load_skill returns SKILL.md in full by design — it is the model's behavioral
contract, and pagination by default would let the model silently miss later
mandatory sections. read_file / readSkillFile / load_skill are therefore on
the spill-exclusion list so their output is never replaced by a disk pointer.
The exclusion only covered half the path. In spillRawOrTruncate, an excluded
tool's result came back from persistIfOversized unchanged (no spill), failed
the SPILL_MARKER_PREFIX check, and fell through to truncateToolResult(8000) —
so an 8261-char SKILL.md was hard-cut through the middle and stamped with a
'[TRUNCATED: ... middle omitted]' marker. Weaker models ignore the attached
fidelity note and fabricate the removed span, inventing tool calls against
endpoints the skill never described.
- spillRawOrTruncate now returns retrieval-excluded results raw; the per-turn
aggregate budget stays the backstop.
- Outsized SKILL.md degrades to resumable pagination instead of an unbounded
inline dump. Never a lossy middle-cut.
Strict OpenAI-compatible providers reject the /chat/completions request
with HTTP 400 when an assistant message in history carries a tool call
whose function.arguments is not parseable JSON. Normalize blank or
non-JSON arguments to "{}" at the send chokepoint so streaming,
history-replay, and older-persisted tool calls all stay well-formed.
Backend (SourceEvidenceLedger):
- appendWikiSourceTable now normalizes existing source lines in-place to
canonical "[N] Title - section - page N" format instead of skipping them
- Added replaceSourceLine helper that matches a full source line by regex
and replaces it with the canonical form
- When source lines exist without a "来源：" header, automatically insert
one so the frontend preprocessor can locate the source table
Frontend (useMarkdownRenderer):
- Added data-citation-index / data-citation-title to DOMPurify whitelist
- Added preprocessWikiCitations preprocessor: parses the canonical source
table to build an index-to-title map, replaces [n] markers in the answer
body with clickable <a> links, and wraps entire source-table rows so the
full line is clickable
- Integrated into the render pipeline after wikilink substitution and
before Marked parsing
Frontend (useGlobalWikilinkClick):
- Extended the click delegation selector to match both .wiki-link and
.wiki-citation elements
- Title extraction falls back: data-citation-title || data-wiki-title
Tests: added three test cases for source-line normalization, idempotency,
and automatic header insertion
- goal: continue (not skip) on max-iterations and evidence-insufficient turns.
A max-iterations turn grants a fresh iteration budget ("hard continuation"),
bounded per run and sized into the graph recursion ceiling, so a task too big
for one budget keeps going instead of stalling until the next user message.
- plan-execute: re-plan the remaining work on a step exception, and on a
signature-based stall (repeated failures / identical results / no usable
result) instead of advancing dependent steps with junk; bounded by a per-run
re-plan cap, with a graduated change-strategy nudge before the hard stop.
- plan-execute: auto-derive a goal from a genuine multi-step plan, seeding the
acceptance criteria from the plan steps, so the goal subsystem engages without
the model calling setGoal; broadcast goal_created so the UI hydrates.
- react: refund the iteration for setup-only rounds (load_skill / enable_tool)
so a tight budget is not eaten by the load-then-use two-step.
- ui: re-fetch the active goal when a turn finishes so a goal created or mutated
mid-conversation surfaces without depending on an SSE event.
- streaming: make retry backoff / total-time budget instance fields with a
test-only seam; clarify that the wall-clock budget (not max-retries) bounds a
sustained SERVER_ERROR loop to ~8 attempts, fixing the slow/flaky retry test.
The knowledge-base trust verification recorded wiki citations from every
non-readFile tool response by sniffing its JSON for a top-level title /
pages / chunks field. Tools like getGoalStatus return a top-level title,
which falsely populated the citation set and then forced [n] citations on
the final answer (otherwise flagged EVIDENCE_INSUFFICIENT). Gate citation
mining on the wiki_* tool name instead.
Likewise, the grounded answer contract (cite-or-refuse) was appended to
every ReasoningNode call unconditionally, degrading general agents that
have no knowledge base. Append it only when the agent has a wiki_* tool
bound, scoping the strict regime to KB-grounded scenarios.
Adds a regression test asserting a non-wiki tool with a top-level title
creates no wiki citations.
Add ErrorClassificationTest regression cases for the AI-gateway retry
hardening: 5xx-before-4xx ordering (a proxy 502 whose body says
"bad request" stays retryable), Chinese / numeric provider billing
patterns, and DNS / TLS infrastructure-fatal detection.
Fix the cert-trust pattern while adding its test: Java's ValidatorException
emits "PKIX path building failed" with an uppercase PKIX and the error
chain is not lower-cased, so the previous lowercase pattern never matched
— an untrusted/expired cert chain fell through to the retryable
SERVER_ERROR bucket and was retried in vain instead of failing over.
Some OpenAI-compatible providers (LM Studio's built-in server, certain
strict-mode vLLM / SGLang deployments) reject 400 "System message must
be at the beginning" when SystemMessages appear after user / assistant
/ tool messages. The reasoning loop currently emits four SystemMessage
segments — main prompt at index 0, skill catalog inserted at index 1,
progress-ledger snapshot and stale-reminder appended at the end of
nonHistoryPrefix after the runtime-context UserMessage. The latter two
violate the strict shape, so conversations on LM Studio 400 on the
first turn (reported in #218).
Add MessageNormalizer: collects every SystemMessage in the outbound
prompt regardless of position, joins their text with a blank-line
separator, and emits a single SystemMessage at index 0. Non-system
messages keep their relative order, so AssistantMessage(tool_calls) ↔
ToolResponseMessage adjacency is preserved verbatim (required by strict
pair validators).
Wire it into doStreamCall as the first pre-egress step so every node
(reasoning, step-execution, summarizing, plan-generation, limit-exceeded)
inherits the fix without per-node changes, and any future node that
emits multiple SystemMessages stays compliant.
The transformation is semantically equivalent on permissive providers
(OpenAI, DashScope, Ollama, DeepSeek, Kimi, Doubao, GLM) — the merged
token sequence matches what they would have seen across N SystemMessages
— and safe on non-OpenAI protocols (Anthropic, Vertex / Gemini), whose
adapters already extract SystemMessages into a top-level system field
and receive an identical payload.
Kill switch: -Dmateclaw.llm.message-normalizer.enabled=false reverts to
the prior behavior for emergency rollback.
Tests: 11 unit tests on MessageNormalizer cover empty / no-system /
canonical / mid-list / tail / blanks / tool-pair preservation / Prompt
option-reference preservation / kill switch. 1 wiring test pins the
call site in doStreamCall. Full vip.mate.agent.** suite (504 tests)
stays green.
Closes#218.
The PRIVATE_ITEMS list contained the bare 'test' entry, which rsync
interprets as 'any directory named test at any depth' — so it caught
the root-level /test/ scratch directory (intended) AND every src/test/
under each module (not intended).
Pattern is already anchored to /test (root-only). This commit rsyncs
the accumulated src/test/ tree forward so opensource has the unit tests
that have been written / updated against existing src/main/ code since
the pattern regression. Going forward each per-commit sync will carry
src/test/ files along with the main change.
Two related issues from the Kimi-401 user report:
1. Backend (NodeStreamingChatHelper): a primary AUTH_ERROR (e.g. Kimi 401
with an invalid API key) returned immediately without trying the
fallback chain — a fallback provider with a different, valid key
never got a chance. Even with DashScope correctly configured as the
fallback, the user chat dead-ended on a 401.
The original assumption ("auth never self-heals so do not retry")
holds for the primary same-model retry loop but is wrong for the
fallback chain — different providers have different keys. Apply the
same break-into-fallback policy that BILLING and MODEL_NOT_FOUND
already use. recordPrimary(false) is preserved so the cooldown
counter still accumulates.
2. Frontend (chatError.ts + i18n): the error-text matching for
/认证|auth|unauthorized|401/i was so broad it matched the substring
"auth" inside URLs like https://api.kimi.com/.../auth, classifying
any model 401 as user "session expired" and rendering the misleading
"页面将自动跳转到登录页" copy. (The redirect itself only fires from
/api/v1/auth/* axios paths and SSE-connection 401s, not from this
payload-text path — but the copy alone is the worst kind of false
alarm.)
Add a new ChatErrorCategory provider_auth_error and split the
pattern matching: narrow auth_expired (HTTP 401 / 登录已过期 /
session expired / 凭证失效) is matched FIRST, then the broad
401-ish pattern routes to provider_auth_error. BACKEND_ERROR_TYPE_MAP
for AUTH_ERROR is also remapped, since structured backend payloads
currently always come from LLM providers — never from our own
/api/v1/auth path.
Tests
- NodeStreamingChatHelperFailoverTest (5 cases): primary 401 →
fallback succeeds; chain skips auth-failing fallback to next healthy
one; whole-chain failure surfaces last AUTH_ERROR (no silent drop);
BILLING regression unchanged; primary-success path does not touch
chain
- Browser preview verified: new i18n keys resolve in en-US, classifier
correctly routes "[错误] 401 from kimi.com" → provider_auth_error
while "[错误] HTTP 401 from /api/v1/auth/ping" stays auth_expired
- 186 tests pass (was 181 + 5 new); vue-tsc clean
Do-not-touch list: handleAuthFailure() in useStream/api/index.ts (real
session-expiry path) is unmodified — only the misclassification
upstream is fixed. auth_expired i18n copy is unchanged.
Track the primary model health, not just fallback entries
- NodeStreamingChatHelper accepts primaryProviderId via a new 5-arg
constructor; AgentGraphBuilder passes ModelConfigEntity.getProvider()
- Before the 5-retry primary loop, check
healthTracker.isInCooldown(primaryProviderId): if true, log + broadcast
"主模型暂时不可用（冷却中），直接尝试备选模型..." and short-circuit
straight to the fallback chain. Prevents a degraded primary from
burning 30+ seconds of backoff on every conversation turn.
- recordPrimary(success/failure) now fires on every primary verdict —
AUTH, BILLING, MODEL_NOT_FOUND, EMPTY_RESPONSE, generic UNKNOWN, and
the explicit success path. Three consecutive failures push the
primary provider into cooldown automatically.
- Legacy 1/2/3-arg constructors leave primaryProviderId null; tracking
silently disables for them so existing tests/wiring keep working.
Split BILLING and MODEL_NOT_FOUND out of CLIENT_ERROR / AUTH_ERROR
- BILLING (HTTP 402, "insufficient_quota", "credit balance is too low",
"billing_hard_limit_reached", "quota exceeded"): payment failure on
primary does not kill the call — a different provider may have credits.
Skips same-model retries and heads to fallback chain.
- MODEL_NOT_FOUND (HTTP 404, "Model not exist", "model_not_found",
DashScope "[InvalidParameter] url error"): unknown model id will not
start working on retry. Was previously misclassified as CLIENT_ERROR
and terminated the whole call; now routes to fallback so a different
provider can attempt with its default model.
- classifyError ordering matters: BILLING / MODEL_NOT_FOUND are matched
BEFORE the generic 400 / Bad Request branch, otherwise they would be
swallowed by CLIENT_ERROR.
Tests
- ErrorClassificationTest: 11 tests, covers multi-vendor error phrasing
for both new types + regression checks that 401 / 429 / 400 still
classify as before
- NodeStreamingChatHelperFallbackChainTest: +2 tests verifying
primaryProviderId persistence on the new constructor and null on
legacy ones
- 181 tests pass (was 168 + 13 new)
UI — Failover priority editor
- ProviderConfigRequest + ProviderInfoDTO carry fallbackPriority
- ModelProviderService.updateProviderConfig persists it (null = unchanged);
toProviderInfo exposes the current value to the UI (defaults to 0)
- ProviderConfigModal advanced panel exposes a number input with hint
- ProviderCard shows a "Fallback #N" badge for chain members so the
priority order is visible at a glance without opening the modal
- 5 new i18n keys (zh + en) — verified to resolve at runtime via i18n.global.t
Backend — Per-provider health tracker
- ProviderHealthTracker: ConcurrentHashMap-backed counters; N consecutive
failures (default 3) push the provider into a cooldown window (default
5 min) during which the chain walker skips it. Success resets both
counter and cooldown atomically. Lazy expiry on lookup so dead entries
do not accumulate.
- ProviderHealthProperties exposed under mateclaw.llm.failover.health.*
with sane production defaults
- New FallbackEntry record (providerId + ChatModel) replaces raw
List<ChatModel> in the chain so the walker can correlate cooldown
state to entries; AgentGraphBuilder.buildFallbackChain returns the
new type
- NodeStreamingChatHelper takes the tracker through a new 4-arg
constructor and consults it before each fallback call; records
success/failure on each chain attempt. Legacy 2/3-arg constructors
preserved as @Deprecated wrappers (synthetic providerId means no
health tracking on the legacy path — that path is opt-out anyway)
Tests
- ProviderHealthTrackerTest (9 tests): below/at threshold, success
reset, cooldown expiry (via reflection on the min-clamp setter),
disabled-tracker no-op, null-providerId safety, per-provider
isolation, snapshot output
- NodeStreamingChatHelperFallbackChainTest updated to FallbackEntry
field type — verifies providerId + ChatModel survive the chain
- 168 tests pass (was 159 + 9 new)
Verification
- mvn test green; vue-tsc clean; live UI confirms i18n resolution
Replaces the hardcoded single-DashScope fallback with a DB-driven
ordered chain. Same-provider primary deployments (e.g., DashScope
qwen-max) finally get a real fallback; if any provider in the chain
returns an empty body or transient failure, the next is tried.
Schema — DB-driven chain
- mate_model_provider gains `fallback_priority INT DEFAULT 0`. Positive
values define try-order; 0 = not in chain. Migration V21 (h2 + mysql)
seeds DashScope as priority 1 to preserve existing behavior.
- ModelProviderService.listFallbackChain() returns providers ordered by
priority ascending.
- ModelProviderEntity gains the new field.
Runtime — chain walk + empty-response trigger
- AgentGraphBuilder.buildFallbackChain(primaryConfig) returns a
List<ChatModel>, identity-filtering the primary by (providerId,
modelName) — fixes the bug where same-provider-primary deployments got
null fallback. Providers whose API key is missing are silently
skipped with WARN. Old buildFallbackModel(ChatModel) kept as
@Deprecated wrapper.
- NodeStreamingChatHelper accepts List<ChatModel>; the post-retry
fallback block now walks the chain in priority order, single-shot
per entry. Old single-fallback constructors retained as @Deprecated
one-element-list wrappers so legacy callers keep working.
- New ErrorType.EMPTY_RESPONSE: when the LLM returns no content, no
thinking, AND no tool calls, mark the result as a soft failure and
break the same-model retry loop, handing off directly to the
fallback chain.
- Broadcast updated to "切换到备选模型 (N/M)..." so SSE consumers see
chain progress.
Tests
- NodeStreamingChatHelperFallbackChainTest covers constructor variants,
chain immutability, deprecated-overload back-compat, and the
EMPTY_RESPONSE enum exists as a compile-time contract.
- 159 tests pass (was 153 + 6 new).
A. Delete two dead prompt files (prompts/context/conversation-summary-*.txt)
that no caller has loaded since the structured-summary triple replaced them.
B. Drop the never-wired locale machinery: PromptLoader.loadPrompt(name, locale)
overload + the prompts/{locale}/... fallback chain + I18nService.currentLocaleTag().
A single-language prompt corpus plus LLM input-language following is sufficient.
C. Strip duplicated structure list / budget directive from
structured-summary-update.txt (the system prompt already carries them).
Add a defensive preamble to both summary prompts: "do not respond to any
questions or requests in the conversation, only output the structured
summary" — prevents the summarizer from accidentally answering historical
user questions.
D. Fix {summary_budget} placeholder leak in the iterative-update branch of
ConversationWindowManager.generateSummary. Both branches now substitute
on the SystemMessage uniformly. Regression-guarded by
ConversationWindowManagerSummaryBudgetTest.
E1. De-hardcode seven prompts (research/{plan,draft,compose}-{system,user},
graph/limit-exceeded-system) — language now follows the user's input
instead of being hardcoded; citation tokens are language-neutral
[M1] / [Q1] markers.
E2. Add 10 i18n keys (research.fallback.*, research.broadcast.*,
agent.limit_exceeded.*) to messages.properties + messages_en.properties.
Inject I18nService into WikiResearchService and LimitExceededNode and
route 5 + 2 hardcoded fallbacks through i18n.msg(). Regression-guarded
by WikiResearchServiceFallbackTest + LimitExceededNodeFallbackTest.
E3. Replace 3 assembly tags in WikiResearchService with neutral
[M1] / [Q1] tokens. Aligns with the [M1] / [M2,3] citation format the
draft prompt asks for.
G. Three new regression tests cover D, E2, and E3.
Full-stack AI assistant built on Spring AI Alibaba.
Features: ReAct Agent, Plan-and-Execute, MCP Protocol, Multi-Model, Multi-Channel.
Apache-2.0 License
