict,
-            stop: Optional[list[str]] = None, stream: bool = True, user: Optional[str] = None,
-            tools: Optional[list[PromptMessageTool]] = None,) -> Union[LLMResult, Generator]:
-        cohere_llm = CohereLargeLanguageModel()
-        client_config = Config(
-            region_name=credentials["aws_region"]
-        )
-
-        runtime_client = boto3.client(
-            service_name='bedrock-runtime',
-            config=client_config,
-            aws_access_key_id=credentials["aws_access_key_id"],
-            aws_secret_access_key=credentials["aws_secret_access_key"]
-        )
-
-        extra_model_kwargs = {}
-        if stop:
-            extra_model_kwargs['stop_sequences'] = stop
-
-        if tools:
-            tools = cohere_llm._convert_tools(tools)
-            model_parameters['tools'] = tools
-
-        message, chat_histories, tool_results \
-            = cohere_llm._convert_prompt_messages_to_message_and_chat_histories(prompt_messages)
-
-        if tool_results:
-            model_parameters['tool_results'] = tool_results
-
-        payload = {
-            **model_parameters,
-            "message": message,
-            "chat_history": chat_histories,
-        }
-
-        # need workaround for ai21 models which doesn't support streaming
-        if stream:
-            invoke = runtime_client.invoke_model_with_response_stream
-        else:
-            invoke = runtime_client.invoke_model
-
-        def serialize(obj):
-            if isinstance(obj, ChatMessage):
-                return obj.__dict__
-            raise TypeError(f"Type {type(obj)} not serializable")
-
-        try:
-            body_jsonstr=json.dumps(payload, default=serialize)
-            response = invoke(
-                modelId=model,
-                contentType="application/json",
-                accept="*/*",
-                body=body_jsonstr
-            )
-        except ClientError as ex:
-            error_code = ex.response['Error']['Code']
-            full_error_msg = f"{error_code}: {ex.response['Error']['Message']}"
-            raise self._map_client_to_invoke_error(error_code, full_error_msg)
-
-        except (EndpointConnectionError, NoRegionError, ServiceNotInRegionError) as ex:
-            raise InvokeConnectionError(str(ex))
-
-        except UnknownServiceError as ex:
-            raise InvokeServerUnavailableError(str(ex))
-
-        except Exception as ex:
-            raise InvokeError(str(ex))
-
-        if stream:
-            return self._handle_generate_stream_response(model, credentials, response, prompt_messages)
-
-        return self._handle_generate_response(model, credentials, response, prompt_messages)
-
 
     def _generate_with_converse(self, model_info: dict, credentials: dict, prompt_messages: list[PromptMessage], model_parameters: dict,
                 stop: Optional[list[str]] = None, stream: bool = True, user: Optional[str] = None, tools: Optional[list[PromptMessageTool]] = None,) -> Union[LLMResult, Generator]:
@@ -208,14 +129,25 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
 
         if model_info['support_tool_use'] and tools:
             parameters['toolConfig'] = self._convert_converse_tool_config(tools=tools)
+        try:
+            if stream:
+                response = bedrock_client.converse_stream(**parameters)
+                return self._handle_converse_stream_response(model_info['model'], credentials, response, prompt_messages)
+            else:
+                response = bedrock_client.converse(**parameters)
+                return self._handle_converse_response(model_info['model'], credentials, response, prompt_messages)
+        except ClientError as ex:
+            error_code = ex.response['Error']['Code']
+            full_error_msg = f"{error_code}: {ex.response['Error']['Message']}"
+            raise self._map_client_to_invoke_error(error_code, full_error_msg)
+        except (EndpointConnectionError, NoRegionError, ServiceNotInRegionError) as ex:
+            raise InvokeConnectionError(str(ex))
 
-        if stream:
-            response = bedrock_client.converse_stream(**parameters)
-            return self._handle_converse_stream_response(model_info['model'], credentials, response, prompt_messages)
-        else:
-            response = bedrock_client.converse(**parameters)
-            return self._handle_converse_response(model_info['model'], credentials, response, prompt_messages)
+        except UnknownServiceError as ex:
+            raise InvokeServerUnavailableError(str(ex))
 
+        except Exception as ex:
+            raise InvokeError(str(ex))
     def _handle_converse_response(self, model: str, credentials: dict, response: dict,
                                 prompt_messages: list[PromptMessage]) -> LLMResult:
         """
@@ -558,7 +490,6 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
         except ClientError as ex:
             error_code = ex.response['Error']['Code']
             full_error_msg = f"{error_code}: {ex.response['Error']['Message']}"
-
             raise CredentialsValidateFailedError(str(self._map_client_to_invoke_error(error_code, full_error_msg)))
 
         except Exception as ex:
@@ -571,38 +502,9 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
         :param message: PromptMessage to convert.
         :return: String representation of the message.
         """
-        
-        if model_prefix == "anthropic":
-            human_prompt_prefix = "\n\nHuman:"
-            human_prompt_postfix = ""
-            ai_prompt = "\n\nAssistant:"
-
-        elif model_prefix == "meta":
-            # LLAMA3
-            if model_name.startswith("llama3"):
-                human_prompt_prefix = "<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n"
-                human_prompt_postfix = "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
-                ai_prompt = "\n\nAssistant:"
-            else:
-                # LLAMA2
-                human_prompt_prefix = "\n[INST]"
-                human_prompt_postfix = "[\\INST]\n"
-                ai_prompt = ""
-
-        elif model_prefix == "mistral":
-            human_prompt_prefix = "<s>[INST]"
-            human_prompt_postfix = "[\\INST]\n"
-            ai_prompt = "\n\nAssistant:"
-
-        elif model_prefix == "amazon":
-            human_prompt_prefix = "\n\nUser:"
-            human_prompt_postfix = ""
-            ai_prompt = "\n\nBot:"
-        
-        else:
-            human_prompt_prefix = ""
-            human_prompt_postfix = ""
-            ai_prompt = ""
+        human_prompt_prefix = ""
+        human_prompt_postfix = ""
+        ai_prompt = ""
 
         content = message.content
 
@@ -653,13 +555,7 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
         model_prefix = model.split('.')[0]
         model_name = model.split('.')[1]
 
-        if model_prefix == "amazon":
-            payload["textGenerationConfig"] = { **model_parameters }
-            payload["textGenerationConfig"]["stopSequences"] = ["User:"]
-            
-            payload["inputText"] = self._convert_messages_to_prompt(prompt_messages, model_prefix)
-        
-        elif model_prefix == "ai21":
+        if model_prefix == "ai21":
             payload["temperature"] = model_parameters.get("temperature")
             payload["topP"] = model_parameters.get("topP")
             payload["maxTokens"] = model_parameters.get("maxTokens")
@@ -671,28 +567,12 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
                 payload["frequencyPenalty"] = {model_parameters.get("frequencyPenalty")}
             if model_parameters.get("countPenalty"):
                 payload["countPenalty"] = {model_parameters.get("countPenalty")}
-        
-        elif model_prefix == "mistral":
-            payload["temperature"] = model_parameters.get("temperature")
-            payload["top_p"] = model_parameters.get("top_p")
-            payload["max_tokens"] = model_parameters.get("max_tokens")
-            payload["prompt"] = self._convert_messages_to_prompt(prompt_messages, model_prefix)
-            payload["stop"] = stop[:10] if stop else []
-
-        elif model_prefix == "anthropic":
-            payload = { **model_parameters }
-            payload["prompt"] = self._convert_messages_to_prompt(prompt_messages, model_prefix)
-            payload["stop_sequences"] = ["\n\nHuman:"] + (stop if stop else [])
-            
+                    
         elif model_prefix == "cohere":
             payload = { **model_parameters }
             payload["prompt"] = prompt_messages[0].content
             payload["stream"] = stream
         
-        elif model_prefix == "meta":
-            payload = { **model_parameters }
-            payload["prompt"] = self._convert_messages_to_prompt(prompt_messages, model_prefix, model_name)
-
         else:
             raise ValueError(f"Got unknown model prefix {model_prefix}")
         
@@ -783,36 +663,16 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
         # get output text and calculate num tokens based on model / provider
         model_prefix = model.split('.')[0]
 
-        if model_prefix == "amazon":
-            output = response_body.get("results")[0].get("outputText").strip('\n')
-            prompt_tokens = response_body.get("inputTextTokenCount")
-            completion_tokens = response_body.get("results")[0].get("tokenCount")
-
-        elif model_prefix == "ai21":
+        if model_prefix == "ai21":
             output = response_body.get('completions')[0].get('data').get('text')
             prompt_tokens = len(response_body.get("prompt").get("tokens"))
             completion_tokens = len(response_body.get('completions')[0].get('data').get('tokens'))
-
-        elif model_prefix == "anthropic":
-            output = response_body.get("completion")
-            prompt_tokens = self.get_num_tokens(model, credentials, prompt_messages)
-            completion_tokens = self.get_num_tokens(model, credentials, output if output else '')
             
         elif model_prefix == "cohere":
             output = response_body.get("generations")[0].get("text")
             prompt_tokens = self.get_num_tokens(model, credentials, prompt_messages)
             completion_tokens = self.get_num_tokens(model, credentials, output if output else '')
-            
-        elif model_prefix == "meta":
-            output = response_body.get("generation").strip('\n')
-            prompt_tokens = response_body.get("prompt_token_count")
-            completion_tokens = response_body.get("generation_token_count")
-        
-        elif model_prefix == "mistral":
-            output = response_body.get("outputs")[0].get("text")
-            prompt_tokens = response.get('ResponseMetadata').get('HTTPHeaders').get('x-amzn-bedrock-input-token-count')
-            completion_tokens = response.get('ResponseMetadata').get('HTTPHeaders').get('x-amzn-bedrock-output-token-count')
-
+                        
         else:
             raise ValueError(f"Got unknown model prefix {model_prefix} when handling block response")
 
@@ -883,26 +743,10 @@ class BedrockLargeLanguageModel(LargeLanguageModel):
             payload = json.loads(chunk.get('bytes').decode())
 
             model_prefix = model.split('.')[0]
-            if model_prefix == "amazon":
-                content_delta = payload.get("outputText").strip('\n')
-                finish_reason = payload.get("completion_reason")
- 
-            elif model_prefix == "anthropic":
-                content_delta = payload.get("completion")
-                finish_reason = payload.get("stop_reason")
-
-            elif model_prefix == "cohere":
+            if model_prefix == "cohere":
                 content_delta = payload.get("text")
                 finish_reason = payload.get("finish_reason")
             
-            elif model_prefix == "mistral":
-                content_delta = payload.get('outputs')[0].get("text")
-                finish_reason = payload.get('outputs')[0].get("stop_reason")
-
-            elif model_prefix == "meta":
-                content_delta = payload.get("generation").strip('\n')
-                finish_reason = payload.get("stop_reason")
-            
             else:
                 raise ValueError(f"Got unknown model prefix {model_prefix} when handling stream response")
 

api/core/model_runtime/model_providers/bedrock/llm/meta.llama3-1-405b-instruct-v1.0.yaml

@@ -0,0 +1,25 @@
+model: meta.llama3-1-405b-instruct-v1:0
+label:
+  en_US: Llama 3.1 405B Instruct
+model_type: llm
+model_properties:
+  mode: completion
+  context_size: 128000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    default: 0.5
+  - name: top_p
+    use_template: top_p
+    default: 0.9
+  - name: max_gen_len
+    use_template: max_tokens
+    required: true
+    default: 512
+    min: 1
+    max: 2048
+pricing:
+  input: '0.00532'
+  output: '0.016'
+  unit: '0.001'
+  currency: USD

api/core/model_runtime/model_providers/bedrock/llm/meta.llama3-1-70b-instruct-v1.0.yaml

@@ -0,0 +1,25 @@
+model: meta.llama3-1-70b-instruct-v1:0
+label:
+  en_US: Llama 3.1 Instruct 70B
+model_type: llm
+model_properties:
+  mode: completion
+  context_size: 128000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    default: 0.5
+  - name: top_p
+    use_template: top_p
+    default: 0.9
+  - name: max_gen_len
+    use_template: max_tokens
+    required: true
+    default: 512
+    min: 1
+    max: 2048
+pricing:
+  input: '0.00265'
+  output: '0.0035'
+  unit: '0.001'
+  currency: USD

api/core/model_runtime/model_providers/bedrock/llm/meta.llama3-1-8b-instruct-v1.0.yaml

@@ -0,0 +1,25 @@
+model: meta.llama3-1-8b-instruct-v1:0
+label:
+  en_US: Llama 3.1 Instruct 8B
+model_type: llm
+model_properties:
+  mode: completion
+  context_size: 128000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    default: 0.5
+  - name: top_p
+    use_template: top_p
+    default: 0.9
+  - name: max_gen_len
+    use_template: max_tokens
+    required: true
+    default: 512
+    min: 1
+    max: 2048
+pricing:
+  input: '0.0003'
+  output: '0.0006'
+  unit: '0.001'
+  currency: USD

api/core/model_runtime/model_providers/bedrock/llm/mistral.mistral-large-2407-v1.0.yaml

@@ -0,0 +1,29 @@
+model: mistral.mistral-large-2407-v1:0
+label:
+  en_US: Mistral Large 2 (24.07)
+model_type: llm
+features:
+  - tool-call
+model_properties:
+  mode: completion
+  context_size: 128000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    required: false
+    default: 0.7
+  - name: top_p
+    use_template: top_p
+    required: false
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    required: true
+    default: 512
+    min: 1
+    max: 8192
+pricing:
+  input: '0.003'
+  output: '0.009'
+  unit: '0.001'
+  currency: USD

api/core/model_runtime/model_providers/deepseek/llm/deepseek-chat.yaml

@@ -23,7 +23,7 @@ parameter_rules:
     type: int
     default: 4096
     min: 1
-    max: 4096
+    max: 8192
     help:
       zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
       en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
@@ -57,6 +57,18 @@ parameter_rules:
     help:
       zh_Hans: 介于 -2.0 和 2.0 之间的数字。如果该值为正，那么新 token 会根据其在已有文本中的出现频率受到相应的惩罚，降低模型重复相同内容的可能性。
       en_US: A number between -2.0 and 2.0. If the value is positive, new tokens are penalized based on their frequency of occurrence in existing text, reducing the likelihood that the model will repeat the same content.
+  - name: response_format
+    label:
+      zh_Hans: 回复格式
+      en_US: response_format
+    type: string
+    help:
+      zh_Hans: 指定模型必须输出的格式
+      en_US: specifying the format that the model must output
+    required: false
+    options:
+      - text
+      - json_object
 pricing:
   input: '1'
   output: '2'

api/core/model_runtime/model_providers/hunyuan/hunyuan.yaml

@@ -18,6 +18,7 @@ help:
     en_US: https://console.cloud.tencent.com/cam/capi
 supported_model_types:
   - llm
+  - text-embedding
 configurate_methods:
   - predefined-model
 provider_credential_schema:

api/core/model_runtime/model_providers/hunyuan/llm/llm.py

@@ -14,6 +14,7 @@ from core.model_runtime.entities.message_entities import (
     PromptMessage,
     PromptMessageTool,
     SystemPromptMessage,
+    ToolPromptMessage,
     UserPromptMessage,
 )
 from core.model_runtime.errors.invoke import InvokeError
@@ -44,6 +45,17 @@ class HunyuanLargeLanguageModel(LargeLanguageModel):
             "Stream": stream,
             **custom_parameters,
         }
+        # add Tools and ToolChoice
+        if (tools and len(tools) > 0):
+            params['ToolChoice'] = "auto"
+            params['Tools'] = [{
+                "Type": "function", 
+                "Function": {
+                    "Name": tool.name, 
+                    "Description": tool.description,
+                    "Parameters": json.dumps(tool.parameters)
+                }
+            } for tool in tools]
 
         request.from_json_string(json.dumps(params))
         response = client.ChatCompletions(request)
@@ -89,9 +101,43 @@ class HunyuanLargeLanguageModel(LargeLanguageModel):
 
     def _convert_prompt_messages_to_dicts(self, prompt_messages: list[PromptMessage]) -> list[dict]:
         """Convert a list of PromptMessage objects to a list of dictionaries with 'Role' and 'Content' keys."""
-        return [{"Role": message.role.value, "Content": message.content} for message in prompt_messages]
+        dict_list = []
+        for message in prompt_messages:
+            if isinstance(message, AssistantPromptMessage):
+                tool_calls = message.tool_calls
+                if (tool_calls and len(tool_calls) > 0):
+                    dict_tool_calls = [
+                        {
+                            "Id": tool_call.id,
+                            "Type": tool_call.type,
+                            "Function": {
+                                "Name": tool_call.function.name,
+                                "Arguments": tool_call.function.arguments if (tool_call.function.arguments == "") else "{}"
+                            }
+                        } for tool_call in tool_calls]
+                    
+                    dict_list.append({ 
+                        "Role": message.role.value,
+                        # fix set content = "" while tool_call request
+                        # fix [hunyuan] None, [TencentCloudSDKException] code:InvalidParameter message:Messages Content and Contents not allowed empty at the same time.
+                        "Content": " ", # message.content if (message.content is not None) else "", 
+                        "ToolCalls":  dict_tool_calls
+                    })
+                else:
+                    dict_list.append({ "Role": message.role.value, "Content": message.content })
+            elif isinstance(message, ToolPromptMessage):
+                tool_execute_result = { "result": message.content }
+                content =json.dumps(tool_execute_result, ensure_ascii=False)
+                dict_list.append({ "Role": message.role.value, "Content": content, "ToolCallId": message.tool_call_id  })
+            else:
+                dict_list.append({ "Role": message.role.value, "Content": message.content })
+        return dict_list
 
     def _handle_stream_chat_response(self, model, credentials, prompt_messages, resp):
+
+        tool_call = None
+        tool_calls = []
+
         for index, event in enumerate(resp):
             logging.debug("_handle_stream_chat_response, event: %s", event)
 
@@ -109,20 +155,54 @@ class HunyuanLargeLanguageModel(LargeLanguageModel):
             usage = data.get('Usage', {})
             prompt_tokens = usage.get('PromptTokens', 0)
             completion_tokens = usage.get('CompletionTokens', 0)
-            usage = self._calc_response_usage(model, credentials, prompt_tokens, completion_tokens)
+
+            response_tool_calls = delta.get('ToolCalls')
+            if (response_tool_calls is not None):
+                new_tool_calls = self._extract_response_tool_calls(response_tool_calls)
+                if (len(new_tool_calls) > 0):
+                    new_tool_call = new_tool_calls[0]
+                    if (tool_call is None): tool_call = new_tool_call
+                    elif (tool_call.id != new_tool_call.id):
+                        tool_calls.append(tool_call)
+                        tool_call = new_tool_call
+                    else:
+                        tool_call.function.name += new_tool_call.function.name
+                        tool_call.function.arguments += new_tool_call.function.arguments
+                if (tool_call is not None and len(tool_call.function.name) > 0 and len(tool_call.function.arguments) > 0):
+                    tool_calls.append(tool_call)
+                    tool_call = None
 
             assistant_prompt_message = AssistantPromptMessage(
                 content=message_content,
                 tool_calls=[]
             )
+            # rewrite content = "" while tool_call to avoid show content on web page
+            if (len(tool_calls) > 0): assistant_prompt_message.content = ""
+            
+            # add tool_calls to assistant_prompt_message
+            if (finish_reason == 'tool_calls'):
+                assistant_prompt_message.tool_calls = tool_calls
+                tool_call = None
+                tool_calls = []
 
-            delta_chunk = LLMResultChunkDelta(
-                index=index,
-                role=delta.get('Role', 'assistant'),
-                message=assistant_prompt_message,
-                usage=usage,
-                finish_reason=finish_reason,
-            )
+            if (len(finish_reason) > 0):
+                usage = self._calc_response_usage(model, credentials, prompt_tokens, completion_tokens)
+
+                delta_chunk = LLMResultChunkDelta(
+                    index=index,
+                    role=delta.get('Role', 'assistant'),
+                    message=assistant_prompt_message,
+                    usage=usage,
+                    finish_reason=finish_reason,
+                )
+                tool_call = None
+                tool_calls = []
+
+            else:
+                delta_chunk = LLMResultChunkDelta(
+                    index=index,
+                    message=assistant_prompt_message,
+                )
 
             yield LLMResultChunk(
                 model=model,
@@ -177,12 +257,15 @@ class HunyuanLargeLanguageModel(LargeLanguageModel):
         """
         human_prompt = "\n\nHuman:"
         ai_prompt = "\n\nAssistant:"
+        tool_prompt = "\n\nTool:"
         content = message.content
 
         if isinstance(message, UserPromptMessage):
             message_text = f"{human_prompt} {content}"
         elif isinstance(message, AssistantPromptMessage):
             message_text = f"{ai_prompt} {content}"
+        elif isinstance(message, ToolPromptMessage):
+            message_text = f"{tool_prompt} {content}"
         elif isinstance(message, SystemPromptMessage):
             message_text = content
         else:
@@ -203,3 +286,30 @@ class HunyuanLargeLanguageModel(LargeLanguageModel):
         return {
             InvokeError: [TencentCloudSDKException],
         }
+    
+    def _extract_response_tool_calls(self,
+                                     response_tool_calls: list[dict]) \
+            -> list[AssistantPromptMessage.ToolCall]:
+        """
+        Extract tool calls from response
+
+        :param response_tool_calls: response tool calls
+        :return: list of tool calls
+        """
+        tool_calls = []
+        if response_tool_calls:
+            for response_tool_call in response_tool_calls:
+                response_function = response_tool_call.get('Function', {})
+                function = AssistantPromptMessage.ToolCall.ToolCallFunction(
+                    name=response_function.get('Name', ''),
+                    arguments=response_function.get('Arguments', '')
+                )
+
+                tool_call = AssistantPromptMessage.ToolCall(
+                    id=response_tool_call.get('Id', 0),
+                    type='function',
+                    function=function
+                )
+                tool_calls.append(tool_call)
+
+        return tool_calls

api/core/model_runtime/model_providers/hunyuan/text_embedding/hunyuan-text-embedding.yaml

@@ -0,0 +1,5 @@
+model: hunyuan-embedding
+model_type: text-embedding
+model_properties:
+  context_size: 1024
+  max_chunks: 1

api/core/model_runtime/model_providers/hunyuan/text_embedding/text_embedding.py

@@ -0,0 +1,173 @@
+import json
+import logging
+import time
+from typing import Optional
+
+from tencentcloud.common import credential
+from tencentcloud.common.exception import TencentCloudSDKException
+from tencentcloud.common.profile.client_profile import ClientProfile
+from tencentcloud.common.profile.http_profile import HttpProfile
+from tencentcloud.hunyuan.v20230901 import hunyuan_client, models
+
+from core.model_runtime.entities.model_entities import PriceType
+from core.model_runtime.entities.text_embedding_entities import EmbeddingUsage, TextEmbeddingResult
+from core.model_runtime.errors.invoke import (
+    InvokeError,
+)
+from core.model_runtime.errors.validate import CredentialsValidateFailedError
+from core.model_runtime.model_providers.__base.text_embedding_model import TextEmbeddingModel
+
+logger = logging.getLogger(__name__)
+
+class HunyuanTextEmbeddingModel(TextEmbeddingModel):
+    """
+    Model class for Hunyuan text embedding model.
+    """
+
+    def _invoke(self, model: str, credentials: dict,
+                texts: list[str], user: Optional[str] = None) \
+            -> TextEmbeddingResult:
+        """
+        Invoke text embedding model
+
+        :param model: model name
+        :param credentials: model credentials
+        :param texts: texts to embed
+        :param user: unique user id
+        :return: embeddings result
+        """
+
+        if model != 'hunyuan-embedding':
+            raise ValueError('Invalid model name')
+        
+        client = self._setup_hunyuan_client(credentials)
+
+        embeddings = []
+        token_usage = 0
+
+        for input in texts:
+            request = models.GetEmbeddingRequest()
+            params = {
+                "Input": input
+            }
+            request.from_json_string(json.dumps(params))
+            response = client.GetEmbedding(request)
+            usage = response.Usage.TotalTokens
+
+            embeddings.extend([data.Embedding for data in response.Data])
+            token_usage += usage
+
+        result = TextEmbeddingResult(
+            model=model,
+            embeddings=embeddings,
+            usage=self._calc_response_usage(
+                model=model,
+                credentials=credentials,
+                tokens=token_usage
+            )
+        )
+
+        return result
+
+    def validate_credentials(self, model: str, credentials: dict) -> None:
+        """
+        Validate credentials
+        """
+        try:
+            client = self._setup_hunyuan_client(credentials)
+
+            req = models.ChatCompletionsRequest()
+            params = {
+                "Model": model,
+                "Messages": [{
+                    "Role": "user",
+                    "Content": "hello"
+                }],
+                "TopP": 1,
+                "Temperature": 0,
+                "Stream": False
+            }
+            req.from_json_string(json.dumps(params))
+            client.ChatCompletions(req)
+        except Exception as e:
+            raise CredentialsValidateFailedError(f'Credentials validation failed: {e}')
+
+    def _setup_hunyuan_client(self, credentials):
+        secret_id = credentials['secret_id']
+        secret_key = credentials['secret_key']
+        cred = credential.Credential(secret_id, secret_key)
+        httpProfile = HttpProfile()
+        httpProfile.endpoint = "hunyuan.tencentcloudapi.com"
+        clientProfile = ClientProfile()
+        clientProfile.httpProfile = httpProfile
+        client = hunyuan_client.HunyuanClient(cred, "", clientProfile)
+        return client
+    
+    def _calc_response_usage(self, model: str, credentials: dict, tokens: int) -> EmbeddingUsage:
+        """
+        Calculate response usage
+
+        :param model: model name
+        :param credentials: model credentials
+        :param tokens: input tokens
+        :return: usage
+        """
+        # get input price info
+        input_price_info = self.get_price(
+            model=model,
+            credentials=credentials,
+            price_type=PriceType.INPUT,
+            tokens=tokens
+        )
+
+        # transform usage
+        usage = EmbeddingUsage(
+            tokens=tokens,
+            total_tokens=tokens,
+            unit_price=input_price_info.unit_price,
+            price_unit=input_price_info.unit,
+            total_price=input_price_info.total_amount,
+            currency=input_price_info.currency,
+            latency=time.perf_counter() - self.started_at
+        )
+
+        return usage
+    
+    @property
+    def _invoke_error_mapping(self) -> dict[type[InvokeError], list[type[Exception]]]:
+        """
+        Map model invoke error to unified error
+        The key is the error type thrown to the caller
+        The value is the error type thrown by the model,
+        which needs to be converted into a unified error type for the caller.
+
+        :return: Invoke error mapping
+        """
+        return {
+            InvokeError: [TencentCloudSDKException],
+        }
+    
+    def get_num_tokens(self, model: str, credentials: dict, texts: list[str]) -> int:
+        """
+        Get number of tokens for given prompt messages
+
+        :param model: model name
+        :param credentials: model credentials
+        :param texts: texts to embed
+        :return:
+        """
+        # client = self._setup_hunyuan_client(credentials)
+
+        num_tokens = 0
+        for text in texts:
+            num_tokens += self._get_num_tokens_by_gpt2(text)
+            # use client.GetTokenCount to get num tokens
+            # request = models.GetTokenCountRequest()
+            # params = {
+            #     "Prompt": text
+            # }
+            # request.from_json_string(json.dumps(params))
+            # response = client.GetTokenCount(request)
+            # num_tokens += response.TokenCount
+
+        return num_tokens

api/core/model_runtime/model_providers/novita/llm/Nous-Hermes-2-Mixtral-8x7B-DPO.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.0027'
+  output: '0.0027'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/airoboros-l2-70b.yaml

@@ -0,0 +1,41 @@
+model: jondurbin/airoboros-l2-70b
+label:
+  zh_Hans: jondurbin/airoboros-l2-70b
+  en_US: jondurbin/airoboros-l2-70b
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 4096
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.005'
+  output: '0.005'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/dolphin-mixtral-8x22b.yaml

@@ -0,0 +1,41 @@
+model: cognitivecomputations/dolphin-mixtral-8x22b
+label:
+  zh_Hans: cognitivecomputations/dolphin-mixtral-8x22b
+  en_US: cognitivecomputations/dolphin-mixtral-8x22b
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 16000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.009'
+  output: '0.009'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/gemma-2-9b-it.yaml

@@ -0,0 +1,41 @@
+model: google/gemma-2-9b-it
+label:
+  zh_Hans: google/gemma-2-9b-it
+  en_US: google/gemma-2-9b-it
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.0008'
+  output: '0.0008'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/hermes-2-pro-llama-3-8b.yaml

@@ -0,0 +1,41 @@
+model: nousresearch/hermes-2-pro-llama-3-8b
+label:
+  zh_Hans: nousresearch/hermes-2-pro-llama-3-8b
+  en_US: nousresearch/hermes-2-pro-llama-3-8b
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.0014'
+  output: '0.0014'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/l3-70b-euryale-v2.1.yaml

@@ -0,0 +1,41 @@
+model: sao10k/l3-70b-euryale-v2.1
+label:
+  zh_Hans: sao10k/l3-70b-euryale-v2.1
+  en_US: sao10k/l3-70b-euryale-v2.1
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 16000
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.0148'
+  output: '0.0148'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/llama-3-70b-instruct.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.0051'
+  output: '0.0074'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/llama-3-8b-instruct.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.00063'
+  output: '0.00063'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/llama-3.1-405b-instruct.yaml

@@ -0,0 +1,41 @@
+model: meta-llama/llama-3.1-405b-instruct
+label:
+  zh_Hans: meta-llama/llama-3.1-405b-instruct
+  en_US: meta-llama/llama-3.1-405b-instruct
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.03'
+  output: '0.05'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/llama-3.1-70b-instruct.yaml

@@ -0,0 +1,41 @@
+model: meta-llama/llama-3.1-70b-instruct
+label:
+  zh_Hans: meta-llama/llama-3.1-70b-instruct
+  en_US: meta-llama/llama-3.1-70b-instruct
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.0055'
+  output: '0.0076'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/llama-3.1-8b-instruct.yaml

@@ -0,0 +1,41 @@
+model: meta-llama/llama-3.1-8b-instruct
+label:
+  zh_Hans: meta-llama/llama-3.1-8b-instruct
+  en_US: meta-llama/llama-3.1-8b-instruct
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 8192
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.001'
+  output: '0.001'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/lzlv_70b.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.0058'
+  output: '0.0078'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/midnight-rose-70b.yaml

@@ -0,0 +1,41 @@
+model: sophosympatheia/midnight-rose-70b
+label:
+  zh_Hans: sophosympatheia/midnight-rose-70b
+  en_US: sophosympatheia/midnight-rose-70b
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 4096
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.008'
+  output: '0.008'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/mistral-7b-instruct.yaml

@@ -0,0 +1,41 @@
+model: mistralai/mistral-7b-instruct
+label:
+  zh_Hans: mistralai/mistral-7b-instruct
+  en_US: mistralai/mistral-7b-instruct
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+    min: 0
+    max: 2
+    default: 1
+  - name: top_p
+    use_template: top_p
+    min: 0
+    max: 1
+    default: 1
+  - name: max_tokens
+    use_template: max_tokens
+    min: 1
+    max: 2048
+    default: 512
+  - name: frequency_penalty
+    use_template: frequency_penalty
+    min: -2
+    max: 2
+    default: 0
+  - name: presence_penalty
+    use_template: presence_penalty
+    min: -2
+    max: 2
+    default: 0
+pricing:
+  input: '0.00059'
+  output: '0.00059'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/mythomax-l2-13b.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.00119'
+  output: '0.00119'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/nous-hermes-llama2-13b.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.0017'
+  output: '0.0017'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/openhermes-2.5-mistral-7b.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.0017'
+  output: '0.0017'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/llm/wizardlm-2-8x22b.yaml

@@ -34,3 +34,8 @@ parameter_rules:
     min: -2
     max: 2
     default: 0
+pricing:
+  input: '0.0064'
+  output: '0.0064'
+  unit: '0.0001'
+  currency: USD

api/core/model_runtime/model_providers/novita/novita.yaml

@@ -1,6 +1,9 @@
 provider: novita
 label:
   en_US: novita.ai
+description:
+  en_US: An LLM API that matches various application scenarios with high cost-effectiveness.
+  zh_Hans: 适配多种海外应用场景的高性价比 LLM API
 icon_small:
   en_US: icon_s_en.svg
 icon_large:
@@ -11,7 +14,7 @@ help:
     en_US: Get your API key from novita.ai
     zh_Hans: 从 novita.ai 获取 API Key
   url:
-    en_US: https://novita.ai/dashboard/key?utm_source=dify
+    en_US: https://novita.ai/settings#key-management?utm_source=dify&utm_medium=ch&utm_campaign=api
 supported_model_types:
   - llm
 configurate_methods:

api/core/model_runtime/model_providers/openai/tts/tts.py

@@ -114,7 +114,8 @@ class OpenAIText2SpeechModel(_CommonOpenAI, TTSModel):
             # doc: https://platform.openai.com/docs/guides/text-to-speech
             credentials_kwargs = self._to_credential_kwargs(credentials)
             client = OpenAI(**credentials_kwargs)
-            if not voice or voice not in self.get_tts_model_voices(model=model, credentials=credentials):
+            model_support_voice = [x.get("value") for x in self.get_tts_model_voices(model=model, credentials=credentials)]
+            if not voice or voice not in model_support_voice:
                 voice = self._get_model_default_voice(model, credentials)
             word_limit = self._get_model_word_limit(model, credentials)
             if len(content_text) > word_limit:

api/core/model_runtime/model_providers/siliconflow/llm/deepdeek-coder-v2-instruct.yaml

@@ -0,0 +1,30 @@
+model: deepseek-ai/DeepSeek-Coder-V2-Instruct
+label:
+  en_US: deepseek-ai/DeepSeek-Coder-V2-Instruct
+model_type: llm
+features:
+  - agent-thought
+model_properties:
+  mode: chat
+  context_size: 32768
+parameter_rules:
+  - name: temperature
+    use_template: temperature
+  - name: max_tokens
+    use_template: max_tokens
+    type: int
+    default: 512
+    min: 1
+    max: 4096
+    help:
+      zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
+      en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
+  - name: top_p
+    use_template: top_p
+  - name: frequency_penalty
+    use_template: frequency_penalty
+pricing:
+  input: '1.33'
+  output: '1.33'
+  unit: '0.000001'
+  currency: RMB

api/core/model_runtime/model_providers/siliconflow/llm/deepseek-v2-chat.yaml

@@ -1,11 +1,9 @@
 model: deepseek-ai/deepseek-v2-chat
 label:
-  en_US: deepseek-ai/deepseek-v2-chat
+  en_US: deepseek-ai/DeepSeek-V2-Chat
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 32768

api/core/model_runtime/model_providers/siliconflow/llm/glm4-9b-chat.yaml

@@ -1,11 +1,9 @@
 model: zhipuai/glm4-9B-chat
 label:
-  en_US: zhipuai/glm4-9B-chat
+  en_US: THUDM/glm-4-9b-chat
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 32768

api/core/model_runtime/model_providers/siliconflow/llm/qwen2-57b-a14b-instruct.yaml

@@ -1,11 +1,9 @@
 model: alibaba/Qwen2-57B-A14B-Instruct
 label:
-  en_US: alibaba/Qwen2-57B-A14B-Instruct
+  en_US: Qwen/Qwen2-57B-A14B-Instruct
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 32768

api/core/model_runtime/model_providers/siliconflow/llm/qwen2-72b-instruct.yaml

@@ -1,11 +1,9 @@
 model: alibaba/Qwen2-72B-Instruct
 label:
-  en_US: alibaba/Qwen2-72B-Instruct
+  en_US: Qwen/Qwen2-72B-Instruct
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 32768

api/core/model_runtime/model_providers/siliconflow/llm/qwen2-7b-instruct.yaml

@@ -1,11 +1,9 @@
 model: alibaba/Qwen2-7B-Instruct
 label:
-  en_US: alibaba/Qwen2-7B-Instruct
+  en_US: Qwen/Qwen2-7B-Instruct
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 32768

api/core/model_runtime/model_providers/siliconflow/llm/yi-1.5-34b-chat.yaml

@@ -1,11 +1,9 @@
 model: 01-ai/Yi-1.5-34B-Chat
 label:
-  en_US: 01-ai/Yi-1.5-34B-Chat
+  en_US: 01-ai/Yi-1.5-34B-Chat-16K
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 16384

api/core/model_runtime/model_providers/siliconflow/llm/yi-1.5-6b-chat.yaml

@@ -3,9 +3,7 @@ label:
   en_US: 01-ai/Yi-1.5-6B-Chat
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 4096

api/core/model_runtime/model_providers/siliconflow/llm/yi-1.5-9b-chat.yaml

@@ -1,11 +1,9 @@
 model: 01-ai/Yi-1.5-9B-Chat
 label:
-  en_US: 01-ai/Yi-1.5-9B-Chat
+  en_US: 01-ai/Yi-1.5-9B-Chat-16K
 model_type: llm
 features:
-  - multi-tool-call
   - agent-thought
-  - stream-tool-call
 model_properties:
   mode: chat
   context_size: 16384

api/core/model_runtime/model_providers/siliconflow/siliconflow.py

@@ -19,7 +19,7 @@ class SiliconflowProvider(ModelProvider):
             model_instance = self.get_model_instance(ModelType.LLM)
 
             model_instance.validate_credentials(
-                model='deepseek-ai/deepseek-v2-chat',
+                model='deepseek-ai/DeepSeek-V2-Chat',
                 credentials=credentials
             )
         except CredentialsValidateFailedError as ex:

api/core/model_runtime/model_providers/tongyi/llm/llm.py

@@ -501,8 +501,7 @@ You should also complete the text started with ``` but not tell ``` directly.
                     'role': 'assistant',
                     'content': content if not rich_content else [{"text": content}],
                     'tool_calls': [tool_call.model_dump() for tool_call in
-                                   prompt_message.tool_calls] if prompt_message.tool_calls else []
-
+                                   prompt_message.tool_calls] if prompt_message.tool_calls else None
                 })
             elif isinstance(prompt_message, ToolPromptMessage):
                 tongyi_messages.append({

api/core/moderation/output_moderation.py

@@ -6,6 +6,7 @@ from typing import Any, Optional
 from flask import Flask, current_app
 from pydantic import BaseModel, ConfigDict
 
+from configs import dify_config
 from core.app.apps.base_app_queue_manager import AppQueueManager, PublishFrom
 from core.app.entities.queue_entities import QueueMessageReplaceEvent
 from core.moderation.base import ModerationAction, ModerationOutputsResult
@@ -20,8 +21,6 @@ class ModerationRule(BaseModel):
 
 
 class OutputModeration(BaseModel):
-    DEFAULT_BUFFER_SIZE: int = 300
-
     tenant_id: str
     app_id: str
 
@@ -76,10 +75,10 @@ class OutputModeration(BaseModel):
         return final_output
 
     def start_thread(self) -> threading.Thread:
-        buffer_size = int(current_app.config.get('MODERATION_BUFFER_SIZE', self.DEFAULT_BUFFER_SIZE))
+        buffer_size = dify_config.MODERATION_BUFFER_SIZE
         thread = threading.Thread(target=self.worker, kwargs={
             'flask_app': current_app._get_current_object(),
-            'buffer_size': buffer_size if buffer_size > 0 else self.DEFAULT_BUFFER_SIZE
+            'buffer_size': buffer_size if buffer_size > 0 else dify_config.MODERATION_BUFFER_SIZE
         })
 
         thread.start()

api/core/ops/ops_trace_manager.py

@@ -298,34 +298,29 @@ class TraceTask:
         self.file_base_url = os.getenv("FILES_URL", "http://127.0.0.1:5001")
 
     def execute(self):
-        method_name, trace_info = self.preprocess()
-        return trace_info
+        return self.preprocess()
 
     def preprocess(self):
-        if self.trace_type == TraceTaskName.CONVERSATION_TRACE:
-            return TraceTaskName.CONVERSATION_TRACE, self.conversation_trace(**self.kwargs)
-        if self.trace_type == TraceTaskName.WORKFLOW_TRACE:
-            return TraceTaskName.WORKFLOW_TRACE, self.workflow_trace(self.workflow_run, self.conversation_id)
-        elif self.trace_type == TraceTaskName.MESSAGE_TRACE:
-            return TraceTaskName.MESSAGE_TRACE, self.message_trace(self.message_id)
-        elif self.trace_type == TraceTaskName.MODERATION_TRACE:
-            return TraceTaskName.MODERATION_TRACE, self.moderation_trace(self.message_id, self.timer, **self.kwargs)
-        elif self.trace_type == TraceTaskName.SUGGESTED_QUESTION_TRACE:
-            return TraceTaskName.SUGGESTED_QUESTION_TRACE, self.suggested_question_trace(
+        preprocess_map = {
+            TraceTaskName.CONVERSATION_TRACE: lambda: self.conversation_trace(**self.kwargs),
+            TraceTaskName.WORKFLOW_TRACE: lambda: self.workflow_trace(self.workflow_run, self.conversation_id),
+            TraceTaskName.MESSAGE_TRACE: lambda: self.message_trace(self.message_id),
+            TraceTaskName.MODERATION_TRACE: lambda: self.moderation_trace(
                 self.message_id, self.timer, **self.kwargs
-            )
-        elif self.trace_type == TraceTaskName.DATASET_RETRIEVAL_TRACE:
-            return TraceTaskName.DATASET_RETRIEVAL_TRACE, self.dataset_retrieval_trace(
+            ),
+            TraceTaskName.SUGGESTED_QUESTION_TRACE: lambda: self.suggested_question_trace(
                 self.message_id, self.timer, **self.kwargs
-            )
-        elif self.trace_type == TraceTaskName.TOOL_TRACE:
-            return TraceTaskName.TOOL_TRACE, self.tool_trace(self.message_id, self.timer, **self.kwargs)
-        elif self.trace_type == TraceTaskName.GENERATE_NAME_TRACE:
-            return TraceTaskName.GENERATE_NAME_TRACE, self.generate_name_trace(
+            ),
+            TraceTaskName.DATASET_RETRIEVAL_TRACE: lambda: self.dataset_retrieval_trace(
+                self.message_id, self.timer, **self.kwargs
+            ),
+            TraceTaskName.TOOL_TRACE: lambda: self.tool_trace(self.message_id, self.timer, **self.kwargs),
+            TraceTaskName.GENERATE_NAME_TRACE: lambda: self.generate_name_trace(
                 self.conversation_id, self.timer, **self.kwargs
-            )
-        else:
-            return '', {}
+            ),
+        }
+
+        return preprocess_map.get(self.trace_type, lambda: None)()
 
     # process methods for different trace types
     def conversation_trace(self, **kwargs):

api/core/rag/datasource/keyword/jieba/jieba.py

@@ -2,9 +2,9 @@ import json
 from collections import defaultdict
 from typing import Any, Optional
 
-from flask import current_app
 from pydantic import BaseModel
 
+from configs import dify_config
 from core.rag.datasource.keyword.jieba.jieba_keyword_table_handler import JiebaKeywordTableHandler
 from core.rag.datasource.keyword.keyword_base import BaseKeyword
 from core.rag.models.document import Document
@@ -139,7 +139,7 @@ class Jieba(BaseKeyword):
             if keyword_table_dict:
                 return keyword_table_dict['__data__']['table']
         else:
-            keyword_data_source_type = current_app.config['KEYWORD_DATA_SOURCE_TYPE']
+            keyword_data_source_type = dify_config.KEYWORD_DATA_SOURCE_TYPE
             dataset_keyword_table = DatasetKeywordTable(
                 dataset_id=self.dataset.id,
                 keyword_table='',

api/core/rag/datasource/vdb/milvus/milvus_vector.py

@@ -5,6 +5,7 @@ from uuid import uuid4
 
 from pydantic import BaseModel, model_validator
 from pymilvus import MilvusClient, MilvusException, connections
+from pymilvus.milvus_client import IndexParams
 
 from configs import dify_config
 from core.rag.datasource.entity.embedding import Embeddings
@@ -250,11 +251,15 @@ class MilvusVector(BaseVector):
                 # Since primary field is auto-id, no need to track it
                 self._fields.remove(Field.PRIMARY_KEY.value)
 
+                # Create Index params for the collection
+                index_params_obj = IndexParams()
+                index_params_obj.add_index(field_name=Field.VECTOR.value, **index_params)
+
                 # Create the collection
                 collection_name = self._collection_name
-                self._client.create_collection_with_schema(collection_name=collection_name,
-                                                           schema=schema, index_param=index_params,
-                                                           consistency_level=self._consistency_level)
+                self._client.create_collection(collection_name=collection_name,
+                                               schema=schema, index_params=index_params_obj,
+                                               consistency_level=self._consistency_level)
             redis_client.set(collection_exist_cache_key, 1, ex=3600)
 
     def _init_client(self, config) -> MilvusClient:

api/core/rag/datasource/vdb/oracle/oraclevector.py

@@ -55,7 +55,7 @@ CREATE TABLE IF NOT EXISTS {table_name} (
 )
 """
 SQL_CREATE_INDEX = """
-CREATE INDEX idx_docs_{table_name} ON {table_name}(text) 
+CREATE INDEX IF NOT EXISTS idx_docs_{table_name} ON {table_name}(text) 
 INDEXTYPE IS CTXSYS.CONTEXT PARAMETERS 
 ('FILTER CTXSYS.NULL_FILTER SECTION GROUP CTXSYS.HTML_SECTION_GROUP LEXER sys.my_chinese_vgram_lexer')
 """
@@ -248,7 +248,7 @@ class OracleVector(BaseVector):
 
     def delete(self) -> None:
         with self._get_cursor() as cur:
-            cur.execute(f"DROP TABLE IF EXISTS {self.table_name}")
+            cur.execute(f"DROP TABLE IF EXISTS {self.table_name} cascade constraints")
 
     def _create_collection(self, dimension: int):
         cache_key = f"vector_indexing_{self._collection_name}"

api/core/rag/extractor/excel_extractor.py

@@ -3,6 +3,7 @@ import os
 from typing import Optional
 
 import pandas as pd
+from openpyxl import load_workbook
 
 from core.rag.extractor.extractor_base import BaseExtractor
 from core.rag.models.document import Document
@@ -28,26 +29,48 @@ class ExcelExtractor(BaseExtractor):
         self._autodetect_encoding = autodetect_encoding
 
     def extract(self) -> list[Document]:
-        """ Load from Excel file in xls or xlsx format using Pandas."""
+        """ Load from Excel file in xls or xlsx format using Pandas and openpyxl."""
         documents = []
-        # Determine the file extension
         file_extension = os.path.splitext(self._file_path)[-1].lower()
-        # Read each worksheet of an Excel file using Pandas
+
         if file_extension == '.xlsx':
-            excel_file = pd.ExcelFile(self._file_path, engine='openpyxl')
+            wb = load_workbook(self._file_path, data_only=True)
+            for sheet_name in wb.sheetnames:
+                sheet = wb[sheet_name]
+                data = sheet.values
+                cols = next(data)
+                df = pd.DataFrame(data, columns=cols)
+
+                df.dropna(how='all', inplace=True)
+
+                for index, row in df.iterrows():
+                    page_content = []
+                    for col_index, (k, v) in enumerate(row.items()):
+                        if pd.notna(v):
+                            cell = sheet.cell(row=index + 2,
+                                              column=col_index + 1)  # +2 to account for header and 1-based index
+                            if cell.hyperlink:
+                                value = f"[{v}]({cell.hyperlink.target})"
+                                page_content.append(f'"{k}":"{value}"')
+                            else:
+                                page_content.append(f'"{k}":"{v}"')
+                    documents.append(Document(page_content=';'.join(page_content),
+                                              metadata={'source': self._file_path}))
+
         elif file_extension == '.xls':
             excel_file = pd.ExcelFile(self._file_path, engine='xlrd')
+            for sheet_name in excel_file.sheet_names:
+                df = excel_file.parse(sheet_name=sheet_name)
+                df.dropna(how='all', inplace=True)
+
+                for _, row in df.iterrows():
+                    page_content = []
+                    for k, v in row.items():
+                        if pd.notna(v):
+                            page_content.append(f'"{k}":"{v}"')
+                    documents.append(Document(page_content=';'.join(page_content),
+                                              metadata={'source': self._file_path}))
         else:
             raise ValueError(f"Unsupported file extension: {file_extension}")
-        for sheet_name in excel_file.sheet_names:
-            df: pd.DataFrame = excel_file.parse(sheet_name=sheet_name)
-
-            # filter out rows with all NaN values
-            df.dropna(how='all', inplace=True)
-
-            # transform each row into a Document
-            documents += [Document(page_content=';'.join(f'"{k}":"{v}"' for k, v in row.items() if pd.notna(v)),
-                                   metadata={'source': self._file_path},
-                                   ) for _, row in df.iterrows()]
 
         return documents

api/core/rag/extractor/extract_processor.py

@@ -4,9 +4,8 @@ from pathlib import Path
 from typing import Union
 from urllib.parse import unquote
 
-import requests
-
 from configs import dify_config
+from core.helper import ssrf_proxy
 from core.rag.extractor.csv_extractor import CSVExtractor
 from core.rag.extractor.entity.datasource_type import DatasourceType
 from core.rag.extractor.entity.extract_setting import ExtractSetting
@@ -51,7 +50,7 @@ class ExtractProcessor:
 
     @classmethod
     def load_from_url(cls, url: str, return_text: bool = False) -> Union[list[Document], str]:
-        response = requests.get(url, headers={
+        response = ssrf_proxy.get(url, headers={
             "User-Agent": USER_AGENT
         })
 

api/core/rag/extractor/markdown_extractor.py

@@ -54,8 +54,16 @@ class MarkdownExtractor(BaseExtractor):
 
         current_header = None
         current_text = ""
+        code_block_flag = False
 
         for line in lines:
+            if line.startswith("```"):
+                code_block_flag = not code_block_flag
+                current_text += line + "\n"
+                continue
+            if code_block_flag:
+                current_text += line + "\n"
+                continue
             header_match = re.match(r"^#+\s", line)
             if header_match:
                 if current_header is not None:

api/core/tools/provider/builtin/cogview/tools/cogview3.py

@@ -30,7 +30,6 @@ class CogView3Tool(BuiltinTool):
         if not prompt:
             return self.create_text_message('Please input prompt')
         # get size
-        print(tool_parameters.get('prompt', 'square'))
         size = size_mapping[tool_parameters.get('size', 'square')]
         # get n
         n = tool_parameters.get('n', 1)
@@ -58,8 +57,9 @@ class CogView3Tool(BuiltinTool):
         result = []
         for image in response.data:
             result.append(self.create_image_message(image=image.url))
-        result.append(self.create_text_message(
-            f'\nGenerate image source to Seed ID: {seed_id}'))
+            result.append(self.create_json_message({
+                "url": image.url,
+            }))
         return result
 
     @staticmethod

api/core/tools/provider/builtin/firecrawl/firecrawl.py

@@ -1,22 +1,19 @@
 from core.tools.errors import ToolProviderCredentialValidationError
-from core.tools.provider.builtin.firecrawl.tools.crawl import CrawlTool
+from core.tools.provider.builtin.firecrawl.tools.scrape import ScrapeTool
 from core.tools.provider.builtin_tool_provider import BuiltinToolProviderController
 
 
 class FirecrawlProvider(BuiltinToolProviderController):
     def _validate_credentials(self, credentials: dict) -> None:
         try:
-            # Example validation using the Crawl tool
-            CrawlTool().fork_tool_runtime(
+            # Example validation using the ScrapeTool, only scraping title for minimize content
+            ScrapeTool().fork_tool_runtime(
                 runtime={"credentials": credentials}
             ).invoke(
                 user_id='',
                 tool_parameters={
-                    "url": "https://example.com",
-                    "includes": '', 
-                    "excludes": '', 
-                    "limit": 1,
-                    "onlyMainContent": True,
+                    "url": "https://google.com",
+                    "onlyIncludeTags": 'title'
                 }
             )
         except Exception as e:

api/core/tools/provider/builtin/firecrawl/firecrawl.yaml

@@ -31,8 +31,5 @@ credentials_for_provider:
     label:
       en_US: Firecrawl server's Base URL
       zh_Hans: Firecrawl服务器的API URL
-      pt_BR: Firecrawl server's Base URL
     placeholder:
-      en_US: https://www.firecrawl.dev
-      zh_HansL: https://www.firecrawl.dev
-      pt_BR: https://www.firecrawl.dev
+      en_US: https://api.firecrawl.dev

api/core/tools/provider/builtin/firecrawl/firecrawl_appx.py

@@ -1,3 +1,4 @@
+import json
 import logging
 import time
 from collections.abc import Mapping
@@ -8,6 +9,7 @@ from requests.exceptions import HTTPError
 
 logger = logging.getLogger(__name__)
 
+
 class FirecrawlApp:
     def __init__(self, api_key: str | None = None, base_url: str | None = None):
         self.api_key = api_key
@@ -25,14 +27,16 @@ class FirecrawlApp:
         return headers
 
     def _request(
-        self,
-        method: str,
-        url: str,
-        data: Mapping[str, Any] | None = None,
-        headers: Mapping[str, str] | None = None,
-        retries: int = 3,
-        backoff_factor: float = 0.3,
+            self,
+            method: str,
+            url: str,
+            data: Mapping[str, Any] | None = None,
+            headers: Mapping[str, str] | None = None,
+            retries: int = 3,
+            backoff_factor: float = 0.3,
     ) -> Mapping[str, Any] | None:
+        if not headers:
+            headers = self._prepare_headers()
         for i in range(retries):
             try:
                 response = requests.request(method, url, json=data, headers=headers)
@@ -47,47 +51,51 @@ class FirecrawlApp:
 
     def scrape_url(self, url: str, **kwargs):
         endpoint = f'{self.base_url}/v0/scrape'
-        headers = self._prepare_headers()
         data = {'url': url, **kwargs}
-        response = self._request('POST', endpoint, data, headers)
         logger.debug(f"Sent request to {endpoint=} body={data}")
+        response = self._request('POST', endpoint, data)
         if response is None:
             raise HTTPError("Failed to scrape URL after multiple retries")
         return response
 
     def search(self, query: str, **kwargs):
         endpoint = f'{self.base_url}/v0/search'
-        headers = self._prepare_headers()
         data = {'query': query, **kwargs}
-        response = self._request('POST', endpoint, data, headers)
         logger.debug(f"Sent request to {endpoint=} body={data}")
+        response = self._request('POST', endpoint, data)
         if response is None:
             raise HTTPError("Failed to perform search after multiple retries")
         return response
 
     def crawl_url(
-        self, url: str, wait: bool = False, poll_interval: int = 5, idempotency_key: str | None = None, **kwargs
+            self, url: str, wait: bool = True, poll_interval: int = 5, idempotency_key: str | None = None, **kwargs
     ):
         endpoint = f'{self.base_url}/v0/crawl'
         headers = self._prepare_headers(idempotency_key)
-        data = {'url': url, **kwargs['params']}
-        response = self._request('POST', endpoint, data, headers)
+        data = {'url': url, **kwargs}
         logger.debug(f"Sent request to {endpoint=} body={data}")
+        response = self._request('POST', endpoint, data, headers)
         if response is None:
             raise HTTPError("Failed to initiate crawl after multiple retries")
         job_id: str = response['jobId']
         if wait:
             return self._monitor_job_status(job_id=job_id, poll_interval=poll_interval)
-        return job_id
+        return response
 
     def check_crawl_status(self, job_id: str):
         endpoint = f'{self.base_url}/v0/crawl/status/{job_id}'
-        headers = self._prepare_headers()
-        response = self._request('GET', endpoint, headers=headers)
+        response = self._request('GET', endpoint)
         if response is None:
             raise HTTPError(f"Failed to check status for job {job_id} after multiple retries")
         return response
 
+    def cancel_crawl_job(self, job_id: str):
+        endpoint = f'{self.base_url}/v0/crawl/cancel/{job_id}'
+        response = self._request('DELETE', endpoint)
+        if response is None:
+            raise HTTPError(f"Failed to cancel job {job_id} after multiple retries")
+        return response
+
     def _monitor_job_status(self, job_id: str, poll_interval: int):
         while True:
             status = self.check_crawl_status(job_id)
@@ -96,3 +104,21 @@ class FirecrawlApp:
             elif status['status'] == 'failed':
                 raise HTTPError(f'Job {job_id} failed: {status["error"]}')
             time.sleep(poll_interval)
+
+
+def get_array_params(tool_parameters: dict[str, Any], key):
+    param = tool_parameters.get(key)
+    if param:
+        return param.split(',')
+
+
+def get_json_params(tool_parameters: dict[str, Any], key):
+    param = tool_parameters.get(key)
+    if param:
+        try:
+            # support both single quotes and double quotes
+            param = param.replace("'", '"')
+            param = json.loads(param)
+        except:
+            raise ValueError(f"Invalid {key} format.")
+        return param

api/core/tools/provider/builtin/firecrawl/tools/crawl.py

@@ -1,36 +1,48 @@
-import json
-from typing import Any, Union
+from typing import Any
 
 from core.tools.entities.tool_entities import ToolInvokeMessage
-from core.tools.provider.builtin.firecrawl.firecrawl_appx import FirecrawlApp
+from core.tools.provider.builtin.firecrawl.firecrawl_appx import FirecrawlApp, get_array_params, get_json_params
 from core.tools.tool.builtin_tool import BuiltinTool
 
 
 class CrawlTool(BuiltinTool):
-    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]:
-        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'], base_url=self.runtime.credentials['base_url'])
+    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> ToolInvokeMessage:
+        """
+        the crawlerOptions and pageOptions comes from doc here:
+        https://docs.firecrawl.dev/api-reference/endpoint/crawl
+        """
+        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'],
+                           base_url=self.runtime.credentials['base_url'])
+        crawlerOptions = {}
+        pageOptions = {}
 
-        options = {
-            'crawlerOptions': {
-                'excludes': tool_parameters.get('excludes', '').split(',') if tool_parameters.get('excludes') else [],
-                'includes': tool_parameters.get('includes', '').split(',') if tool_parameters.get('includes') else [],
-                'limit': tool_parameters.get('limit', 5)
-            },
-            'pageOptions': {
-                'onlyMainContent': tool_parameters.get('onlyMainContent', False)
-            }
-        }
+        wait_for_results = tool_parameters.get('wait_for_results', True)
+
+        crawlerOptions['excludes'] = get_array_params(tool_parameters, 'excludes')
+        crawlerOptions['includes'] = get_array_params(tool_parameters, 'includes')
+        crawlerOptions['returnOnlyUrls'] = tool_parameters.get('returnOnlyUrls', False)
+        crawlerOptions['maxDepth'] = tool_parameters.get('maxDepth')
+        crawlerOptions['mode'] = tool_parameters.get('mode')
+        crawlerOptions['ignoreSitemap'] = tool_parameters.get('ignoreSitemap', False)
+        crawlerOptions['limit'] = tool_parameters.get('limit', 5)
+        crawlerOptions['allowBackwardCrawling'] = tool_parameters.get('allowBackwardCrawling', False)
+        crawlerOptions['allowExternalContentLinks'] = tool_parameters.get('allowExternalContentLinks', False)
+
+        pageOptions['headers'] = get_json_params(tool_parameters, 'headers')
+        pageOptions['includeHtml'] = tool_parameters.get('includeHtml', False)
+        pageOptions['includeRawHtml'] = tool_parameters.get('includeRawHtml', False)
+        pageOptions['onlyIncludeTags'] = get_array_params(tool_parameters, 'onlyIncludeTags')
+        pageOptions['removeTags'] = get_array_params(tool_parameters, 'removeTags')
+        pageOptions['onlyMainContent'] = tool_parameters.get('onlyMainContent', False)
+        pageOptions['replaceAllPathsWithAbsolutePaths'] = tool_parameters.get('replaceAllPathsWithAbsolutePaths', False)
+        pageOptions['screenshot'] = tool_parameters.get('screenshot', False)
+        pageOptions['waitFor'] = tool_parameters.get('waitFor', 0)
 
         crawl_result = app.crawl_url(
-            url=tool_parameters['url'], 
-            params=options,
-            wait=True
+            url=tool_parameters['url'],
+            wait=wait_for_results,
+            crawlerOptions=crawlerOptions,
+            pageOptions=pageOptions
         )
 
-        if not isinstance(crawl_result, str):
-            crawl_result = json.dumps(crawl_result, ensure_ascii=False, indent=4)
-
-        if not crawl_result:
-            return self.create_text_message("Crawl request failed.")
-
-        return self.create_text_message(crawl_result)
+        return self.create_json_message(crawl_result)

api/core/tools/provider/builtin/firecrawl/tools/crawl.yaml

@@ -3,76 +3,243 @@ identity:
   author: Richards Tu
   label:
     en_US: Crawl
-    zh_Hans: 爬取
+    zh_Hans: 深度爬取
 description:
   human:
-    en_US: Extract data from a website by crawling through a URL.
-    zh_Hans: 通过URL从网站中提取数据。
+    en_US: Recursively search through a urls subdomains, and gather the content.
+    zh_Hans: 递归爬取一个网址的子域名，并收集内容。
   llm: This tool initiates a web crawl to extract data from a specified URL. It allows configuring crawler options such as including or excluding URL patterns, generating alt text for images using LLMs (paid plan required), limiting the maximum number of pages to crawl, and returning only the main content of the page. The tool can return either a list of crawled documents or a list of URLs based on the provided options.
 parameters:
   - name: url
     type: string
     required: true
     label:
-      en_US: URL to crawl
-      zh_Hans: 要爬取的URL
+      en_US: Start URL
+      zh_Hans: 起始URL
     human_description:
-      en_US: The URL of the website to crawl and extract data from.
-      zh_Hans: 要爬取并提取数据的网站URL。
+      en_US: The base URL to start crawling from.
+      zh_Hans: 要爬取网站的起始URL。
     llm_description: The URL of the website that needs to be crawled. This is a required parameter.
     form: llm
+  - name: wait_for_results
+    type: boolean
+    default: true
+    label:
+      en_US: Wait For Results
+      zh_Hans: 等待爬取结果
+    human_description:
+      en_US: If you choose not to wait, it will directly return a job ID. You can use this job ID to check the crawling results or cancel the crawling task, which is usually very useful for a large-scale crawling task.
+      zh_Hans: 如果选择不等待，则会直接返回一个job_id，可以通过job_id查询爬取结果或取消爬取任务，这通常对于一个大型爬取任务来说非常有用。
+    form: form
+############## Crawl Options #######################
   - name: includes
     type: string
     required: false
     label:
       en_US: URL patterns to include
       zh_Hans: 要包含的URL模式
+    placeholder:
+      en_US: Use commas to separate multiple tags
+      zh_Hans: 多个标签时使用半角逗号分隔
     human_description:
-      en_US: Specify URL patterns to include during the crawl. Only pages matching these patterns will be crawled, you can use ',' to separate multiple patterns.
-      zh_Hans: 指定爬取过程中要包含的URL模式。只有与这些模式匹配的页面才会被爬取。
+      en_US: |
+       Only pages matching these patterns will be crawled. Example: blog/*, about/*
+      zh_Hans: 只有与这些模式匹配的页面才会被爬取。示例：blog/*, about/*
     form: form
-    default: ''
   - name: excludes
     type: string
-    required: false
     label:
       en_US: URL patterns to exclude
       zh_Hans: 要排除的URL模式
+    placeholder:
+      en_US: Use commas to separate multiple tags
+      zh_Hans: 多个标签时使用半角逗号分隔
     human_description:
-      en_US: Specify URL patterns to exclude during the crawl. Pages matching these patterns will be skipped, you can use ',' to separate multiple patterns.
-      zh_Hans: 指定爬取过程中要排除的URL模式。匹配这些模式的页面将被跳过。
+      en_US: |
+       Pages matching these patterns will be skipped. Example: blog/*, about/*
+      zh_Hans: 匹配这些模式的页面将被跳过。示例：blog/*, about/*
+    form: form
+  - name: returnOnlyUrls
+    type: boolean
+    default: false
+    label:
+      en_US: return Only Urls
+      zh_Hans: 仅返回URL
+    human_description:
+      en_US: |
+       If true, returns only the URLs as a list on the crawl status. Attention: the return response will be a list of URLs inside the data, not a list of documents.
+      zh_Hans: 只返回爬取到的网页链接，而不是网页内容本身。
+    form: form
+  - name: maxDepth
+    type: number
+    label:
+      en_US: Maximum crawl depth
+      zh_Hans: 爬取深度
+    human_description:
+      en_US: Maximum depth to crawl relative to the entered URL. A maxDepth of 0 scrapes only the entered URL. A maxDepth of 1 scrapes the entered URL and all pages one level deep. A maxDepth of 2 scrapes the entered URL and all pages up to two levels deep. Higher values follow the same pattern.
+      zh_Hans: 相对于输入的URL，爬取的最大深度。maxDepth为0时，仅抓取输入的URL。maxDepth为1时，抓取输入的URL以及所有一级深层页面。maxDepth为2时，抓取输入的URL以及所有两级深层页面。更高值遵循相同模式。
+    form: form
+    min: 0
+  - name: mode
+    type: select
+    required: false
+    form: form
+    options:
+      - value: default
+        label:
+          en_US: default
+      - value: fast
+        label:
+          en_US: fast
+    default: default
+    label:
+      en_US: Crawl Mode
+      zh_Hans: 爬取模式
+    human_description:
+      en_US: The crawling mode to use. Fast mode crawls 4x faster websites without sitemap, but may not be as accurate and shouldn't be used in heavy js-rendered websites.
+      zh_Hans: 使用fast模式将不会使用其站点地图，比普通模式快4倍，但是可能不够准确，也不适用于大量js渲染的网站。
+  - name: ignoreSitemap
+    type: boolean
+    default: false
+    label:
+      en_US: ignore Sitemap
+      zh_Hans: 忽略站点地图
+    human_description:
+      en_US: Ignore the website sitemap when crawling.
+      zh_Hans: 爬取时忽略网站站点地图。
     form: form
-    default: 'blog/*'
   - name: limit
     type: number
     required: false
     label:
-      en_US: Maximum number of pages to crawl
+      en_US: Maximum pages to crawl
       zh_Hans: 最大爬取页面数
     human_description:
       en_US: Specify the maximum number of pages to crawl. The crawler will stop after reaching this limit.
       zh_Hans: 指定要爬取的最大页面数。爬虫将在达到此限制后停止。
     form: form
     min: 1
-    max: 20
     default: 5
+  - name: allowBackwardCrawling
+    type: boolean
+    default: false
+    label:
+      en_US: allow Backward Crawling
+      zh_Hans: 允许向后爬取
+    human_description:
+      en_US: Enables the crawler to navigate from a specific URL to previously linked pages. For instance, from 'example.com/product/123' back to 'example.com/product'
+      zh_Hans: 使爬虫能够从特定URL导航到之前链接的页面。例如，从'example.com/product/123'返回到'example.com/product'
+    form: form
+  - name: allowExternalContentLinks
+    type: boolean
+    default: false
+    label:
+      en_US: allow External Content Links
+      zh_Hans: 允许爬取外链
+    human_description:
+      en_US: Allows the crawler to follow links to external websites.
+      zh_Hans:
+    form: form
+############## Page Options #######################
+  - name: headers
+    type: string
+    label:
+      en_US: headers
+      zh_Hans: 请求头
+    human_description:
+      en_US: |
+        Headers to send with the request. Can be used to send cookies, user-agent, etc. Example: {"cookies": "testcookies"}
+      zh_Hans: |
+        随请求发送的头部。可以用来发送cookies、用户代理等。示例：{"cookies": "testcookies"}
+    placeholder:
+      en_US: Please enter an object that can be serialized in JSON
+      zh_Hans: 请输入可以json序列化的对象
+    form: form
+  - name: includeHtml
+    type: boolean
+    default: false
+    label:
+      en_US: include Html
+      zh_Hans: 包含HTML
+    human_description:
+      en_US: Include the HTML version of the content on page. Will output a html key in the response.
+      zh_Hans: 返回中包含一个HTML版本的内容，将以html键返回。
+    form: form
+  - name: includeRawHtml
+    type: boolean
+    default: false
+    label:
+      en_US: include Raw Html
+      zh_Hans: 包含原始HTML
+    human_description:
+      en_US: Include the raw HTML content of the page. Will output a rawHtml key in the response.
+      zh_Hans: 返回中包含一个原始HTML版本的内容，将以rawHtml键返回。
+    form: form
+  - name: onlyIncludeTags
+    type: string
+    label:
+      en_US: only Include Tags
+      zh_Hans: 仅抓取这些标签
+    placeholder:
+      en_US: Use commas to separate multiple tags
+      zh_Hans: 多个标签时使用半角逗号分隔
+    human_description:
+      en_US: |
+        Only include tags, classes and ids from the page in the final output. Use comma separated values. Example: script, .ad, #footer
+      zh_Hans: |
+        仅在最终输出中包含HTML页面的这些标签，可以通过标签名、类或ID来设定，使用逗号分隔值。示例：script, .ad, #footer
+    form: form
   - name: onlyMainContent
     type: boolean
-    required: false
+    default: false
     label:
-      en_US: Only return the main content of the page
-      zh_Hans: 仅返回页面的主要内容
+      en_US: only Main Content
+      zh_Hans: 仅抓取主要内容
     human_description:
-      en_US: If enabled, the crawler will only return the main content of the page, excluding headers, navigation, footers, etc.
-      zh_Hans: 如果启用，爬虫将仅返回页面的主要内容，不包括标题、导航、页脚等。
+      en_US: Only return the main content of the page excluding headers, navs, footers, etc.
+      zh_Hans: 只返回页面的主要内容，不包括头部、导航栏、尾部等。
+    form: form
+  - name: removeTags
+    type: string
+    label:
+      en_US: remove Tags
+      zh_Hans: 要移除这些标签
+    human_description:
+      en_US: |
+        Tags, classes and ids to remove from the page. Use comma separated values. Example: script, .ad, #footer
+      zh_Hans: |
+        要在最终输出中移除HTML页面的这些标签，可以通过标签名、类或ID来设定，使用逗号分隔值。示例：script, .ad, #footer
+    placeholder:
+      en_US: Use commas to separate multiple tags
+      zh_Hans: 多个标签时使用半角逗号分隔
+    form: form
+  - name: replaceAllPathsWithAbsolutePaths
+    type: boolean
+    default: false
+    label:
+      en_US: All AbsolutePaths
+      zh_Hans: 使用绝对路径
+    human_description:
+      en_US: Replace all relative paths with absolute paths for images and links.
+      zh_Hans: 将所有图片和链接的相对路径替换为绝对路径。
+    form: form
+  - name: screenshot
+    type: boolean
+    default: false
+    label:
+      en_US: screenshot
+      zh_Hans: 截图
+    human_description:
+      en_US: Include a screenshot of the top of the page that you are scraping.
+      zh_Hans: 提供正在抓取的页面的顶部的截图。
+    form: form
+  - name: waitFor
+    type: number
+    min: 0
+    label:
+      en_US: wait For
+      zh_Hans: 等待时间
+    human_description:
+      en_US: Wait x amount of milliseconds for the page to load to fetch content.
+      zh_Hans: 等待x毫秒以使页面加载并获取内容。
     form: form
-    options:
-      - value: 'true'
-        label:
-          en_US: 'Yes'
-          zh_Hans: 是
-      - value: 'false'
-        label:
-          en_US: 'No'
-          zh_Hans: 否
-    default: 'false'

api/core/tools/provider/builtin/firecrawl/tools/crawl_job.py

@@ -0,0 +1,20 @@
+from typing import Any
+
+from core.tools.entities.tool_entities import ToolInvokeMessage
+from core.tools.provider.builtin.firecrawl.firecrawl_appx import FirecrawlApp
+from core.tools.tool.builtin_tool import BuiltinTool
+
+
+class CrawlJobTool(BuiltinTool):
+    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> ToolInvokeMessage:
+        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'],
+                           base_url=self.runtime.credentials['base_url'])
+        operation = tool_parameters.get('operation', 'get')
+        if operation == 'get':
+            result = app.check_crawl_status(job_id=tool_parameters['job_id'])
+        elif operation == 'cancel':
+            result = app.cancel_crawl_job(job_id=tool_parameters['job_id'])
+        else:
+            raise ValueError(f'Invalid operation: {operation}')
+
+        return self.create_json_message(result)

api/core/tools/provider/builtin/firecrawl/tools/crawl_job.yaml

@@ -0,0 +1,37 @@
+identity:
+  name: crawl_job
+  author: hjlarry
+  label:
+    en_US: Crawl Job
+    zh_Hans: 爬取任务处理
+description:
+  human:
+    en_US: Retrieve the scraping results based on the job ID, or cancel the scraping task.
+    zh_Hans: 根据爬取任务ID获取爬取结果，或者取消爬取任务
+  llm: Retrieve the scraping results based on the job ID, or cancel the scraping task.
+parameters:
+  - name: job_id
+    type: string
+    required: true
+    label:
+      en_US: Job ID
+    human_description:
+      en_US: Set wait_for_results to false in the Crawl tool can get the job ID.
+      zh_Hans: 在深度爬取工具中将等待爬取结果设置为否可以获取Job ID。
+    llm_description: Set wait_for_results to false in the Crawl tool can get the job ID.
+    form: llm
+  - name: operation
+    type: select
+    required: true
+    options:
+      - value: get
+        label:
+          en_US: get crawl status
+      - value: cancel
+        label:
+          en_US: cancel crawl job
+    label:
+      en_US: operation
+      zh_Hans: 操作
+    llm_description: choose the operation to perform. `get` is for getting the crawl status, `cancel` is for cancelling the crawl job.
+    form: llm

api/core/tools/provider/builtin/firecrawl/tools/scrape.py

@@ -1,26 +1,39 @@
-import json
-from typing import Any, Union
+from typing import Any
 
 from core.tools.entities.tool_entities import ToolInvokeMessage
-from core.tools.provider.builtin.firecrawl.firecrawl_appx import FirecrawlApp
+from core.tools.provider.builtin.firecrawl.firecrawl_appx import FirecrawlApp, get_array_params, get_json_params
 from core.tools.tool.builtin_tool import BuiltinTool
 
 
 class ScrapeTool(BuiltinTool):
-    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]:
-        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'], base_url=self.runtime.credentials['base_url'])
 
-        crawl_result = app.scrape_url(
-            url=tool_parameters['url'],
-            wait=True
-        )
+    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> ToolInvokeMessage:
+        """
+        the pageOptions and extractorOptions comes from doc here:
+        https://docs.firecrawl.dev/api-reference/endpoint/scrape
+        """
+        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'],
+                           base_url=self.runtime.credentials['base_url'])
 
-        if isinstance(crawl_result, dict):
-            result_message = json.dumps(crawl_result, ensure_ascii=False, indent=4)
-        else:
-            result_message = str(crawl_result)
+        pageOptions = {}
+        extractorOptions = {}
 
-        if not crawl_result:
-            return self.create_text_message("Scrape request failed.")
+        pageOptions['headers'] = get_json_params(tool_parameters, 'headers')
+        pageOptions['includeHtml'] = tool_parameters.get('includeHtml', False)
+        pageOptions['includeRawHtml'] = tool_parameters.get('includeRawHtml', False)
+        pageOptions['onlyIncludeTags'] = get_array_params(tool_parameters, 'onlyIncludeTags')
+        pageOptions['removeTags'] = get_array_params(tool_parameters, 'removeTags')
+        pageOptions['onlyMainContent'] = tool_parameters.get('onlyMainContent', False)
+        pageOptions['replaceAllPathsWithAbsolutePaths'] = tool_parameters.get('replaceAllPathsWithAbsolutePaths', False)
+        pageOptions['screenshot'] = tool_parameters.get('screenshot', False)
+        pageOptions['waitFor'] = tool_parameters.get('waitFor', 0)
 
-        return self.create_text_message(result_message)
+        extractorOptions['mode'] = tool_parameters.get('mode', '')
+        extractorOptions['extractionPrompt'] = tool_parameters.get('extractionPrompt', '')
+        extractorOptions['extractionSchema'] = get_json_params(tool_parameters, 'extractionSchema')
+
+        crawl_result = app.scrape_url(url=tool_parameters['url'],
+                                      pageOptions=pageOptions,
+                                      extractorOptions=extractorOptions)
+
+        return self.create_json_message(crawl_result)

api/core/tools/provider/builtin/firecrawl/tools/scrape.yaml

@@ -3,7 +3,7 @@ identity:
   author: ahasasjeb
   label:
     en_US: Scrape
-    zh_Hans: 抓取
+    zh_Hans: 单页面抓取
 description:
   human:
     en_US: Extract data from a single URL.
@@ -21,3 +21,160 @@ parameters:
       zh_Hans: 要抓取并提取数据的网站URL。
     llm_description: The URL of the website that needs to be crawled. This is a required parameter.
     form: llm
+############## Page Options #######################
+  - name: headers
+    type: string
+    label:
+      en_US: headers
+      zh_Hans: 请求头
+    human_description:
+      en_US: |
+        Headers to send with the request. Can be used to send cookies, user-agent, etc. Example: {"cookies": "testcookies"}
+      zh_Hans: |
+        随请求发送的头部。可以用来发送cookies、用户代理等。示例：{"cookies": "testcookies"}
+    placeholder:
+      en_US: Please enter an object that can be serialized in JSON
+      zh_Hans: 请输入可以json序列化的对象
+    form: form
+  - name: includeHtml
+    type: boolean
+    default: false
+    label:
+      en_US: include Html
+      zh_Hans: 包含HTML
+    human_description:
+      en_US: Include the HTML version of the content on page. Will output a html key in the response.
+      zh_Hans: 返回中包含一个HTML版本的内容，将以html键返回。
+    form: form
+  - name: includeRawHtml
+    type: boolean
+    default: false
+    label:
+      en_US: include Raw Html
+      zh_Hans: 包含原始HTML
+    human_description:
+      en_US: Include the raw HTML content of the page. Will output a rawHtml key in the response.
+      zh_Hans: 返回中包含一个原始HTML版本的内容，将以rawHtml键返回。
+    form: form
+  - name: onlyIncludeTags
+    type: string
+    label:
+      en_US: only Include Tags
+      zh_Hans: 仅抓取这些标签
+    placeholder:
+      en_US: Use commas to separate multiple tags
+      zh_Hans: 多个标签时使用半角逗号分隔
+    human_description:
+      en_US: |
+        Only include tags, classes and ids from the page in the final output. Use comma separated values. Example: script, .ad, #footer
+      zh_Hans: |
+        仅在最终输出中包含HTML页面的这些标签，可以通过标签名、类或ID来设定，使用逗号分隔值。示例：script, .ad, #footer
+    form: form
+  - name: onlyMainContent
+    type: boolean
+    default: false
+    label:
+      en_US: only Main Content
+      zh_Hans: 仅抓取主要内容
+    human_description:
+      en_US: Only return the main content of the page excluding headers, navs, footers, etc.
+      zh_Hans: 只返回页面的主要内容，不包括头部、导航栏、尾部等。
+    form: form
+  - name: removeTags
+    type: string
+    label:
+      en_US: remove Tags
+      zh_Hans: 要移除这些标签
+    human_description:
+      en_US: |
+        Tags, classes and ids to remove from the page. Use comma separated values. Example: script, .ad, #footer
+      zh_Hans: |
+        要在最终输出中移除HTML页面的这些标签，可以通过标签名、类或ID来设定，使用逗号分隔值。示例：script, .ad, #footer
+    placeholder:
+      en_US: Use commas to separate multiple tags
+      zh_Hans: 多个标签时使用半角逗号分隔
+    form: form
+  - name: replaceAllPathsWithAbsolutePaths
+    type: boolean
+    default: false
+    label:
+      en_US: All AbsolutePaths
+      zh_Hans: 使用绝对路径
+    human_description:
+      en_US: Replace all relative paths with absolute paths for images and links.
+      zh_Hans: 将所有图片和链接的相对路径替换为绝对路径。
+    form: form
+  - name: screenshot
+    type: boolean
+    default: false
+    label:
+      en_US: screenshot
+      zh_Hans: 截图
+    human_description:
+      en_US: Include a screenshot of the top of the page that you are scraping.
+      zh_Hans: 提供正在抓取的页面的顶部的截图。
+    form: form
+  - name: waitFor
+    type: number
+    min: 0
+    label:
+      en_US: wait For
+      zh_Hans: 等待时间
+    human_description:
+      en_US: Wait x amount of milliseconds for the page to load to fetch content.
+      zh_Hans: 等待x毫秒以使页面加载并获取内容。
+    form: form
+############## Extractor Options #######################
+  - name: mode
+    type: select
+    options:
+      - value: markdown
+        label:
+          en_US: markdown
+      - value: llm-extraction
+        label:
+          en_US: llm-extraction
+      - value: llm-extraction-from-raw-html
+        label:
+          en_US: llm-extraction-from-raw-html
+      - value: llm-extraction-from-markdown
+        label:
+          en_US: llm-extraction-from-markdown
+    label:
+      en_US: Extractor Mode
+      zh_Hans: 提取模式
+    human_description:
+      en_US: |
+       The extraction mode to use. 'markdown': Returns the scraped markdown content, does not perform LLM extraction. 'llm-extraction': Extracts information from the cleaned and parsed content using LLM.
+      zh_Hans: 使用的提取模式。“markdown”：返回抓取的markdown内容，不执行LLM提取。“llm-extractioin”：使用LLM按Extractor Schema从内容中提取信息。
+    form: form
+  - name: extractionPrompt
+    type: string
+    label:
+      en_US: Extractor Prompt
+      zh_Hans: 提取时的提示词
+    human_description:
+      en_US: A prompt describing what information to extract from the page, applicable for LLM extraction modes.
+      zh_Hans: 当使用LLM提取模式时，用于给LLM描述提取规则。
+    form: form
+  - name: extractionSchema
+    type: string
+    label:
+      en_US: Extractor Schema
+      zh_Hans: 提取时的结构
+    placeholder:
+      en_US: Please enter an object that can be serialized in JSON
+    human_description:
+      en_US: |
+        The schema for the data to be extracted, required only for LLM extraction modes. Example: {
+            "type": "object",
+            "properties": {"company_mission": {"type": "string"}},
+            "required": ["company_mission"]
+        }
+      zh_Hans: |
+       当使用LLM提取模式时，使用该结构去提取，示例：{
+            "type": "object",
+            "properties": {"company_mission": {"type": "string"}},
+            "required": ["company_mission"]
+        }
+    form: form

api/core/tools/provider/builtin/firecrawl/tools/search.py

@@ -1,5 +1,4 @@
-import json
-from typing import Any, Union
+from typing import Any
 
 from core.tools.entities.tool_entities import ToolInvokeMessage
 from core.tools.provider.builtin.firecrawl.firecrawl_appx import FirecrawlApp
@@ -7,20 +6,23 @@ from core.tools.tool.builtin_tool import BuiltinTool
 
 
 class SearchTool(BuiltinTool):
-    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]:
-        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'], base_url=self.runtime.credentials['base_url'])
-
-        crawl_result = app.search(
+    def _invoke(self, user_id: str, tool_parameters: dict[str, Any]) -> ToolInvokeMessage:
+        """
+        the pageOptions and searchOptions comes from doc here:
+        https://docs.firecrawl.dev/api-reference/endpoint/search
+        """
+        app = FirecrawlApp(api_key=self.runtime.credentials['firecrawl_api_key'],
+                           base_url=self.runtime.credentials['base_url'])
+        pageOptions = {}
+        pageOptions['onlyMainContent'] = tool_parameters.get('onlyMainContent', False)
+        pageOptions['fetchPageContent'] = tool_parameters.get('fetchPageContent', True)
+        pageOptions['includeHtml'] = tool_parameters.get('includeHtml', False)
+        pageOptions['includeRawHtml'] = tool_parameters.get('includeRawHtml', False)
+        searchOptions = {'limit': tool_parameters.get('limit')}
+        search_result = app.search(
             query=tool_parameters['keyword'],
-            wait=True
+            pageOptions=pageOptions,
+            searchOptions=searchOptions
         )
 
-        if isinstance(crawl_result, dict):
-            result_message = json.dumps(crawl_result, ensure_ascii=False, indent=4)
-        else:
-            result_message = str(crawl_result)
-
-        if not crawl_result:
-            return self.create_text_message("Search request failed.")
-
-        return self.create_text_message(result_message)
+        return self.create_json_message(search_result)

api/core/tools/provider/builtin/firecrawl/tools/search.yaml

@@ -21,3 +21,55 @@ parameters:
       zh_Hans: 输入关键词即可使用Firecrawl API进行搜索。
     llm_description: Efficiently extract keywords from user text.
     form: llm
+############## Page Options #######################
+  - name: onlyMainContent
+    type: boolean
+    default: false
+    label:
+      en_US: only Main Content
+      zh_Hans: 仅抓取主要内容
+    human_description:
+      en_US: Only return the main content of the page excluding headers, navs, footers, etc.
+      zh_Hans: 只返回页面的主要内容，不包括头部、导航栏、尾部等。
+    form: form
+  - name: fetchPageContent
+    type: boolean
+    default: true
+    label:
+      en_US: fetch Page Content
+      zh_Hans: 抓取页面内容
+    human_description:
+      en_US: Fetch the content of each page. If false, defaults to a basic fast serp API.
+      zh_Hans: 获取每个页面的内容。如果为否，则使用基本的快速搜索结果页面API。
+    form: form
+  - name: includeHtml
+    type: boolean
+    default: false
+    label:
+      en_US: include Html
+      zh_Hans: 包含HTML
+    human_description:
+      en_US: Include the HTML version of the content on page. Will output a html key in the response.
+      zh_Hans: 返回中包含一个HTML版本的内容，将以html键返回。
+    form: form
+  - name: includeRawHtml
+    type: boolean
+    default: false
+    label:
+      en_US: include Raw Html
+      zh_Hans: 包含原始HTML
+    human_description:
+      en_US: Include the raw HTML content of the page. Will output a rawHtml key in the response.
+      zh_Hans: 返回中包含一个原始HTML版本的内容，将以rawHtml键返回。
+    form: form
+############## Search Options #######################
+  - name: limit
+    type: number
+    min: 0
+    label:
+      en_US: Maximum results
+      zh_Hans: 最大结果数量
+    human_description:
+      en_US: Maximum number of results. Max is 20 during beta.
+      zh_Hans: 最大结果数量。在测试阶段，最大为20。
+    form: form

api/core/tools/provider/builtin/jina/tools/jina_reader.py

@@ -60,11 +60,13 @@ class JinaReaderTool(BuiltinTool):
         if tool_parameters.get('no_cache', False):
             headers['X-No-Cache'] = 'true'
 
+        max_retries = tool_parameters.get('max_retries', 3)
         response = ssrf_proxy.get(
             str(URL(self._jina_reader_endpoint + url)),
             headers=headers,
             params=request_params,
             timeout=(10, 60),
+            max_retries=max_retries
         )
 
         if tool_parameters.get('summary', False):

api/core/tools/provider/builtin/jina/tools/jina_reader.yaml

@@ -150,3 +150,17 @@ parameters:
       pt_BR: Habilitar resumo para a saída
     llm_description: enable summary
     form: form
+  - name: max_retries
+    type: number
+    required: false
+    default: 3
+    label:
+      en_US: Retry
+      zh_Hans: 重试
+      pt_BR: Repetir
+    human_description:
+      en_US: Number of times to retry the request if it fails
+      zh_Hans: 请求失败时重试的次数
+      pt_BR: Número de vezes para repetir a solicitação se falhar
+    llm_description: Number of times to retry the request if it fails
+    form: form