Split-out PR B from #3584. Normalizes Antigravity/agy provider quotas: prefers retrieveUserQuota for live consumption, falls back to fetchAvailableModels and local usage_history, sanitizes cached Provider Limits so retired upstream IDs are not re-exposed, and schedules a deduplicated post-usage refresh. Maintainer follow-up: decoupled the post-usage refresh via a lightweight usageEvents bus (usageHistory no longer dynamic-imports providerLimits) so it does not pull the executors/translator graph into the typecheck-core surface — typecheck:core stays at 0. Integrated into release/v3.8.21. Thanks @dhaern!
Split-out PR A from #3584. Normalizes the Antigravity/agy Gemini 3.5 Flash tier IDs to clean public names (gemini-3.5-flash-low/medium/high), maps them to the live upstream IDs at the executor boundary, and removes Antigravity from the global model resolver so the executor owns wire normalization. Maintainer follow-up: kept gemini-3.5-flash-preview as a hidden backward-compat alias routing to the High tier (so saved combos/configs keep working). Live-validated the tier set via the agy CLI catalog. Integrated into release/v3.8.21. Thanks @dhaern!
Split-out PR C from #3584. Isolates textual reasoning wrappers (<think>/<thinking>/<thought>/<internal_thought>, including malformed/open tags) into reasoning_content across both the non-streaming sanitizer and the Gemini streaming translator, with split-chunk buffering. Additive to the existing textual tool-call pipeline; does not touch the #3569 native functionResponse path. Integrated into release/v3.8.21. Thanks @dhaern!
* chore(release): open v3.8.20 development cycle
* fix(images): prefer bare combos over image aliases (#3527)
Integrated into release/v3.8.20
* fix(translator): map Codex local_shell tool (#3534)
Integrated into release/v3.8.20
* fix(usage): make opencode-go quota fetcher fail-open instead of throwing 500 (#3522)
Integrated into release/v3.8.20
* Fix Runtime page breaker state rendering (#3533)
Integrated into release/v3.8.20
* Expose provider breaker degradation threshold setting (#3535)
Integrated into release/v3.8.20
* fix(executor): strip provider prefix from versioned built-in tool model field (#3532)
Integrated into release/v3.8.20
* feat(providers): add Claude Fable 5 support (#3524)
Integrated into release/v3.8.20
* feat(resilience): add global provider cooldown tracking to prevent combo re-walking (#3556)
Integrated into release/v3.8.20 (default OFF, opt-in)
* fix(translator): scope thoughtSignature bypass to Antigravity/CLI only (#3560)
Integrated into release/v3.8.20. Co-authored-by: Six7Day <six7day@gmail.com>
* fix(routing): normalize thinking:disabled for combo-substituted models that reject it (#3554) (#3563)
Integrated into release/v3.8.20
* fix(usage): accept 0/empty budget limits so the dashboard can save and clear (#3537) (#3564)
Integrated into release/v3.8.20
* docs(changelog): credit @Six7Day for #3560 thoughtSignature fix (#3414)
The #3560 squash co-author trailer landed inline (unparsed by GitHub), so add
an explicit CHANGELOG credit ensuring @Six7Day (original #3414) and @oyi77 are
on the public record for the Gemini thoughtSignature fix.
* fix(gamification): dedup badge unlock via user_badges so events don't re-fire every request (#3472) (#3565)
Integrated into release/v3.8.20
* fix(routing): pass through 'auto' keyword on codex /v1/responses instead of rewriting to codex/auto (#3509) (#3566)
Integrated into release/v3.8.20
* fix(cli-tools): normalize apiKey null in guide-settings schema so cloud-mode config saves (#3552) (#3567)
Integrated into release/v3.8.20
* fix(catalog): reclassify PublicAI from keyless to one-time-initial (requires API key) (#3558) (#3568)
Integrated into release/v3.8.20
* fix(gemini-web): surface missing Playwright browser as actionable 503 + cooldown hint, not a retryable 500 loop (#3516) (#3570)
Integrated into release/v3.8.20
* fix(security): sanitize raw err.message in web executors + embeddings/search response bodies (Rule #12) (#3494, #3495) (#3573)
Integrated into release/v3.8.20
* fix(dashboard): point CustomHostsManager + FeatureFlagsGrid at real routes (#3486, #3487) (#3574)
Integrated into release/v3.8.20
* chore(providers): remove dead krutrim entry (#3483) + docs(api): fix agent-bridge per-agent state route (#3489) (#3575)
Integrated into release/v3.8.20
* docs(api): correct API_REFERENCE.md paths for skills/plugins/admin/cache/acp/system-info (#3497) (#3577)
Integrated into release/v3.8.20
* fix(proxy): drive SOCKS5 UI option from runtime ENABLE_SOCKS5_PROXY, not build-time NEXT_PUBLIC (#3508) (#3579)
Integrated into release/v3.8.20
* fix(playground): filter playground models by node prefix so custom-endpoint models appear (#3505) (#3581)
Integrated into release/v3.8.20
* fix(usage): show an informative message instead of a blank Kiro quota card when no usage breakdown (#3506) (#3582)
Integrated into release/v3.8.20
* docs(changelog): add the #3506 Kiro quota entry (missed in #3582 due to a stale-base CHANGELOG anchor) (#3583)
Integrated into release/v3.8.20
* fix(auto-update): use stable PROJECT_ROOT walker, not frozen process.cwd() (#3561)
Integrated into release/v3.8.20. Auto-update PROJECT_ROOT now uses a stable __dirname-anchored upward walker instead of the no-op process.cwd() resolver.
* fix: address PR #3518 review comments (lifecycle hooks, regex, indentation, route params) (#3562)
Integrated into release/v3.8.20. Addresses #3518 review: regex literals, logs/[id] route params (Next 16), indentation, and wires plugin lifecycle hooks (onInstall/onActivate/onDeactivate/onUninstall) in the loader so manager.ts can register them. Adds Rule #18 regression test.
* docs(changelog): credit @ViFigueiredo (#3423) for PROJECT_ROOT + log #3561/#3562 (v3.8.20)
* fix: openai to gemini incorrectly translates historical tool calls into text (#3569)
Integrated into release/v3.8.20. Standard Gemini direct path now maps historical tool calls to native functionCall/functionResponse parts (signaturelessToolCallMode: native) instead of inert text — validated against the real Gemini API (gemini-2.5-flash returns 200 for signatureless native functionCall, even with tools+thinking; Hard Rule #18). Eliminates the text-serialization leak. Antigravity/CLI sentinel path (#3560) untouched.
* docs(changelog)+test: reconcile standard-Gemini native mode (#3569) — update round-2 rationale comment + log VPS validation
* docs(changelog): reconcile v3.8.20 — add 9 missing bullets + move [Unreleased] to versioned section
* docs(changelog): complete v3.8.20 reconciliation — 27 bullets, 11 contributors
---------
Co-authored-by: Alexander Averyanov <alex@averyan.ru>
Co-authored-by: Hakan Kurşun <bykamaka@gmail.com>
Co-authored-by: Wilson <pedbookmed@gmail.com>
Co-authored-by: Randi <55005611+rdself@users.noreply.github.com>
Co-authored-by: Giorgos Giakoumettis <giorgos@yiakoumettis.gr>
Co-authored-by: PizzaV <103120356+pizzav-xyz@users.noreply.github.com>
Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com>
Co-authored-by: Markus Hartung <mail@hartmark.se>
* chore(release): open v3.8.19 development cycle
* chore(release): sync electron lockfile to 3.8.19
* feat(quality): quality-gate ratchet + anti-hallucination/rule-enforcement guardrails (Phases 0-6) (#3471)
* feat(quality): generic ratchet comparator (multi-metric, regression-only)
* chore(ci): Fase 0 quality-gate fixes — reconcile coverage gate (40->60), tier npm audit, wire orphaned contract gates, re-enable cheap husky pre-commit
* feat(quality): ratchet engine (collector + frozen baseline + CI job) and provider-consistency gate
- collect-metrics.mjs: emits quality-metrics.json (ESLint warnings + coverage when present)
- quality-baseline.json: frozen baseline (eslintWarnings=3482, regression-only)
- ci.yml: quality-gate job (ratchet + step summary + artifact) and check:provider-consistency in lint job
- check-provider-consistency.ts: every REGISTRY id must be a canonical provider (found krutrim half-registered → allowlisted as known pre-existing, blocks any NEW orphan)
- TDD: 9 tests (5 ratchet + 4 provider-consistency)
* feat(quality): Fase 2 anti-hallucination gates — fetch-targets, openapi-routes, deps allowlist
- check-fetch-targets: every dashboard fetch(/api/...) resolves to a real route.ts; found 7 pre-existing dashboard->route mismatches frozen as KNOWN_MISSING for triage
- check-openapi-routes: every openapi.yaml path resolves to a real route; found 1 stale spec entry (agent-bridge agents/{id}/state) frozen as KNOWN_STALE_SPEC
- check-deps: anti-slopsquatting allowlist (105 deps); new deps need explicit human-reviewed entry
- all wired into CI lint/docs jobs; TDD +12 tests (21 total across 5 gates)
* docs(quality): add quality-gates report + implementation plan to repo root
* feat(quality): Fase 3a — file-size ratchet (freeze 91 files >800 LOC, cap 800 for new)
- check-file-size.mjs: frozen files can only shrink; new files must be <= cap (kills the next 12k-line god-component)
- file-size-baseline.json: 91 files frozen at current LOC (largest 12883)
- wired into CI lint job; TDD 5 tests; --update ratchets the baseline down on shrink
* feat(quality): Fase 3b — duplication ratchet (jscpd@4, baseline 5.72%)
- check-duplication.mjs: runs jscpd@4 (pinned; v5 is an incompatible Rust rewrite) over src+open-sse, fails if duplication % rises vs frozen baseline (5.72%, measured: 1358 clones / 22967 dup lines). Targets the executor copy-paste (48/50 override execute() wholesale)
- wired into the parallel quality-gate CI job (off the lint critical path); TDD 4 tests; --update ratchets down
- snapshot now complete: coverage ~82.6%, eslint 3482 (98.5% no-explicit-any), duplication 5.72%, 91 files >800 LOC
* feat(quality): Fase 4a — anti test-masking gate
- check-test-masking.mjs: for each MODIFIED test file in a PR, flags net assert removal + new assert.ok(true) tautologies (base...HEAD diff). Directly enforces CLAUDE.md 'never weaken asserts to go green'
- wired into pr-test-policy CI job (reuses base fetch); no-op outside PR; TDD 5 tests
* feat(quality): Fase 4b — coverage ratchet (conservative floors, CI consumes merged coverage)
- quality-baseline.json: coverage.{statements,lines,functions,branches} floors (80/80/82/73, real ~82.58/82.58/84.23/75.22 with margin; tighten via --update after a green main run)
- check-quality-ratchet.mjs: --allow-missing (local quality:gate skips coverage.* without a coverage run; CI runs strict)
- ci.yml quality-gate job: needs test-coverage + downloads merged coverage-report so the ratchet enforces 'coverage cannot drop'
- TDD +1 test (6 total)
* feat(quality): Fase 6 — 8 new gates (Rule #11/#12, migrations, known-symbols, route-guard, complexity, docs-symbols, db-rules)
Deterministic gates, each freezing pre-existing violations in a documented allowlist (ratchet) so they pass now and block only NEW regressions:
- check-error-helper (Rule #12): 7 executors/handlers forwarding raw err.message frozen
- check-public-creds (Rule #11): 5 literal client_ids (Claude/Codex/Qwen/Kimi/Copilot) frozen
- check-migration-numbering: gaps 026/055 + dup 041 frozen (prevents the git-rm-deleted-migration incident)
- check-known-symbols: 93 executors conformance + 15 combo strategies + 18 translator pairs
- check-route-guard-membership (#15/#17): all 25 spawn-capable routes verified local-only (0 gaps)
- check-complexity: cyclomatic>15 / fn-length>80 ratchet (baseline 1739)
- check-docs-symbols: 30 stale doc /api refs frozen (docs hallucination)
- check-db-rules (#2/#5): 25 unexported db modules + 15 raw-SQL routes frozen
Wired into CI (lint / docs-sync-strict / quality-gate jobs). 115 TDD tests, all green. ESLint ratchet held at 3482.
* docs(quality): Phase 7 plan (security/dead-code/mutation/community tooling) — GATED to 2026-06-16
Stored, not active. 7 suggested gates + all discussed OSS/Community tools (SonarQube Community + osv-scanner + CodeQL + knip + sonarjs + type-coverage + lockfile-lint + Stryker + size-limit + axe-core + semcheck + agent-lsp + Qlty). Activation gate: do not start before 2026-06-16 (use Phases 0-6 in production for 1 week, validate in practice, then evolve).
* docs(quality): Phase 6A critical-audit plan + Phase 7 additions — gated to 2026-06-16 (#3530)
PLANO-QUALITY-GATES-FASE6A.md (12-task audit of Phases 0-6: orphan tests, stale-allowlist enforcement, scope gaps) + Phase 7 additions (gitleaks, actionlint+zizmor, license compliance). Both stored, activation gated to 2026-06-16. Tasks 6A.1/6A.2 were fast-tracked separately (#3536).
* feat(quality): 6A.1+6A.2 — test-discovery gate, 135 orphan tests re-wired, 2 production bug fixes, vitest in CI (#3536)
check-test-discovery gate (TDD; 195 orphans found, 135 re-wired into the node runner, 60 frozen+annotated). Triage fixed 2 real production bugs: missing BYPASS_PREFIX_NOT_ALLOWED zod refine (spawn-capable prefixes accepted into the bypass list, Hard Rules #15/#17) and resetDbInstance not firing stateReset resetters (stale schema memo → 503 instead of 403; also hit backup-restore). New test-vitest CI job: test:vitest blocking (146/146), test:vitest:ui informational (14 pre-existing fails, triage 2026-06-16).
* chore: ignore generated yt-downloader artifact files
Add dated yt-downloader output files to .gitignore to prevent
local automation artifacts from being accidentally committed.
* chore(quality): green-light the quality-gate — conscious file-size + eslintWarnings re-baselines (#3538)
file-size: 9 files frozen at current sizes (v3.8.18-era growth + core.ts +7 from #3536 fix). eslintWarnings 3482→3501: the published v3.8.18 tag already measures 3501 (delta predates the quality-gate job); v3.8.19 cycle is neutral. Reduction + --require-tighten = Phase 6A (2026-06-16).
* fix(check): exclude internal planning docs (docs/superpowers/) from the docs-symbols gate
docs/superpowers/plans/*.md are historical implementation-plan snapshots that
may cite planned/abandoned routes — not claims about the current code. Three
such refs entered during the v3.8.18 cycle, before this gate was on the
pipeline, and would have blocked the v3.8.19 release merge.
* chore(release): v3.8.19 — 2026-06-09
CHANGELOG section for the quality-infrastructure release (7 commits, 1:1
coverage), [3.8.18] label corrected to its real release date, local prompt
artifacts ignored.
* test: hermetic auth context for 2 re-wired suites + real headroom on the breaker reset-timeout flake
CI shards exposed what the dev DATA_DIR was masking locally: detect.test.ts
and managementCliToken.test.ts asserted 401/403/reject outcomes that only
exist when login protection is configured — on a fresh CI DB isAuthRequired()
is false and the policy anonymous-allows. Both now create an isolated
DATA_DIR with requireLogin+password (the established pattern).
observability-fase04: the breaker reset-timeout test ran with a 5ms margin
(resetTimeout 10 / sleep 15) — lazy HALF_OPEN refresh under shard contention
flipped the first OPEN assert. Now 250/300ms.
* test: align bypass-prefix schema test to the restored layer-1 contract + real waitFor headroom
appearance-widget-settings-schema asserted that /api/cli-tools/runtime/ was
ACCEPTED into the bypass list — written against the buggy schema (missing
BYPASS_PREFIX_NOT_ALLOWED refine, restored in #3536) and consecrating the
bug the AC-8 orphan test guards against. Split into accept-safe +
reject-spawn-capable cases. chatcore waitFor ceiling 1500→10000ms (green
runs return immediately; observed 1580ms expiry on 2-core CI runners).
* test(chatcore): fix structurally-broken pending-detail predicate (flatten before find)
pendingRequests.details[connectionId] is Record<modelKey, PendingRequestDetail[]>
— the upstream-timeout test's waitFor tested each ARRAY's .providerRequest
(always undefined), so it could never resolve and expired (failed on 3 CI jobs;
reproduced deterministically isolated, including at the published v3.8.18 tag).
Flatten to the actual details + declare the call_log_pipeline_enabled dependency
explicitly + waitFor ceiling with real CI headroom.
* chore(quality): re-baseline coverage floors to the honest post-re-wire denominator + changelog coverage for the stabilization commits
The 135 re-wired tests import modules that were never loaded before, so the
c8 denominator grew: the old ~82.5% was inflated by never-imported modules
being invisible. CI merged coverage now measures 78.4/78.4/83.84/75.73 —
floors set ~2pt below (76.5/76.5; functions/branches floors already hold).
Tightening via --require-tighten is Phase 6A work (2026-06-16).
* chore(release): open v3.8.18 development cycle
* fix(catalog): stop Codex CLI model-catalog refresh from erroring (#3481)
Codex's model-catalog refresh (codex_models_manager) does
GET /v1/models?client_version=<v> and decodes a JSON object with a
TOP-LEVEL `models` array. OmniRoute answers in the OpenAI-standard
`{object,data}` shape, so codex fails with "missing field `models`"
and logs "failed to refresh available models" on every startup.
Detect codex clients via the `originator` / `user-agent` = `codex_*`
headers they send and add an EMPTY top-level `models: []` so the decode
succeeds. Non-codex OpenAI clients keep the byte-identical `{object,data}`
response.
The array is intentionally empty: codex replaces its built-in per-model
agent prompt (`base_instructions`, ~21k chars) with whatever a populated
entry carries for the selected model, so emitting our catalog would drop
the agent prompt to nothing and break codex's agent behaviour (verified
empirically against codex 0.137). An empty list keeps codex on its
built-in model info — same inference as before, minus the error.
Validated end-to-end with the real handler against codex 0.137:
"failed to refresh available models" → 0 occurrences, instructions
preserved (built-in Codex agent prompt, not empty).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* chore: ignore quality reports and local prompt artifacts
Add generated quality gate reports, metrics files, and local setup prompt
artifacts to .gitignore to prevent committing environment-specific or
temporary files.
* fix(provider): detect Responses API format when body has `input` but … (#3490)
Integrated into release/v3.8.18
* fix(sse): normalize numeric provider ids to strings (#3451)
Integrated into release/v3.8.18
* feat(browserPool): resolve Playwright proxy from proxy_registry DB (#3492)
Integrated into release/v3.8.18
* fix(theoldllm): generate X-Request-Token server-side, drop Playwright (#3491)
Integrated into release/v3.8.18
* feat(plugins): add lifecycle hooks and theme-manager plugin (#3473)
Integrated into release/v3.8.18
* fix(combo): parallel pre-screen + circuit-breaker fast-exit for priority combos (#3169)
Integrated into release/v3.8.18
* feat(ui): unifi active and finished requests into single view #1422 (#3401)
Integrated into release/v3.8.18
* docs(changelog): record #3401, #3473, #3492, #3490, #3451, #3491, #3169 under v3.8.18
* feat(docs): add doc accuracy gate + refresh AGENTS.md counts (#3510)
Integrated into release/v3.8.18
* fix(sse): drop empty-choices chunks without usage instead of injecting retry text (#3513)
PR #3422 ('allow OpenAI usage-only empty choices chunks') reintroduced the
assistant-content injection '[OmniRoute] Upstream returned an empty response.
Please retry.' for empty `choices: []` chunks that carry no valid usage. Clients
(Goose/opencode) feed that text back as a turn and spin in a retry loop -- the
exact regression #3400 had fixed by dropping the chunk.
Restore the drop behavior for the no-usage case while preserving #3422's
standards-compliant forwarding of usage-only `include_usage` final chunks.
Realign the mislabeled stream-utils test (it asserted the injection) and add a
dedicated regression guard.
Reported-by: @mochizzan
Refs: #3502, #3388, #3400, #3422
* fix(authz): fall back to URL token when Authorization isn't a usable Bearer (#3504)
Integrated into release/v3.8.18
* fix(playground): authenticate via session, test key policy by id (#3503)
Integrated into release/v3.8.18
* docs(changelog): record #3510, #3504, #3503 under v3.8.18
* fix: llama base url normalization (#3519)
* docs(changelog): reconcile v3.8.18 — add #3519, #3513, #3435-repair, gitignore chore (full commit↔changelog coverage)
* fix(opencode-plugin): bound regex quantifiers in normaliseFreeLabel (polynomial-ReDoS)
CodeQL js/polynomial-redos: unbounded \s* before an anchored \s*$ allowed
O(n²) backtracking on attacker-influenced display names. Bounded to {0,8}/{1,8}
(ample for any real label spacing). Plugin builds + 254 tests green.
* fix(types): restore clean typecheck:core for v3.8.18 release gate
- getPendingRequests() typed to real shape (was widened to object) → fixes
unknown 'count' in the unified-requests view (#3401)
- streamChunks log payload cast to its declared type (callLogs.ts)
- preScreenTargets aligned to canonical IsModelAvailable signature (#3169),
Promise.resolve-normalized so .catch never hits a bare boolean
All 5 gates green: lint(0 err) + typecheck:core + cycles + docs-all + unit + vitest(146).
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Andrey Borodulin <borodulin@gmail.com>
Co-authored-by: Dmitrii Safronov <zimniy@cyberbrain.cc>
Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com>
Co-authored-by: PizzaV <103120356+pizzav-xyz@users.noreply.github.com>
Co-authored-by: Markus Hartung <mail@hartmark.se>
Co-authored-by: Felipe Almeman <4226997+zhiru@users.noreply.github.com>
The custom-models loop in getUnifiedModelsResponse gated every model through hasEligibleConnectionForModel(getConnectionsForProvider(...)). no-auth providers (theoldllm, etc.) never create DB connection rows, so that returned [] and the gate dropped every imported/custom model for them — the Playground dropdown showed nothing for imported models while built-in/custom models on auth providers worked. Built-in models survived because they go through providerSupportsModel(), which already has a no-auth bypass (#2798).
The custom-model gate now applies the same no-auth bypass, keeping the eligibility check (with parentProviderType) intact for auth providers.
Co-authored-by: tjengbudi <tjengbudi@users.noreply.github.com>
Co-authored-by: a2belugin <a2belugin@users.noreply.github.com>
Claude Code -> claude-opus-4-8 turns intermittently died with 'tool call could not be parsed (retry also failed)'. OmniRoute's claude identity cloak rebuilt the anthropic-beta header from scratch and unconditionally forced interleaved-thinking-2025-05-14 (+ advanced-tool-use / effort for heavy agents), even when the client never negotiated them. The forced interleaved-thinking conflicts with tool_choice-forced turns, producing malformed opus tool_use streams (and sibling 400 'Thinking may not be enabled when tool_choice forces tool use').
selectBetaFlags now takes the client's inbound anthropic-beta: when present, thinking/effort betas are only emitted if the client requested them. Opaque clients (no header — the OAuth cloak path) keep the full set unchanged, so existing behavior and the #2454 model-tier gating are preserved.
Co-authored-by: Forcerecon <Forcerecon@users.noreply.github.com>
Vertex AI's FunctionCall/FunctionResponse protos have no id field; emitting it made Vertex reject tool calls with 400 'Unknown name id'. The id is now stripped only when the routed provider is vertex/vertex-partner (threaded via credentials._provider), preserving it for the public Gemini API where Gemini 3+ uses it for signature matching.
Co-authored-by: nullbytef0x <nullbytef0x@users.noreply.github.com>
- Unit/integration tests: update hardcoded 42→43 in 7 test files
(agentSkillTools-mcp, agentSkills-catalog, agentSkills-generator,
agent-skills-content, agent-skills-discovery, listCapabilities-a2a)
to match the 43rd skill (config-codex-cli) added in the previous commit.
- Include CONFIG_SKILL_IDS in integration content test ALL_IDS so
skills/config-codex-cli/ is no longer "unexpected".
- listCapabilities.ts: change totalSkills from literal 42 to catalog.length
so it adapts to catalog growth automatically.
- computeCoverage assertions: include config.have in totalSkills check.
- CI: switch E2E artifact from upload-artifact path (ambiguous stripping)
to explicit tar archive. Fixes "Could not find a production build in
./.build/next" — the previous approach's download path was double-nested
(.build/next/next/...) due to upload-artifact LCA computation. tar -czf
stores .build/next/... relative to CWD; tar -xzf restores them verbatim.
- Also exclude .build/next/cache from the tar to keep archive lean.
- feat(translator): strip client_metadata in Responses→Chat translation
(Mistral 422 extra_forbidden fix); add regression test.
- Remove 429 from PROVIDER_BREAKER_FAILURE_STATUSES; 429 belongs to
connection cooldown, not whole-provider breaker (CLAUDE.md §resilience).
PR #3366 correctly added 429 to PROVIDER_FAILURE_ERROR_CODES in
accountFallback.ts (combo infinite-retry fix) but the parallel change
to chat.ts was wrong — the integration test from v3.8.10 confirms this.
- Align stream-utils tests to PR #3399 (SYNTHETIC_CLAUDE_EMPTY_RESPONSE_TEXT
→ "", message.content → null) and PR #3355 (malformed tool-call buffer
now emitted as plain text, not suppressed).
- Align services-branch-hardening test to PR #3399 (pinnedModel always
null from applyComboAgentMiddleware; server-side session pinning replaced
client-side <omniModel> tag extraction).
- Align combo-routing-engine context-cache tests to PR #3399 (no <omniModel>
tag in output, no X-OmniRoute-Model header, priority routing unchanged).
