Repository navigation
fix(cli-tools): resolve API key resolution and model mapping bugs - #1263
Conversation
…CLI tools Two critical bugs prevented the Dashboard CLI Tools feature from working: Bug #1: keyId stripped by Zod validation - Zod's safeParse() stripped the keyId field before backend could read it - The real API key was never resolved from DB and written to settings.json - Fix: Extract keyId from rawBody BEFORE validation in all 6 CLI settings routes - Affected: claude, codex, droid, cline, openclaw, kilo settings routes Bug #2: Key matching logic compared full key vs masked key - Frontend compared full key (sk-31c4ea6a...) against masked key (sk-31c4****dfa6) - They never matched, so the dropdown never showed the correct selected key - Fix: Mask the token from file before comparing against the masked API keys list - Affected: ClaudeToolCard.tsx Also fixes model mapping being scrambled when key resolution failed. Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
There was a problem hiding this comment.
Code Review
This pull request updates several CLI tool settings routes to extract keyId before Zod validation to prevent field stripping, ensuring full API keys can be resolved from the database. It also introduces token masking in the ClaudeToolCard component to match stored tokens against masked API keys and allows for custom model paths in provider connections. Feedback includes a critical issue regarding an undefined helper function toNonEmptyString, a logic error where API key validation occurs before ID resolution, and concerns about the fragility of hardcoded masking logic.
| const url = `${baseUrl}/models`; | ||
| // Use modelsPath from provider node if available, otherwise default to /models | ||
| const psd = asRecord(connection.providerSpecificData); | ||
| const modelsPath = toNonEmptyString(psd.modelsPath) || "/models"; |
There was a problem hiding this comment.
The function toNonEmptyString is not defined in this file, nor is it imported from any of the modules listed in the imports section. This will result in a ReferenceError at runtime when this code path is executed. You should either define this helper function or use an inline check to ensure psd.modelsPath is a valid string.
| const modelsPath = toNonEmptyString(psd.modelsPath) || "/models"; | |
| const modelsPath = typeof psd.modelsPath === "string" && psd.modelsPath.trim() ? psd.modelsPath.trim() : "/models"; |
| // The dashboard sends masked key strings — resolving by ID guarantees | ||
| // we always write the full key value to the config file. | ||
| const keyId = typeof rawBody?.keyId === "string" ? rawBody.keyId.trim() : null; | ||
| // Resolve real key from DB by ID |
There was a problem hiding this comment.
The if (!apiKey) check (lines 177-182) occurs before the keyId resolution logic. If a client provides a keyId but omits the apiKey in the request body (expecting the backend to resolve it from the database), the request will fail with a 400 error at line 177. To fix this, the validation check for apiKey should be moved after the if (keyId) block.
| const matchedKey = apiKeys?.find((k) => k.key === tokenFromFile); | ||
| // (#523) Keys from /api/keys are masked (first 8 + "****" + last 4). | ||
| // Mask the token from file to compare against the masked list. | ||
| const maskedToken = tokenFromFile.slice(0, 8) + "****" + tokenFromFile.slice(-4); |
There was a problem hiding this comment.
This masking logic is hardcoded and assumes the token is at least 12 characters long. It is also fragile as it depends on the exact masking format (first 8 + last 4) used by the /api/keys endpoint. If the backend masking logic changes or if a shorter token is encountered, this comparison will fail to match correctly.
d89e38e
into
cheaper-inference:release/v3.6.6
* fix(streaming): #1211 greedy strip omniModel tags to prevent literal \n\n artifacts - Changed regex quantifier from ? to * in combo.ts, comboAgentMiddleware.ts, and contextHandoff.ts to greedily strip all JSON-escaped newline sequences surrounding <omniModel> tags in SSE streaming chunks - Added \r to the character class for cross-platform robustness - Fixed Playwright strict-mode violation in combo-unification.spec.ts - Bumped OpenAPI version and CHANGELOG to 3.6.6 * fix: 3 bugs found during issue triage (#1175, #1187/#1218, #1202) - fix(gemini): strip VS Code JSON Schema extensions from tool schemas (#1175) Add enumDescriptions, markdownDescription, markdownEnumDescriptions, enumItemLabels and tags to UNSUPPORTED_SCHEMA_CONSTRAINTS so the Gemini sanitizer removes them before forwarding. GitHub Copilot injects these non-standard fields into tool definitions, causing Gemini to reject with 'Unknown name enumDescriptions at functionDeclarations[n].parameters'. - fix(health-check): unwrap proxy config object before passing to getAccessToken (#1187 #1218) resolveProxyForConnection() returns { proxy, level, levelId } but the health check loop was passing the full wrapper to getAccessToken(), which expects the inner config object (.host, .port etc). The proxy dispatcher validated .host on the wrapper (undefined) and threw 'Context proxy host is required', silently marking every connection as unhealthy every sweep. Fix mirrors the pattern already used in chatHelpers.ts: proxyResult?.proxy || null. - fix(ui): debounce models.dev sync interval slider to save only on release (#1202) The slider's onChange fired updateInterval() on every drag tick, sending a PATCH per pixel of movement. Rapid API responses overwrote UI state mid-drag. Introduce draftIntervalHours for smooth visual feedback; the PATCH fires on onMouseUp / onBlur once the user releases the control. * fix(providers): update Xiaomi MiMo token-plan endpoints (#1238) Integrated into release/v3.6.6 * fix(cc-compatible): trim beta flags and preserve cache passthrough (#1230) Integrated into release/v3.6.6 * feat(memory+skills): full-featured memory & skills systems with tests (#1228) Integrated into release/v3.6.6 * fix: forward client x-initiator header to GitHub Copilot upstream (#1227) Integrated into release/v3.6.6 * feat(bailian-quota): add Alibaba Coding Plan quota monitoring (#1235) * fix: resolve v3.6.6 backlog bugs (#1206, #1211, #1220, #1231) - fix(core): #1206 inject startup guard against app/ and src/app/ conflict - fix(health): #1220 add HEALTHCHECK_STAGGER_MS to prevent token refresh bursting - fix(proxy): #1231 prioritize HTTP 429 over quota body heuristics - fix(sse): #1211 strip leading double-newlines in responses API stream * fix(tests): resolve memory migration and skills route pagination bugs from PR overlaps * docs: Update CHANGELOG.md with v3.6.6 features (#1182, #1165, #1177) * chore(release): bump version to 3.6.6 Update package versions for the electron app and open-sse package. Sync llm.txt metadata and feature headings with the 3.6.6 release. * feat(core): harden outbound provider calls and add cooldown retries Add guarded outbound fetch helpers with private/local URL blocking, controlled retries, timeout normalization, and route-level status propagation for provider validation and model discovery. Introduce cooldown-aware chat retries with configurable requestRetry and maxRetryIntervalSec settings, model-scoped cooldown responses, and improved rate-limit learning from headers and error bodies so short upstream lockouts can recover automatically. Also align Antigravity and Codex header handling, require API keys for Pollinations, validate web runtime env at startup, restore sanitized Gemini tool names in translated responses, and inject a synthetic Claude text block when upstream SSE completes empty. * feat(models): add glmt preset and hybrid token counting Introduce GLM Thinking as a first-class provider preset with shared GLM model metadata, pricing, usage sync, dashboard support, and provider request defaults for higher token budgets and longer timeouts. Use provider-side /messages/count_tokens when a Claude-compatible upstream supports it, while preserving estimated fallback behavior for missing models, missing credentials, and upstream failures. Also add startup seeding for default model aliases and normalize common cross-proxy model dialects so canonical slashful model ids do not get misrouted during resolution. * feat(api): add sync tokens and v1 websocket bridge Add dedicated sync token storage, issuance, revocation, and bundle download routes backed by stable config bundle versioning and ETag support. Expose the v1 websocket handshake route and custom Next server bridge so OpenAI-compatible websocket traffic can be upgraded and proxied through the dashboard and API bridge. Expand compliance auditing with structured metadata, pagination, request context, auth and provider credential events, and SSRF-blocked validation logging. * docs: Update all documentation for v3.6.6 - CHANGELOG: Add WebSocket bridge, GLM Thinking preset, safe outbound fetch/SSRF guard, cooldown-aware retries, compliance audit v2, model alias seeding, and all Internal Improvements for the 3 new commits - README: Expand v3.6.x highlights table with 10 new features; add SafeOutboundFetch, CooldownAwareRetry, SSRF guard, TPS metric, sync tokens, WebSocket bridge to Resilience/Observability/Deployment tables - ARCHITECTURE: Bump date; add new modules to executive summary, API routes, SSE core services, Auth/Security section; add SSRF/Outbound guard failure mode (section 6); expand module mapping - ENVIRONMENT: Add OMNIROUTE_CRYPT_KEY/OMNIROUTE_API_KEY_BASE64 legacy aliases, OUTBOUND_SSRF_GUARD_ENABLED, CODEX_CLIENT_VERSION, and REQUEST_RETRY/MAX_RETRY_INTERVAL_SEC cooldown retry settings - FEATURES: Add 6 new feature sections — V1 WebSocket Bridge, Sync Tokens & Config Bundle, GLM Thinking Preset, Safe Outbound Fetch & SSRF Guard, Cooldown-Aware Retries, Compliance Audit v2 * fix: use api64 for proxy test (#1255) Integrated into release/v3.6.6 — IPv6 proxy test fix * fix(page): update custom models section to include all providers #1200 (#1256) Integrated into release/v3.6.6 — Gemini custom model picker fix * fix: provide default client_id fallbacks to prevent broken OAuth requests (#1246) Integrated into release/v3.6.6 — OAuth client_id default fallbacks * fix: translate max_tokens/max_completion_tokens → max_output_tokens in Chat→Responses translator (#1245) Integrated into release/v3.6.6 — max_tokens → max_output_tokens Responses API translation + unit tests * feat(oauth): support cursor-agent CLI as Cursor credential source (#1258) Integrated into release/v3.6.6 — cursor-agent CLI credential source support * fix(cc-compatible): restore upstream SSE and correct stream/combo timeout behavior (#1257) Integrated into release/v3.6.6 — CC-compatible upstream SSE restore + stream timeout fix + README table repair * fix(cli-tools): resolve API key resolution and model mapping bugs in CLI tools (#1263) Integrated into release/v3.6.6 * feat(cli-tools): add Qwen Code CLI integration (#1266) Integrated into release/v3.6.6 * fix(i18n): add missing zh-CN translations and fix logger imports (#1269) Integrated into release/v3.6.6 * fix(i18n): add Chinese i18n support to dashboard components (#1274) Integrated into release/v3.6.6 * feat: update Pollinations to require API key, remove free tier flag (#1177) * feat: friendly error messages for crypto/encryption failures (#1165) * feat: add TPS (tokens per second) metric column to request logs (#1182) * feat: merge custom/imported models into filter list for all providers (#1191) * feat(fallback): Fix provider-profile-driven lockouts (#1267) This integrates rdself's unify-provider-profile-locks PR manually to handle structural conflicts. * fix(claude): proper Anthropic SDK integration (#1271) * fix(healthcheck): use correct proxy wrapper format for getAccessToken (#1272) * chore(release): v3.6.6 — skills registry stability fix + final integration * fix(auth): harden bootstrap auth and memory dashboard behavior Restrict unauthenticated writes to /api/settings/require-login to the initial bootstrap window while keeping read-only checks public. This prevents post-setup config changes without blocking first-run login setup, and the onboarding flow now logs in immediately after setting the password. Restore memory API filtering and pagination behavior by supporting q searches, honoring offset-based requests, and avoiding unrelated fallback results when FTS misses. Update dashboard stats fallback to use the response totals consistently. Package the MCP server with explicit file entries and add regression tests for bootstrap auth and memory route behavior * fix(codex): remove max_output_tokens from body for compatibility * chore(release): v3.6.6 — include PR 1274 fixes in changelog * chore: exclude additional build artifacts and internal directories from npm package distribution * fix: update Gemini OAuth test to match registry defaults + codex UI improvements * fix: restore .mjs refs for scripts/ in test imports after ts migration * fix: restore next.config.mjs ref in dev-origins test * fix: implement db migration safety checks and codex config format * fix: disable mass-migration abort during unit tests based on auto-backup flag * fix: update script regex in auto-update tests to use .mjs * feat: Add Perplexity Web (Session) provider (#1289) Integrated into release/v3.6.6 * fix(cli): resolve codex routing config parsing, standardize select model button positioning, and clarify oauth documentation * docs(changelog): record recent cli, provider, and test updates Document the latest fixes for Codex routing configuration parsing and Lobehub provider icon fallback behavior. Add the note that the remaining JavaScript test files were migrated to TypeScript ES modules to reflect the completed test stack transition. * chore(release): merge #1286 minor improvements manually to avoid testing conflict * chore(test): rename perplexity-web.test.mjs to .ts to maintain 100% TS codebase * chore(docs): update CHANGELOG.md for perplexity-web provider * fix(security): resolve CodeQL incomplete URL substring sanitization via URL parsing in test mocks * fix: integrate compressContext() into chatCore.ts request pipeline Proactively compress oversized contexts before sending to upstream providers, preventing context_length_exceeded errors. Compression triggers at 85% of model's context limit using the existing 3-layer compressContext() function. - Import compressContext, estimateTokens, getTokenLimit from contextManager - Add compression check after translation, before executor dispatch - Estimate tokens and compare against 85% threshold of model's context limit - Apply 3-layer compression (trim tools, compress thinking, purify history) - Log compression events with before/after token counts and layers applied - Audit compression events for observability - Add unit tests verifying integration behavior Closes #1290 * fix(tests): align reasoning expectations with GLM thinking structure * fix: prevent orphaned tool_result messages in purifyHistory() When purifyHistory() drops oldest messages to fit context window, it can split tool_use/tool_result pairs — keeping the tool_result but dropping the tool_use that initiated it. This causes upstream providers to reject the request with format errors. Add fixToolPairs() that runs after each purification pass to remove: - OpenAI format: orphaned role='tool' messages without matching tool_calls ID - Claude format: orphaned tool_result content blocks without matching tool_use ID Closes #1291 * fix(tests): supply tool_use in mock so it is not dropped * chore: convert remaining test to TypeScript * fix(tests): restore compatibility with compressContext threshold test after tsx migration * docs: finalize v3.6.6 release documentation * fix(core): finalize provider removal, type issues, and codex API key config * fix(dashboard): render Web/Cookie, Search, Audio provider sections and fix TypeScript errors * fix: increase MCP web_search timeout to 60s (#1278) * fix: route combo testing properly for embedding models (#1260) * fix: accumulate excluded accounts in combo fallback loop (#1233) * fix: strip leading whitespace and newlines from first streaming chunk (#1211) * docs: clarify VPS and Docker settings for OAuth credentials (#1204) * fix: return real retry-after for pipeline gates (#1301) Integrated into release/v3.6.6 — returns real Retry-After values from pipeline gates * feat: streaming semantic cache, Cursor auto-version detection, and call-log enhancements (#1296) Integrated into release/v3.6.6 — streaming semantic cache, Cursor auto-version detection, call-log cache_source tracking * feat(api): support more OpenAI types (image, embeddings, audio-transcriptions, audio-speech) (#1297) Integrated into release/v3.6.6 — adds embeddings, audio-transcriptions, audio-speech, and images-generations support for custom OpenAI-compatible providers, plus Pollinations image registry * deps: bump hono from 4.12.12 to 4.12.14 (#1302) Integrated into release/v3.6.6 * deps: bump hono from 4.12.12 to 4.12.14 (#1306) Integrated into release/v3.6.6 * chore: stabilization fixes for v3.6.6 (#1298, #1254, #59, CI) * fix(providers): match correct endpoint for Xiaomi MiMo, strip routing prefix for custom openai endpoints (#1303, #1261) * feat(storage): add database backup cleanup controls * chore(release): v3.6.6 — Final Stabilization Push * Backport call log storage refactor to release/v3.6.6 (#1307) Integrated into release/v3.6.6 * deps: update dompurify to 3.4.0 to resolve CVE-XYZ (#60) * test: disable sqlite auto backup in CI to resolve E2E timeout (#24481475058) * chore(docs): sync CHANGELOG for v3.6.6 with missing features and fixes * chore(release): prep v3.6.6 infrastructure and type safety fixes - Migrated legacy .mjs scripts to .ts (bin, prepublish, policies) - Resolved pre-commit strict lint (t11 budget) errors in combo.ts - Explicitly typed all TS bindings in pack-artifact policies - Updated package.json commands to run Node via tsx/esm internally - Hardened CI/CD with explicit node version 22.22.2 checks - Completed stage validations for v3.6.6 final release * chore: fix TS build errors and e2e timeouts in CI - Migrate nodeRuntimeSupport to TS interfaces avoiding implicit any - Increase visibility timeouts in skills-marketplace E2E test to 15s to bypass CI flakiness - Complete migration of .mjs scripts to .ts ensuring type safety * chore(release): sync package version 3.6.6 across workspaces * test(e2e): universally increase UI component visibility timeouts from 5s to 15s to bypass CI starvation * chore(build): inject baseUrl, paths, and types:node into MITM tsconfig within prepublish hook to fix missing types in CI check --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com> Co-authored-by: Jack <5443152+hijak@users.noreply.github.com> Co-authored-by: Randi <55005611+rdself@users.noreply.github.com> Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com> Co-authored-by: Samuel Cedric <ceds.sam@gmail.com> Co-authored-by: Max Garmash <max@37bytes.com> Co-authored-by: Markus Hartung <mail@hartmark.se> Co-authored-by: Gi99lin <74502520+Gi99lin@users.noreply.github.com> Co-authored-by: Payne <baboialex95@gmail.com> Co-authored-by: Benson K B <bensonkbmca@gmail.com> Co-authored-by: clousky2020 <33016567+clousky2020@users.noreply.github.com> Co-authored-by: Ravi Tharuma <25951435+RaviTharuma@users.noreply.github.com> Co-authored-by: oyi77 <oyi77@users.noreply.github.com> Co-authored-by: Hdsje <vovan877@gmail.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: xiaoge1688 <moyekongling@gmail.com>
…#4343) * fix(cli): honor TAILSCALE_AUTHKEY for non-interactive tailscale login (port from 9router#1263) startTailscaleLogin built `tailscale up` without ever reading process.env.TAILSCALE_AUTHKEY, so a pre-authenticated / headless daemon waited for an interactive auth URL and timed out (~15s). When TAILSCALE_AUTHKEY is set it is now passed via `--auth-key=` (an argv element to spawn(binary, args) — no shell interpolation, Hard Rule #13); when unset, behavior is unchanged. The arg builder is extracted into a pure exported `tailscaleUpArgs()` for testing. Reported-by: ipeterpetrus (decolua/9router#1263) Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> * chore(quality): rebaseline tailscaleTunnel.ts file-size to 1202 (#1263 +13) --------- Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com>
…4397) * chore(release): open v3.8.31 development cycle * fix(mitm): exact host membership in MITM hosts test (CodeQL false positive) (#4386) getMitmToolHosts returns string[], so .includes(host) is Array.prototype.includes (exact membership). CodeQL's js/incomplete-url-substring-sanitization heuristic misreads it as a String.includes() URL-substring sanitization check and raises a HIGH alert. Switch to .some(h => h === host) — identical semantics, explicit intent, no flagged pattern. Surfaced post-v3.8.30 (#4325) once the test landed on main. Test-only change (no runtime behavior); the suite still passes and the CodeQL re-scan on merge clears the alert. * fix(codex): request reasoning summaries (#4359) Adds reasoning.summary=auto + reasoning.encrypted_content include for Codex. Thanks @xz-dev. * fix(embeddings): inject NVIDIA NIM input_type for asymmetric embed models (#4341) NVIDIA NIM asymmetric embedding models (e.g. nvidia/nv-embedqa-e5-v5) reject requests without an `input_type` ("query" | "passage") with 400 "'input_type' parameter is required". The embedding registry now carries a model-level default param for the asymmetric NVIDIA model, and the embeddings handler injects a model's default params into the upstream body only when the client omitted them, leaving a client-supplied value untouched. Reported-by: hydraromania (decolua/9router#1378) Co-authored-by: hydraromania <252583922+hydraromania@users.noreply.github.com> * fix(api): migrate deprecated Codex [features].codex_hooks to [features].hooks (#4342) Codex renamed the `codex_hooks` feature flag to `hooks`; recent Codex CLI versions ignore the old key and warn. When OmniRoute rewrites an existing config.toml (configure/reset Codex provider) it now renames [features].codex_hooks -> [features].hooks, preserving the value and never clobbering an already-present `hooks`, then drops the deprecated key. The migration is a no-op when the flag is absent and runs on both the POST and DELETE config paths. Reported-by: Bian-Sh (decolua/9router#1327) Co-authored-by: Bian-Sh <24520547+Bian-Sh@users.noreply.github.com> * fix(translator): drop the null flush on the same-format response path (#4344) The streaming response translator's same-format fast path returned `[chunk]` unconditionally, so the end-of-stream null/flush signal (chunk === null) propagated as a literal `[null]`. Downstream this surfaced as an empty `data: null` SSE event between chunks and crashed strict clients (e.g. Factory Droid BYOK on /v1/responses). The fast path now returns `[]` for the null flush while still passing real chunks through unchanged. Reported-by: thaitryhand (decolua/9router#1052) Co-authored-by: thaitryhand <248103256+thaitryhand@users.noreply.github.com> * fix(translator): strip assistant echo fields on the OpenAI target path (Mistral 422) (#4350) Strict OpenAI-compatible upstreams (e.g. mistral/codestral-latest) reject client-only assistant echo fields sent back as input with 422 extra_forbidden (the report hit messages[].assistant.reasoning_content via Codex /responses). Only reasoning_content was stripped on the OpenAI target path; the sibling fields reasoning / refusal / annotations / cache_control leaked through. They are now all dropped on the non-reasoner OpenAI target path. `audio` is intentionally preserved (OpenAI audio models reference a prior assistant audio response by id; Mistral never emits audio). Reported-by: xxy9468615 (decolua/9router#1649) Co-authored-by: xxy9468615 <63351664+xxy9468615@users.noreply.github.com> * fix(cli): honor TAILSCALE_AUTHKEY for non-interactive tailscale login (#4343) * fix(cli): honor TAILSCALE_AUTHKEY for non-interactive tailscale login (port from 9router#1263) startTailscaleLogin built `tailscale up` without ever reading process.env.TAILSCALE_AUTHKEY, so a pre-authenticated / headless daemon waited for an interactive auth URL and timed out (~15s). When TAILSCALE_AUTHKEY is set it is now passed via `--auth-key=` (an argv element to spawn(binary, args) — no shell interpolation, Hard Rule #13); when unset, behavior is unchanged. The arg builder is extracted into a pure exported `tailscaleUpArgs()` for testing. Reported-by: ipeterpetrus (decolua/9router#1263) Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> * chore(quality): rebaseline tailscaleTunnel.ts file-size to 1202 (#1263 +13) --------- Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> * fix(dashboard): OAuth modal surfaces the real error on a non-JSON response (#4351) * fix(dashboard): OAuth modal surfaces real error on non-JSON responses (port from 9router#1318) The OAuth connect/reauth modal called `await res.json()` unconditionally, so a non-JSON error response (e.g. a plain-text 500 page from a build/OAuth endpoint) threw `Unexpected token 'I'...` and hid the real failure. New shared helpers parseResponseBody / getErrorMessage (src/shared/utils/api.ts) read the body safely (JSON when JSON, raw text otherwise) and produce a clean message either way; every modal fetch site now uses them. Reported-by: DNNYF (decolua/9router#1318) Co-authored-by: DNNYF <74033321+DNNYF@users.noreply.github.com> * fix(dashboard): type OAuth modal response body as Record<string, unknown> (t11 any-budget) Switch the parseResponseBody casts from Record<string, any> to Record<string, unknown> so OAuthModal.tsx stays within its t11 explicit-any budget. getErrorMessage already takes unknown; the success paths typecheck clean under strict:false. No runtime change. --------- Co-authored-by: DNNYF <74033321+DNNYF@users.noreply.github.com> * fix(translator): accept AI SDK-style { type: image, image: data-URL } content parts (#4345) * fix(translator): accept AI SDK-style { type: image, image: "data:..." } parts (port from 9router#1330) Several OpenAI-input translators only recognized images shaped as `image_url.url` (or an object with `.source`/`.url`), so an AI SDK-style content part where `image` is a bare data-URL STRING was silently dropped before reaching a vision provider (OpenCode is one affected client; the gap is generic). The OpenAI->Claude, OpenAI->Kiro and OpenAI->Gemini/Antigravity translators now parse a string `image` data URL into each provider's native image shape (Claude base64 source, Kiro images[].source.bytes, Gemini inlineData). Reported-by: mugnimaestra (decolua/9router#1330) Co-authored-by: mugnimaestra <13349159+mugnimaestra@users.noreply.github.com> * chore(quality): freeze openai-to-kiro.ts file-size at 807 (#1330 +9, over 800 cap) --------- Co-authored-by: mugnimaestra <13349159+mugnimaestra@users.noreply.github.com> * fix(dashboard): show a disabled connection's last error in the row (#4352) * fix(dashboard): show a disabled connection's last error in the row (port from 9router#1447) The provider card's error badge counts a disabled connection (isActive === false) that has an error — its effective status is still error/expired/unavailable — but the connection row hid the lastError text for disabled rows, so the operator saw the count without the cause. The row's error-visibility decision is extracted into shouldShowConnectionLastError() and now shows the error whenever there is one, regardless of the active toggle. Reported-by: ntdung6868 (decolua/9router#1447) Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * chore(quality): rebaseline ConnectionRow.tsx file-size to 942 (#1447 +1 import) --------- Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * fix(providers): bound the OAuth connection-test probe with a timeout (#4347) * fix(providers): bound OAuth connection-test probe with a timeout (port from 9router#1449) The OAuth path of "Test Connection One-by-One" called bare fetch() with no AbortController/signal, so a provider probe that accepted the socket but never responded wedged the test queue forever. Both the initial probe and the post-refresh retry are now bounded with AbortSignal.timeout(30s) — matching the API-key path's existing budget — and a timed-out probe resolves as a failure with a clear "Test timed out after 30s" message in the route's normal error shape. Reported-by: ntdung6868 (decolua/9router#1449) Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * chore(quality): rebaseline providers test route file-size to 887 (#1449 + sibling #1444 growth) --------- Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * fix(providers): label a deactivated account distinctly from a revoked token (#4353) A Codex connection whose OAuth refresh is fully healthy but whose ChatGPT account has been deactivated by the provider gets a 401 from the upstream API. The connection test labeled that the same as a bad credential ("Token invalid or revoked" -> upstream_auth_error), so an operator could not tell a deactivated account from a revoked token. The test now reads the 401/403 body and, when it indicates account deactivation, classifies it as account_deactivated (which the dashboard already renders as "Account Deactivated"); a plain auth 401 is unchanged. Reported-by: ntdung6868 (decolua/9router#1444) Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * fix(db): cascade-delete orphaned model aliases when a provider is removed (#4348) * fix(db): cascade-delete orphaned model aliases when a provider is removed (port from 9router#1409) Deleting a custom provider removed its connections and node but left the imported model-alias rows (key=<alias>, value="<providerId>/<model>") behind, so re-importing the same provider was blocked by stale "already exists" aliases. Add a deleteModelAliasesForProvider(providerId) DB helper that drops every alias whose stored value begins with "<providerId>/", and call it from the provider-node DELETE handler so a fresh import is unblocked. Reported-by: nguyenvanhuy0612 (decolua/9router#1409) Co-authored-by: nguyenvanhuy0612 <57367674+nguyenvanhuy0612@users.noreply.github.com> * chore(quality): rebaseline models.ts file-size to 1221 (#1409 + sibling #1294 growth) --------- Co-authored-by: nguyenvanhuy0612 <57367674+nguyenvanhuy0612@users.noreply.github.com> * fix(api): persist max_input_tokens/max_output_tokens when adding a custom model (#4349) The POST /api/provider-models handler read the rest of the body but never the two token-limit fields, and addCustomModel() had no parameter for them, so the form values were dropped on write while the DB layer and /v1/models catalog already round-trip inputTokenLimit/outputTokenLimit. Accept the two optional limits in the schema, forward them through the handler, and persist them in addCustomModel(). TDD: failing-then-passing unit test. Reported-by: codename-zen (decolua/9router#1294) Co-authored-by: codename-zen <263238141+codename-zen@users.noreply.github.com> * docs: feature-documentation catch-up (v3.8.20 → v3.8.30) (#4391) One-time reconciliation of the docs with every user-facing feature shipped since v3.8.20 (we had never done a dedicated pass, so debt had accumulated): - README: new '✨ What's New' section (curated v3.8.20→v3.8.30 highlights). - New guides: CLI-INTEGRATIONS (all setup-*/launch commands), MITM-TPROXY-DECRYPT (transparent-decrypt epic), CONTEXT_EDITING (delegated Anthropic clear_tool_uses). - Refreshed: AUTO-COMBO (auto/<category>:<tier> + Arena-ELO), API_REFERENCE (x-omniroute-no-memory), MEMORY (int8 quantization + off-by-default), RESILIENCE (model-lockout success-decay), RTK, AGENTBRIDGE, TRAFFIC_INSPECTOR, GUARDRAILS, CLOUD_AGENT, ENVIRONMENT, SETUP_GUIDE, CLI-TOOLS, MCP-SERVER. - Regenerated PROVIDER_REFERENCE (231 providers); synced the count in README/CLAUDE/AGENTS. - Allowlisted external-tool env vars (OPENAI_API_BASE, PROMPTFOO_PROVIDER_KEY) and the STREAM_RECOVERY config-object name in the docs-accuracy gates. All claims source-verified; check:docs-all (sync/counts/env/links/fabricated) passes. Going forward this runs every release via generate-release step 6b. * fix(executors): don't inject thinking when tool_choice forces a tool (native Claude) (#4389) Forced tool_choice now strips the adaptive thinking injection to avoid Anthropic 400. Thanks @NomenAK. * fix(translator): Gemini accepts HTTP/HTTPS image URLs (port from 9router#344) (#4373) OpenAI-style `image_url` parts with an `http://` or `https://` URL reached `convertOpenAIContentToParts` and were dropped with only a `console.warn`, because Gemini's `inlineData` requires base64 (the helper is synchronous and cannot fetch+encode upstream assets). Gemini's `Part` schema, however, natively accepts `fileData: { fileUri }` for remote URIs — the model fetches the asset itself. The helper now emits a `fileData` part (`mimeType: "image/*"`, inferred upstream on fetch) for HTTP/HTTPS URLs instead of silently dropping them. Vision requests that pass a URL — not a data: URI — now reach Gemini intact. No behavioral change for: - `data:` URIs → still emitted as `inlineData` with the parsed media type. - Unsupported schemes (e.g. `ftp:`) → still skipped (Gemini would reject them). The openai-to-claude side already passed HTTP/HTTPS URLs through as `source: { type: "url", url }` (lines 573–578) — the upstream PR's Claude-side change was already covered. Regression test: tests/unit/gemini-helper-http-image-url-port344.test.ts (4 cases: https URL, http URL, data: URI no-regression, unsupported-scheme guard). Inspired-by: decolua/9router#344 Co-authored-by: Ibrahim Ryan <ryan@nuevanext.com> * fix(executors): strip stream_options for qwen non-streaming / thinking Claude Code requests (port from 9router#663) (#4374) Claude-Code-compatible providers force the executor-level `stream` flag on via `upstreamStream = stream || isClaudeCodeCompatible` (open-sse/handlers/chatCore.ts), but the outgoing body keeps the caller's original `stream: false`. The shared `stream && targetFormat === "openai"` branch in DefaultExecutor.transformRequest then injected `stream_options: { include_usage: true }` onto a body that still said `stream: false`, and qwen upstream rejected the request with `400 "'stream_options' only set this when you set stream: true"`. The same rejection surfaced when the body carried `thinking` / `enable_thinking`. The qwen branch now skips the injection (and strips any client-sent `stream_options`) when the body explicitly says `stream: false` or requests thinking, leaving regular qwen streaming requests with the include_usage injection intact. Other providers are unaffected. Adds a TDD regression with 4 cases covering both opt-out paths and the normal-streaming positive control. Inspired-by: decolua/9router#663 Co-authored-by: anuragg-saxenaa <anuragg.saxenaa@gmail.com> * fix(security): scope OAuth callback postMessage to a trusted-origin allowlist (port from 9router#998) (#4372) The OAuth callback at `/callback` previously fell back to `window.opener.postMessage({ code, state, ... }, "*")` whenever the opener was cross-origin. The fallback was intended to support remote-OmniRoute + local-loopback callbacks (where opener and callback live on different origins), but the same code path also delivers the OAuth code/state to any hostile opener that pops the well-known callback URL — letting that attacker complete the OAuth flow as the user. Replace the wildcard fallback with iteration over a fixed allowlist: `window.location.origin` (same-origin parent — the popup-mode dashboard) plus Codex's fixed loopback helper (`http://localhost:1455` and the IPv4 literal `http://127.0.0.1:1455`). The browser drops `postMessage` to any opener whose actual origin is not in `targetOrigin`, so the message reaches only known parents and is silently dropped for any other. The same-origin fallback path is unchanged — methods 2 (`BroadcastChannel`) and 3 (`localStorage` storage event) still cover same-origin openers that COOP severed. The `openerSameOrigin` probe stays in place to drive the auto-close vs manual-copy UI decision (no behavior change for the success path). Adds a regression test (`tests/unit/ui/oauth-callback-postmessage-scope.test.tsx`) that mounts the page with a stubbed cross-origin opener and asserts no `postMessage` call ever uses `"*"` and every call lands on an allowlisted origin. The test failed against the pre-fix code (red), passes after the fix (green) — TDD per CLAUDE.md hard rule #18. Partial port of upstream decolua/9router#998: the upstream PR also re-enabled TLS verification on a DNS-bypass fetch in `open-sse/utils/proxyFetch.js`; that part is N/A here because OmniRoute's `proxyFetch.ts` never disabled TLS verification (no `rejectUnauthorized: false` anywhere in the file). Inspired-by: decolua/9router#998 Co-authored-by: aeonframework <aeon@aeonframework.dev> * fix(sse): default combo per-target timeout to 120s for fast failover (#4365) Combo per-target timeout inherited the full FETCH_TIMEOUT_MS (600s) when a combo did not set its own targetTimeoutMs, so a single hung/slow target stalled the whole combo for up to 10 minutes before falling through to the next model. Introduce DEFAULT_COMBO_TARGET_TIMEOUT_MS (120s) as the unset-default in resolveComboTargetTimeoutMs (new 3rd arg) and wire it in phaseComboSetup. The upstream ceiling (600s) and per-combo opt-out (targetTimeoutMs, up to the ceiling) are preserved; single non-combo requests are unchanged. For streaming requests this only bounds time-to-first-headers, so token generation is not cut short. TDD: failing-then-passing unit test in tests/unit/combo-config.test.ts. * refactor(combo): de-dup exhausted-target skip predicate across both dispatchers (#4362) Primeiro incremento da de-dup dos 2 dispatchers de combo (handleComboChat + handleRoundRobinCombo). O bloco de pre-check #1731/#1731v2 (skip de target já exhausted no provider/connection) era BYTE-IDÊNTICO nos dois (mesmas condições, mesmas mensagens), diferindo só na tag de log e no control-flow. - comboPredicates.ts: getExhaustedTargetSkipReason(target, exhaustedProviders, exhaustedConnections) — predicate PURO que retorna a mensagem de skip (ou null); cada dispatcher mantém seu próprio log-tag + control-flow (return null / continue) + fallbackCount. No mutate do stryker (cobertura de mutação). - combo.ts: −20 linhas (os 2 blocos viram 1 chamada cada). - 7 testes de caracterização travam condições + strings exatas. Comportamento preservado: 376/376 testes combo (357 caracterização + 7 novos), integração sse-correctness 5/5, typecheck 0, complexity neutro (1895), file-size encolhe. Próximo incremento: de-dup do error-handling/exhausted-tracking (handleTargetError). * refactor(combo): de-dup upstream-error exhaustion classification across both dispatchers (#4366) Segundo incremento da de-dup dos 2 dispatchers (handleTargetError). Após cada erro de target, ambos rodavam um bloco quase-idêntico que marca o provider exhausted (#1731), a conexão connection-errored (#1731v2) ou o provider transiently rate-limited. - combo/targetExhaustion.ts: applyComboTargetExhaustion(target, opts) — atualiza os 3 Sets de exhaustion e retorna providerExhausted. As MUTAÇÕES de Set (que dirigem o skip de targets, lidas por getExhaustedTargetSkipReason) são BYTE-IDÊNTICAS nos dois; as diferenças reais viram parâmetros: tag, allAccountsRateLimited (termo extra do RR, false no handleComboChat), exhaustedLogLevel (info no handleComboChat, debug no RR). Connection-level extraído p/ markConnectionLevelExhaustion (privado, <15 complexity). - combo.ts: −73 linhas; 4 imports órfãos removidos. - 7 testes de caracterização travam as mutações + o return. ÚNICA mudança de comportamento: o WORDING das mensagens de log do RR ganha o sufixo 'on remaining targets' (cosmético; mesmo #code, mesmas mutações, mesmos níveis de log). 376/376 combo (caracterização preservada), integração sse 5/5, typecheck 0, complexity neutro (1895), file-size encolhe. * refactor(chatCore): extract checkHeapPressureGuard leaf (god-file decomposition start) (#4371) Primeiro incremento da decomposição do chatCore.ts (5127 LOC, hot-path mais quente). O guard de memória do topo do handleChatCore (rejeita 503 quando o heap V8 passa o threshold de shed) vira um leaf testável, co-locado com o threshold em heapPressure.ts. - heapPressure.ts: checkHeapPressureGuard(heapUsedMb, thresholdMb) — retorna o result 503 pronto ou null. Byte-idêntico ao guard inline (mesmo check, mesma 503, mesmo warn). A figura de heap fica em telemetria INTERNA, nunca no response do cliente (Hard Rule #12). - chatCore.ts: o bloco inline (~22 ln) vira 3 linhas; import órfão de HEAP_PRESSURE_THRESHOLD_MB trocado por checkHeapPressureGuard. - 3 testes novos (incl. assert Rule #12: o MB medido não vaza no payload). complexity-baseline 1895->1896: drift de base pós-#4338 (medido com minhas mudanças stashed = 1896); esta mudança é complexity-NEUTRA (helper complexity 2, handleChatCore só perde código). 190/190 chatcore tests, typecheck 0, file-size encolhe. * Localize CLI and stabilize fetch, memory, and coverage handling (#4383) en-only i18n, fetch-start-timeout hardening, EngineConfigPage icon fix, CI build-artifact-reuse overhaul. Memory production hunk dropped as a no-op (tests kept). Thanks @JxnLexn. * test(combo): reset circuit breakers between stream-readiness cases (restore green) (#4396) The combo-dispatch cases in combo-stream-readiness-fallback.test.ts deliberately fail `glm` (zombie streams / repeated 504s), which legitimately trips the per-provider circuit breaker. That OPEN state is a module-level singleton, so it leaked into the next test and combo.ts then SKIPPED `glm/*` targets entirely ("Skipping … circuit breaker OPEN"). That made "combo does not retry stream readiness timeouts on the same model" never attempt glm/zombie — expected ['glm/zombie','openai/gpt-5.4-mini'] but got ['openai/gpt-5.4-mini']. This was a pre-existing red on release/v3.8.31 (present at the cycle-open tip), order-dependent: the test passes in isolation, fails after the preceding cases. Add a test.beforeEach(resetAllCircuitBreakers) so each scenario starts from a clean breaker slate. Test-isolation only — the breaker behavior is correct and no production code or assertion changes. Full combo suite: 390/390 green. * fix(cli): decline confirm() cleanly on non-interactive stdin + document contexts workflow The `contexts remove` command already has `--yes` to skip confirmation, but when run without it under a non-interactive stdin (pipe, CI, EOF) the [y/N] prompt could never be answered — the readline question stayed pending and Node warned about an "unsettled top-level await" at exit. confirm() now detects `!process.stdin.isTTY` and declines cleanly (returns false), pointing at `--yes` for non-interactive use. Exported confirm() for a regression test. Docs: REMOTE-MODE.md gains a full "Managing contexts" section (list/current/use to switch between remote and local, add/show/rename, remove with --yes, export/import), fixes a `context current` -> `contexts current` typo, and the README remote-mode snippet now shows switching back to local. Verified against the live CLI: command signatures, --yes, and the non-TTY decline path all behave as documented. Tests: cli-contexts.test.ts asserts confirm() declines on non-TTY stdin (RED before, GREEN after). All docs gates (fabricated/links/symbols) pass. * ci(t11): bump any-budget for executors/base.ts (2 false-positive "any" strings) Unblocks the Fast Quality Gates on release/v3.8.31: `check:any-budget:t11` was red on `open-sse/executors/base.ts` for ALL PRs (pre-existing base drift, unrelated to this branch). The checker counts `\bany\b` after stripping comments but NOT strings, and the native-Claude tool_choice logic uses the API value `"any"` in two string literals (`tb.tool_choice === "any"`, `.type === "any"`). There are zero actual TypeScript `any` types in the file — budget set to the matched count, mirroring the existing cursor.ts false-positive entry right below it. * perf: combos UI split + next config + 1-click redis + bifrost sidecar (#3932) (#4381) Combos UI split + next.config perf + 1-click local Redis launcher + bifrost relay. Review fixes (co-author): --rm/--restart conflict, error sanitization, UI/route names, dead-guard re-doc, bifrost Zod, IPv6 + CLI test fixes. Thanks @KooshaPari. * fix(plugin): prefix OC static-catalog combo+raw keys with providerId (#4384) OC parses model ids on '/'; combo keys now carry 'omniroute/' (was 'combo/'). Live-validated against the VPS via OpenCode. Thanks @herjarsa. * docs(env): document TAILSCALE_AUTHKEY (env/docs contract drift on .31) Second pre-existing base-drift fix needed to get Fast Quality Gates green: the `repository contract is in sync` test (check:env-doc-sync) was red on ALL .31 PRs. PR #4343 added a code reference to `process.env.TAILSCALE_AUTHKEY` (src/lib/tailscaleTunnel.ts) for non-interactive `tailscale up`, but never added the var to .env.example / docs/reference/ENVIRONMENT.md — the contract requires code vars to appear in both. Add the (commented) entry to .env.example next to TAILSCALE_BIN and a row to ENVIRONMENT.md. Verified: `node scripts/check/check-env-doc-sync.mjs` → in sync. Unrelated to this branch's confirm()/docs change; surfaced because touching a quality script triggers the TIA fail-safe full unit suite. * chore(release): v3.8.31 — 2026-06-20 Finalize the v3.8.31 release: reconcile the CHANGELOG (full commit-to-bullet coverage, 26 bullets across Features/Fixed/Security/Maintenance), refresh the README What's New section, back-fill the .30/.31 sections into the 41 i18n CHANGELOG mirrors, and add TAILSCALE_AUTHKEY to the env contract (.env.example + ENVIRONMENT.md). * chore(release): align mitm-hosts test comment with main to clear merge conflict The same CodeQL false-positive fix landed twice — #4386 on release/v3.8.31 and #4387 directly on main — with only the explanatory comment differing (the assertion is byte-identical). Adopt main's comment wording on the release branch so the release PR merges without a comment-only conflict. * test(translator): align stale openai->gemini remote-URL tests with #4373 Third pre-existing base-drift fix to get Fast Quality Gates green on .31. PR #4373 ('Gemini accepts HTTP/HTTPS image URLs', port of 9router#344) intentionally changed convertOpenAIContentToParts so remote http(s) image URLs pass through as a native `fileData: { fileUri }` part instead of the old #2807 drop+warn — and added its own test (gemini-helper-http-image-url-port344.test.ts) for the new behavior. But it left three tests in translator-openai-to-gemini.test.ts asserting the OLD drop+warn contract, so they fail deterministically (verified: the file fails in isolation on clean .31). These only surface under the TIA fail-safe FULL suite, which a quality- script touch triggers. Align the three stale tests to the real, intended behavior (captured by running the function): remote URLs -> fileData.fileUri (mimeType image/*), still never inlineData (the sync path cannot fetch+encode). This is test-vs-code alignment to a deliberate, separately-tested change — not a weakened assertion. 40/40 in the file; 94/94 across the translator + env-doc combo that previously failed. * chore(quality): reconcile complexity baseline 1896->1900 (/review-prs v3.8.31 batch) (#4410) * fix(release): reconcile full-CI drift for v3.8.31 (gemini tests #4373, any-budget #4389, masking allowlist #4384) The release PR's full CI surfaced cumulative cycle drift the per-PR fast gates skip: - tests/unit/translator-openai-to-gemini.test.ts: realign 3 cases to #4373's HTTP/HTTPS-URL fileData pass-through (they asserted the old warn-and-drop). - scripts/check/check-t11-any-budget.mjs: base.ts budget 0→2 — #4389 compares tool_choice against the string literal "any" (not a TS any type). - config/quality/test-masking-allowlist.json: allowlist #4384's opencode combos net-assert reduction (obsolete combo/ namespace removed). No production behavior change. * chore(release): re-trigger full CI for v3.8.31 finalization Force a fresh pull_request CI run on the head carrying the cycle-drift fixes (gemini #4373 tests, any-budget #4389, masking allowlist #4384) — the prior synchronize event did not spawn a ci.yml run. * chore: re-trigger CI (no Actions runs registered for prior push) --------- Co-authored-by: Xiangzhe <32761048+xz-dev@users.noreply.github.com> Co-authored-by: hydraromania <252583922+hydraromania@users.noreply.github.com> Co-authored-by: Bian-Sh <24520547+Bian-Sh@users.noreply.github.com> Co-authored-by: thaitryhand <248103256+thaitryhand@users.noreply.github.com> Co-authored-by: xxy9468615 <63351664+xxy9468615@users.noreply.github.com> Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> Co-authored-by: DNNYF <74033321+DNNYF@users.noreply.github.com> Co-authored-by: mugnimaestra <13349159+mugnimaestra@users.noreply.github.com> Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> Co-authored-by: nguyenvanhuy0612 <57367674+nguyenvanhuy0612@users.noreply.github.com> Co-authored-by: codename-zen <263238141+codename-zen@users.noreply.github.com> Co-authored-by: Anton <39598727+NomenAK@users.noreply.github.com> Co-authored-by: Ibrahim Ryan <ryan@nuevanext.com> Co-authored-by: anuragg-saxenaa <anuragg.saxenaa@gmail.com> Co-authored-by: aeonframework <aeon@aeonframework.dev> Co-authored-by: Jan Leon <Jan.gaschler@gmail.com> Co-authored-by: KooshaPari <42529354+KooshaPari@users.noreply.github.com> Co-authored-by: Hernan Javier Ardila Sanchez <hjasgr@gmail.com>
* fix(streaming): cheaper-inference#1211 greedy strip omniModel tags to prevent literal \n\n artifacts - Changed regex quantifier from ? to * in combo.ts, comboAgentMiddleware.ts, and contextHandoff.ts to greedily strip all JSON-escaped newline sequences surrounding <omniModel> tags in SSE streaming chunks - Added \r to the character class for cross-platform robustness - Fixed Playwright strict-mode violation in combo-unification.spec.ts - Bumped OpenAPI version and CHANGELOG to 3.6.6 * fix: 3 bugs found during issue triage (cheaper-inference#1175, cheaper-inference#1187/cheaper-inference#1218, cheaper-inference#1202) - fix(gemini): strip VS Code JSON Schema extensions from tool schemas (cheaper-inference#1175) Add enumDescriptions, markdownDescription, markdownEnumDescriptions, enumItemLabels and tags to UNSUPPORTED_SCHEMA_CONSTRAINTS so the Gemini sanitizer removes them before forwarding. GitHub Copilot injects these non-standard fields into tool definitions, causing Gemini to reject with 'Unknown name enumDescriptions at functionDeclarations[n].parameters'. - fix(health-check): unwrap proxy config object before passing to getAccessToken (cheaper-inference#1187 cheaper-inference#1218) resolveProxyForConnection() returns { proxy, level, levelId } but the health check loop was passing the full wrapper to getAccessToken(), which expects the inner config object (.host, .port etc). The proxy dispatcher validated .host on the wrapper (undefined) and threw 'Context proxy host is required', silently marking every connection as unhealthy every sweep. Fix mirrors the pattern already used in chatHelpers.ts: proxyResult?.proxy || null. - fix(ui): debounce models.dev sync interval slider to save only on release (cheaper-inference#1202) The slider's onChange fired updateInterval() on every drag tick, sending a PATCH per pixel of movement. Rapid API responses overwrote UI state mid-drag. Introduce draftIntervalHours for smooth visual feedback; the PATCH fires on onMouseUp / onBlur once the user releases the control. * fix(providers): update Xiaomi MiMo token-plan endpoints (cheaper-inference#1238) Integrated into release/v3.6.6 * fix(cc-compatible): trim beta flags and preserve cache passthrough (cheaper-inference#1230) Integrated into release/v3.6.6 * feat(memory+skills): full-featured memory & skills systems with tests (cheaper-inference#1228) Integrated into release/v3.6.6 * fix: forward client x-initiator header to GitHub Copilot upstream (cheaper-inference#1227) Integrated into release/v3.6.6 * feat(bailian-quota): add Alibaba Coding Plan quota monitoring (cheaper-inference#1235) * fix: resolve v3.6.6 backlog bugs (cheaper-inference#1206, cheaper-inference#1211, cheaper-inference#1220, cheaper-inference#1231) - fix(core): cheaper-inference#1206 inject startup guard against app/ and src/app/ conflict - fix(health): cheaper-inference#1220 add HEALTHCHECK_STAGGER_MS to prevent token refresh bursting - fix(proxy): cheaper-inference#1231 prioritize HTTP 429 over quota body heuristics - fix(sse): cheaper-inference#1211 strip leading double-newlines in responses API stream * fix(tests): resolve memory migration and skills route pagination bugs from PR overlaps * docs: Update CHANGELOG.md with v3.6.6 features (cheaper-inference#1182, cheaper-inference#1165, cheaper-inference#1177) * chore(release): bump version to 3.6.6 Update package versions for the electron app and open-sse package. Sync llm.txt metadata and feature headings with the 3.6.6 release. * feat(core): harden outbound provider calls and add cooldown retries Add guarded outbound fetch helpers with private/local URL blocking, controlled retries, timeout normalization, and route-level status propagation for provider validation and model discovery. Introduce cooldown-aware chat retries with configurable requestRetry and maxRetryIntervalSec settings, model-scoped cooldown responses, and improved rate-limit learning from headers and error bodies so short upstream lockouts can recover automatically. Also align Antigravity and Codex header handling, require API keys for Pollinations, validate web runtime env at startup, restore sanitized Gemini tool names in translated responses, and inject a synthetic Claude text block when upstream SSE completes empty. * feat(models): add glmt preset and hybrid token counting Introduce GLM Thinking as a first-class provider preset with shared GLM model metadata, pricing, usage sync, dashboard support, and provider request defaults for higher token budgets and longer timeouts. Use provider-side /messages/count_tokens when a Claude-compatible upstream supports it, while preserving estimated fallback behavior for missing models, missing credentials, and upstream failures. Also add startup seeding for default model aliases and normalize common cross-proxy model dialects so canonical slashful model ids do not get misrouted during resolution. * feat(api): add sync tokens and v1 websocket bridge Add dedicated sync token storage, issuance, revocation, and bundle download routes backed by stable config bundle versioning and ETag support. Expose the v1 websocket handshake route and custom Next server bridge so OpenAI-compatible websocket traffic can be upgraded and proxied through the dashboard and API bridge. Expand compliance auditing with structured metadata, pagination, request context, auth and provider credential events, and SSRF-blocked validation logging. * docs: Update all documentation for v3.6.6 - CHANGELOG: Add WebSocket bridge, GLM Thinking preset, safe outbound fetch/SSRF guard, cooldown-aware retries, compliance audit v2, model alias seeding, and all Internal Improvements for the 3 new commits - README: Expand v3.6.x highlights table with 10 new features; add SafeOutboundFetch, CooldownAwareRetry, SSRF guard, TPS metric, sync tokens, WebSocket bridge to Resilience/Observability/Deployment tables - ARCHITECTURE: Bump date; add new modules to executive summary, API routes, SSE core services, Auth/Security section; add SSRF/Outbound guard failure mode (section 6); expand module mapping - ENVIRONMENT: Add OMNIROUTE_CRYPT_KEY/OMNIROUTE_API_KEY_BASE64 legacy aliases, OUTBOUND_SSRF_GUARD_ENABLED, CODEX_CLIENT_VERSION, and REQUEST_RETRY/MAX_RETRY_INTERVAL_SEC cooldown retry settings - FEATURES: Add 6 new feature sections — V1 WebSocket Bridge, Sync Tokens & Config Bundle, GLM Thinking Preset, Safe Outbound Fetch & SSRF Guard, Cooldown-Aware Retries, Compliance Audit v2 * fix: use api64 for proxy test (cheaper-inference#1255) Integrated into release/v3.6.6 — IPv6 proxy test fix * fix(page): update custom models section to include all providers cheaper-inference#1200 (cheaper-inference#1256) Integrated into release/v3.6.6 — Gemini custom model picker fix * fix: provide default client_id fallbacks to prevent broken OAuth requests (cheaper-inference#1246) Integrated into release/v3.6.6 — OAuth client_id default fallbacks * fix: translate max_tokens/max_completion_tokens → max_output_tokens in Chat→Responses translator (cheaper-inference#1245) Integrated into release/v3.6.6 — max_tokens → max_output_tokens Responses API translation + unit tests * feat(oauth): support cursor-agent CLI as Cursor credential source (cheaper-inference#1258) Integrated into release/v3.6.6 — cursor-agent CLI credential source support * fix(cc-compatible): restore upstream SSE and correct stream/combo timeout behavior (cheaper-inference#1257) Integrated into release/v3.6.6 — CC-compatible upstream SSE restore + stream timeout fix + README table repair * fix(cli-tools): resolve API key resolution and model mapping bugs in CLI tools (cheaper-inference#1263) Integrated into release/v3.6.6 * feat(cli-tools): add Qwen Code CLI integration (cheaper-inference#1266) Integrated into release/v3.6.6 * fix(i18n): add missing zh-CN translations and fix logger imports (cheaper-inference#1269) Integrated into release/v3.6.6 * fix(i18n): add Chinese i18n support to dashboard components (cheaper-inference#1274) Integrated into release/v3.6.6 * feat: update Pollinations to require API key, remove free tier flag (cheaper-inference#1177) * feat: friendly error messages for crypto/encryption failures (cheaper-inference#1165) * feat: add TPS (tokens per second) metric column to request logs (cheaper-inference#1182) * feat: merge custom/imported models into filter list for all providers (cheaper-inference#1191) * feat(fallback): Fix provider-profile-driven lockouts (cheaper-inference#1267) This integrates rdself's unify-provider-profile-locks PR manually to handle structural conflicts. * fix(claude): proper Anthropic SDK integration (cheaper-inference#1271) * fix(healthcheck): use correct proxy wrapper format for getAccessToken (cheaper-inference#1272) * chore(release): v3.6.6 — skills registry stability fix + final integration * fix(auth): harden bootstrap auth and memory dashboard behavior Restrict unauthenticated writes to /api/settings/require-login to the initial bootstrap window while keeping read-only checks public. This prevents post-setup config changes without blocking first-run login setup, and the onboarding flow now logs in immediately after setting the password. Restore memory API filtering and pagination behavior by supporting q searches, honoring offset-based requests, and avoiding unrelated fallback results when FTS misses. Update dashboard stats fallback to use the response totals consistently. Package the MCP server with explicit file entries and add regression tests for bootstrap auth and memory route behavior * fix(codex): remove max_output_tokens from body for compatibility * chore(release): v3.6.6 — include PR 1274 fixes in changelog * chore: exclude additional build artifacts and internal directories from npm package distribution * fix: update Gemini OAuth test to match registry defaults + codex UI improvements * fix: restore .mjs refs for scripts/ in test imports after ts migration * fix: restore next.config.mjs ref in dev-origins test * fix: implement db migration safety checks and codex config format * fix: disable mass-migration abort during unit tests based on auto-backup flag * fix: update script regex in auto-update tests to use .mjs * feat: Add Perplexity Web (Session) provider (cheaper-inference#1289) Integrated into release/v3.6.6 * fix(cli): resolve codex routing config parsing, standardize select model button positioning, and clarify oauth documentation * docs(changelog): record recent cli, provider, and test updates Document the latest fixes for Codex routing configuration parsing and Lobehub provider icon fallback behavior. Add the note that the remaining JavaScript test files were migrated to TypeScript ES modules to reflect the completed test stack transition. * chore(release): merge cheaper-inference#1286 minor improvements manually to avoid testing conflict * chore(test): rename perplexity-web.test.mjs to .ts to maintain 100% TS codebase * chore(docs): update CHANGELOG.md for perplexity-web provider * fix(security): resolve CodeQL incomplete URL substring sanitization via URL parsing in test mocks * fix: integrate compressContext() into chatCore.ts request pipeline Proactively compress oversized contexts before sending to upstream providers, preventing context_length_exceeded errors. Compression triggers at 85% of model's context limit using the existing 3-layer compressContext() function. - Import compressContext, estimateTokens, getTokenLimit from contextManager - Add compression check after translation, before executor dispatch - Estimate tokens and compare against 85% threshold of model's context limit - Apply 3-layer compression (trim tools, compress thinking, purify history) - Log compression events with before/after token counts and layers applied - Audit compression events for observability - Add unit tests verifying integration behavior Closes cheaper-inference#1290 * fix(tests): align reasoning expectations with GLM thinking structure * fix: prevent orphaned tool_result messages in purifyHistory() When purifyHistory() drops oldest messages to fit context window, it can split tool_use/tool_result pairs — keeping the tool_result but dropping the tool_use that initiated it. This causes upstream providers to reject the request with format errors. Add fixToolPairs() that runs after each purification pass to remove: - OpenAI format: orphaned role='tool' messages without matching tool_calls ID - Claude format: orphaned tool_result content blocks without matching tool_use ID Closes cheaper-inference#1291 * fix(tests): supply tool_use in mock so it is not dropped * chore: convert remaining test to TypeScript * fix(tests): restore compatibility with compressContext threshold test after tsx migration * docs: finalize v3.6.6 release documentation * fix(core): finalize provider removal, type issues, and codex API key config * fix(dashboard): render Web/Cookie, Search, Audio provider sections and fix TypeScript errors * fix: increase MCP web_search timeout to 60s (cheaper-inference#1278) * fix: route combo testing properly for embedding models (cheaper-inference#1260) * fix: accumulate excluded accounts in combo fallback loop (cheaper-inference#1233) * fix: strip leading whitespace and newlines from first streaming chunk (cheaper-inference#1211) * docs: clarify VPS and Docker settings for OAuth credentials (cheaper-inference#1204) * fix: return real retry-after for pipeline gates (cheaper-inference#1301) Integrated into release/v3.6.6 — returns real Retry-After values from pipeline gates * feat: streaming semantic cache, Cursor auto-version detection, and call-log enhancements (cheaper-inference#1296) Integrated into release/v3.6.6 — streaming semantic cache, Cursor auto-version detection, call-log cache_source tracking * feat(api): support more OpenAI types (image, embeddings, audio-transcriptions, audio-speech) (cheaper-inference#1297) Integrated into release/v3.6.6 — adds embeddings, audio-transcriptions, audio-speech, and images-generations support for custom OpenAI-compatible providers, plus Pollinations image registry * deps: bump hono from 4.12.12 to 4.12.14 (cheaper-inference#1302) Integrated into release/v3.6.6 * deps: bump hono from 4.12.12 to 4.12.14 (cheaper-inference#1306) Integrated into release/v3.6.6 * chore: stabilization fixes for v3.6.6 (cheaper-inference#1298, cheaper-inference#1254, cheaper-inference#59, CI) * fix(providers): match correct endpoint for Xiaomi MiMo, strip routing prefix for custom openai endpoints (cheaper-inference#1303, cheaper-inference#1261) * feat(storage): add database backup cleanup controls * chore(release): v3.6.6 — Final Stabilization Push * Backport call log storage refactor to release/v3.6.6 (cheaper-inference#1307) Integrated into release/v3.6.6 * deps: update dompurify to 3.4.0 to resolve CVE-XYZ (cheaper-inference#60) * test: disable sqlite auto backup in CI to resolve E2E timeout (#24481475058) * chore(docs): sync CHANGELOG for v3.6.6 with missing features and fixes * chore(release): prep v3.6.6 infrastructure and type safety fixes - Migrated legacy .mjs scripts to .ts (bin, prepublish, policies) - Resolved pre-commit strict lint (t11 budget) errors in combo.ts - Explicitly typed all TS bindings in pack-artifact policies - Updated package.json commands to run Node via tsx/esm internally - Hardened CI/CD with explicit node version 22.22.2 checks - Completed stage validations for v3.6.6 final release * chore: fix TS build errors and e2e timeouts in CI - Migrate nodeRuntimeSupport to TS interfaces avoiding implicit any - Increase visibility timeouts in skills-marketplace E2E test to 15s to bypass CI flakiness - Complete migration of .mjs scripts to .ts ensuring type safety * chore(release): sync package version 3.6.6 across workspaces * test(e2e): universally increase UI component visibility timeouts from 5s to 15s to bypass CI starvation * chore(build): inject baseUrl, paths, and types:node into MITM tsconfig within prepublish hook to fix missing types in CI check --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com> Co-authored-by: Jack <5443152+hijak@users.noreply.github.com> Co-authored-by: Randi <55005611+rdself@users.noreply.github.com> Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com> Co-authored-by: Samuel Cedric <ceds.sam@gmail.com> Co-authored-by: Max Garmash <max@37bytes.com> Co-authored-by: Markus Hartung <mail@hartmark.se> Co-authored-by: Gi99lin <74502520+Gi99lin@users.noreply.github.com> Co-authored-by: Payne <baboialex95@gmail.com> Co-authored-by: Benson K B <bensonkbmca@gmail.com> Co-authored-by: clousky2020 <33016567+clousky2020@users.noreply.github.com> Co-authored-by: Ravi Tharuma <25951435+RaviTharuma@users.noreply.github.com> Co-authored-by: oyi77 <oyi77@users.noreply.github.com> Co-authored-by: Hdsje <vovan877@gmail.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: xiaoge1688 <moyekongling@gmail.com>
…cheaper-inference#4343) * fix(cli): honor TAILSCALE_AUTHKEY for non-interactive tailscale login (port from 9router#1263) startTailscaleLogin built `tailscale up` without ever reading process.env.TAILSCALE_AUTHKEY, so a pre-authenticated / headless daemon waited for an interactive auth URL and timed out (~15s). When TAILSCALE_AUTHKEY is set it is now passed via `--auth-key=` (an argv element to spawn(binary, args) — no shell interpolation, Hard Rule cheaper-inference#13); when unset, behavior is unchanged. The arg builder is extracted into a pure exported `tailscaleUpArgs()` for testing. Reported-by: ipeterpetrus (decolua/9router#1263) Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> * chore(quality): rebaseline tailscaleTunnel.ts file-size to 1202 (cheaper-inference#1263 +13) --------- Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com>
…heaper-inference#4397) * chore(release): open v3.8.31 development cycle * fix(mitm): exact host membership in MITM hosts test (CodeQL false positive) (cheaper-inference#4386) getMitmToolHosts returns string[], so .includes(host) is Array.prototype.includes (exact membership). CodeQL's js/incomplete-url-substring-sanitization heuristic misreads it as a String.includes() URL-substring sanitization check and raises a HIGH alert. Switch to .some(h => h === host) — identical semantics, explicit intent, no flagged pattern. Surfaced post-v3.8.30 (cheaper-inference#4325) once the test landed on main. Test-only change (no runtime behavior); the suite still passes and the CodeQL re-scan on merge clears the alert. * fix(codex): request reasoning summaries (cheaper-inference#4359) Adds reasoning.summary=auto + reasoning.encrypted_content include for Codex. Thanks @xz-dev. * fix(embeddings): inject NVIDIA NIM input_type for asymmetric embed models (cheaper-inference#4341) NVIDIA NIM asymmetric embedding models (e.g. nvidia/nv-embedqa-e5-v5) reject requests without an `input_type` ("query" | "passage") with 400 "'input_type' parameter is required". The embedding registry now carries a model-level default param for the asymmetric NVIDIA model, and the embeddings handler injects a model's default params into the upstream body only when the client omitted them, leaving a client-supplied value untouched. Reported-by: hydraromania (decolua/9router#1378) Co-authored-by: hydraromania <252583922+hydraromania@users.noreply.github.com> * fix(api): migrate deprecated Codex [features].codex_hooks to [features].hooks (cheaper-inference#4342) Codex renamed the `codex_hooks` feature flag to `hooks`; recent Codex CLI versions ignore the old key and warn. When OmniRoute rewrites an existing config.toml (configure/reset Codex provider) it now renames [features].codex_hooks -> [features].hooks, preserving the value and never clobbering an already-present `hooks`, then drops the deprecated key. The migration is a no-op when the flag is absent and runs on both the POST and DELETE config paths. Reported-by: Bian-Sh (decolua/9router#1327) Co-authored-by: Bian-Sh <24520547+Bian-Sh@users.noreply.github.com> * fix(translator): drop the null flush on the same-format response path (cheaper-inference#4344) The streaming response translator's same-format fast path returned `[chunk]` unconditionally, so the end-of-stream null/flush signal (chunk === null) propagated as a literal `[null]`. Downstream this surfaced as an empty `data: null` SSE event between chunks and crashed strict clients (e.g. Factory Droid BYOK on /v1/responses). The fast path now returns `[]` for the null flush while still passing real chunks through unchanged. Reported-by: thaitryhand (decolua/9router#1052) Co-authored-by: thaitryhand <248103256+thaitryhand@users.noreply.github.com> * fix(translator): strip assistant echo fields on the OpenAI target path (Mistral 422) (cheaper-inference#4350) Strict OpenAI-compatible upstreams (e.g. mistral/codestral-latest) reject client-only assistant echo fields sent back as input with 422 extra_forbidden (the report hit messages[].assistant.reasoning_content via Codex /responses). Only reasoning_content was stripped on the OpenAI target path; the sibling fields reasoning / refusal / annotations / cache_control leaked through. They are now all dropped on the non-reasoner OpenAI target path. `audio` is intentionally preserved (OpenAI audio models reference a prior assistant audio response by id; Mistral never emits audio). Reported-by: xxy9468615 (decolua/9router#1649) Co-authored-by: xxy9468615 <63351664+xxy9468615@users.noreply.github.com> * fix(cli): honor TAILSCALE_AUTHKEY for non-interactive tailscale login (cheaper-inference#4343) * fix(cli): honor TAILSCALE_AUTHKEY for non-interactive tailscale login (port from 9router#1263) startTailscaleLogin built `tailscale up` without ever reading process.env.TAILSCALE_AUTHKEY, so a pre-authenticated / headless daemon waited for an interactive auth URL and timed out (~15s). When TAILSCALE_AUTHKEY is set it is now passed via `--auth-key=` (an argv element to spawn(binary, args) — no shell interpolation, Hard Rule cheaper-inference#13); when unset, behavior is unchanged. The arg builder is extracted into a pure exported `tailscaleUpArgs()` for testing. Reported-by: ipeterpetrus (decolua/9router#1263) Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> * chore(quality): rebaseline tailscaleTunnel.ts file-size to 1202 (cheaper-inference#1263 +13) --------- Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> * fix(dashboard): OAuth modal surfaces the real error on a non-JSON response (cheaper-inference#4351) * fix(dashboard): OAuth modal surfaces real error on non-JSON responses (port from 9router#1318) The OAuth connect/reauth modal called `await res.json()` unconditionally, so a non-JSON error response (e.g. a plain-text 500 page from a build/OAuth endpoint) threw `Unexpected token 'I'...` and hid the real failure. New shared helpers parseResponseBody / getErrorMessage (src/shared/utils/api.ts) read the body safely (JSON when JSON, raw text otherwise) and produce a clean message either way; every modal fetch site now uses them. Reported-by: DNNYF (decolua/9router#1318) Co-authored-by: DNNYF <74033321+DNNYF@users.noreply.github.com> * fix(dashboard): type OAuth modal response body as Record<string, unknown> (t11 any-budget) Switch the parseResponseBody casts from Record<string, any> to Record<string, unknown> so OAuthModal.tsx stays within its t11 explicit-any budget. getErrorMessage already takes unknown; the success paths typecheck clean under strict:false. No runtime change. --------- Co-authored-by: DNNYF <74033321+DNNYF@users.noreply.github.com> * fix(translator): accept AI SDK-style { type: image, image: data-URL } content parts (cheaper-inference#4345) * fix(translator): accept AI SDK-style { type: image, image: "data:..." } parts (port from 9router#1330) Several OpenAI-input translators only recognized images shaped as `image_url.url` (or an object with `.source`/`.url`), so an AI SDK-style content part where `image` is a bare data-URL STRING was silently dropped before reaching a vision provider (OpenCode is one affected client; the gap is generic). The OpenAI->Claude, OpenAI->Kiro and OpenAI->Gemini/Antigravity translators now parse a string `image` data URL into each provider's native image shape (Claude base64 source, Kiro images[].source.bytes, Gemini inlineData). Reported-by: mugnimaestra (decolua/9router#1330) Co-authored-by: mugnimaestra <13349159+mugnimaestra@users.noreply.github.com> * chore(quality): freeze openai-to-kiro.ts file-size at 807 (cheaper-inference#1330 +9, over 800 cap) --------- Co-authored-by: mugnimaestra <13349159+mugnimaestra@users.noreply.github.com> * fix(dashboard): show a disabled connection's last error in the row (cheaper-inference#4352) * fix(dashboard): show a disabled connection's last error in the row (port from 9router#1447) The provider card's error badge counts a disabled connection (isActive === false) that has an error — its effective status is still error/expired/unavailable — but the connection row hid the lastError text for disabled rows, so the operator saw the count without the cause. The row's error-visibility decision is extracted into shouldShowConnectionLastError() and now shows the error whenever there is one, regardless of the active toggle. Reported-by: ntdung6868 (decolua/9router#1447) Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * chore(quality): rebaseline ConnectionRow.tsx file-size to 942 (cheaper-inference#1447 +1 import) --------- Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * fix(providers): bound the OAuth connection-test probe with a timeout (cheaper-inference#4347) * fix(providers): bound OAuth connection-test probe with a timeout (port from 9router#1449) The OAuth path of "Test Connection One-by-One" called bare fetch() with no AbortController/signal, so a provider probe that accepted the socket but never responded wedged the test queue forever. Both the initial probe and the post-refresh retry are now bounded with AbortSignal.timeout(30s) — matching the API-key path's existing budget — and a timed-out probe resolves as a failure with a clear "Test timed out after 30s" message in the route's normal error shape. Reported-by: ntdung6868 (decolua/9router#1449) Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * chore(quality): rebaseline providers test route file-size to 887 (cheaper-inference#1449 + sibling cheaper-inference#1444 growth) --------- Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * fix(providers): label a deactivated account distinctly from a revoked token (cheaper-inference#4353) A Codex connection whose OAuth refresh is fully healthy but whose ChatGPT account has been deactivated by the provider gets a 401 from the upstream API. The connection test labeled that the same as a bad credential ("Token invalid or revoked" -> upstream_auth_error), so an operator could not tell a deactivated account from a revoked token. The test now reads the 401/403 body and, when it indicates account deactivation, classifies it as account_deactivated (which the dashboard already renders as "Account Deactivated"); a plain auth 401 is unchanged. Reported-by: ntdung6868 (decolua/9router#1444) Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> * fix(db): cascade-delete orphaned model aliases when a provider is removed (cheaper-inference#4348) * fix(db): cascade-delete orphaned model aliases when a provider is removed (port from 9router#1409) Deleting a custom provider removed its connections and node but left the imported model-alias rows (key=<alias>, value="<providerId>/<model>") behind, so re-importing the same provider was blocked by stale "already exists" aliases. Add a deleteModelAliasesForProvider(providerId) DB helper that drops every alias whose stored value begins with "<providerId>/", and call it from the provider-node DELETE handler so a fresh import is unblocked. Reported-by: nguyenvanhuy0612 (decolua/9router#1409) Co-authored-by: nguyenvanhuy0612 <57367674+nguyenvanhuy0612@users.noreply.github.com> * chore(quality): rebaseline models.ts file-size to 1221 (cheaper-inference#1409 + sibling cheaper-inference#1294 growth) --------- Co-authored-by: nguyenvanhuy0612 <57367674+nguyenvanhuy0612@users.noreply.github.com> * fix(api): persist max_input_tokens/max_output_tokens when adding a custom model (cheaper-inference#4349) The POST /api/provider-models handler read the rest of the body but never the two token-limit fields, and addCustomModel() had no parameter for them, so the form values were dropped on write while the DB layer and /v1/models catalog already round-trip inputTokenLimit/outputTokenLimit. Accept the two optional limits in the schema, forward them through the handler, and persist them in addCustomModel(). TDD: failing-then-passing unit test. Reported-by: codename-zen (decolua/9router#1294) Co-authored-by: codename-zen <263238141+codename-zen@users.noreply.github.com> * docs: feature-documentation catch-up (v3.8.20 → v3.8.30) (cheaper-inference#4391) One-time reconciliation of the docs with every user-facing feature shipped since v3.8.20 (we had never done a dedicated pass, so debt had accumulated): - README: new '✨ What's New' section (curated v3.8.20→v3.8.30 highlights). - New guides: CLI-INTEGRATIONS (all setup-*/launch commands), MITM-TPROXY-DECRYPT (transparent-decrypt epic), CONTEXT_EDITING (delegated Anthropic clear_tool_uses). - Refreshed: AUTO-COMBO (auto/<category>:<tier> + Arena-ELO), API_REFERENCE (x-omniroute-no-memory), MEMORY (int8 quantization + off-by-default), RESILIENCE (model-lockout success-decay), RTK, AGENTBRIDGE, TRAFFIC_INSPECTOR, GUARDRAILS, CLOUD_AGENT, ENVIRONMENT, SETUP_GUIDE, CLI-TOOLS, MCP-SERVER. - Regenerated PROVIDER_REFERENCE (231 providers); synced the count in README/CLAUDE/AGENTS. - Allowlisted external-tool env vars (OPENAI_API_BASE, PROMPTFOO_PROVIDER_KEY) and the STREAM_RECOVERY config-object name in the docs-accuracy gates. All claims source-verified; check:docs-all (sync/counts/env/links/fabricated) passes. Going forward this runs every release via generate-release step 6b. * fix(executors): don't inject thinking when tool_choice forces a tool (native Claude) (cheaper-inference#4389) Forced tool_choice now strips the adaptive thinking injection to avoid Anthropic 400. Thanks @NomenAK. * fix(translator): Gemini accepts HTTP/HTTPS image URLs (port from 9router#344) (cheaper-inference#4373) OpenAI-style `image_url` parts with an `http://` or `https://` URL reached `convertOpenAIContentToParts` and were dropped with only a `console.warn`, because Gemini's `inlineData` requires base64 (the helper is synchronous and cannot fetch+encode upstream assets). Gemini's `Part` schema, however, natively accepts `fileData: { fileUri }` for remote URIs — the model fetches the asset itself. The helper now emits a `fileData` part (`mimeType: "image/*"`, inferred upstream on fetch) for HTTP/HTTPS URLs instead of silently dropping them. Vision requests that pass a URL — not a data: URI — now reach Gemini intact. No behavioral change for: - `data:` URIs → still emitted as `inlineData` with the parsed media type. - Unsupported schemes (e.g. `ftp:`) → still skipped (Gemini would reject them). The openai-to-claude side already passed HTTP/HTTPS URLs through as `source: { type: "url", url }` (lines 573–578) — the upstream PR's Claude-side change was already covered. Regression test: tests/unit/gemini-helper-http-image-url-port344.test.ts (4 cases: https URL, http URL, data: URI no-regression, unsupported-scheme guard). Inspired-by: decolua/9router#344 Co-authored-by: Ibrahim Ryan <ryan@nuevanext.com> * fix(executors): strip stream_options for qwen non-streaming / thinking Claude Code requests (port from 9router#663) (cheaper-inference#4374) Claude-Code-compatible providers force the executor-level `stream` flag on via `upstreamStream = stream || isClaudeCodeCompatible` (open-sse/handlers/chatCore.ts), but the outgoing body keeps the caller's original `stream: false`. The shared `stream && targetFormat === "openai"` branch in DefaultExecutor.transformRequest then injected `stream_options: { include_usage: true }` onto a body that still said `stream: false`, and qwen upstream rejected the request with `400 "'stream_options' only set this when you set stream: true"`. The same rejection surfaced when the body carried `thinking` / `enable_thinking`. The qwen branch now skips the injection (and strips any client-sent `stream_options`) when the body explicitly says `stream: false` or requests thinking, leaving regular qwen streaming requests with the include_usage injection intact. Other providers are unaffected. Adds a TDD regression with 4 cases covering both opt-out paths and the normal-streaming positive control. Inspired-by: decolua/9router#663 Co-authored-by: anuragg-saxenaa <anuragg.saxenaa@gmail.com> * fix(security): scope OAuth callback postMessage to a trusted-origin allowlist (port from 9router#998) (cheaper-inference#4372) The OAuth callback at `/callback` previously fell back to `window.opener.postMessage({ code, state, ... }, "*")` whenever the opener was cross-origin. The fallback was intended to support remote-OmniRoute + local-loopback callbacks (where opener and callback live on different origins), but the same code path also delivers the OAuth code/state to any hostile opener that pops the well-known callback URL — letting that attacker complete the OAuth flow as the user. Replace the wildcard fallback with iteration over a fixed allowlist: `window.location.origin` (same-origin parent — the popup-mode dashboard) plus Codex's fixed loopback helper (`http://localhost:1455` and the IPv4 literal `http://127.0.0.1:1455`). The browser drops `postMessage` to any opener whose actual origin is not in `targetOrigin`, so the message reaches only known parents and is silently dropped for any other. The same-origin fallback path is unchanged — methods 2 (`BroadcastChannel`) and 3 (`localStorage` storage event) still cover same-origin openers that COOP severed. The `openerSameOrigin` probe stays in place to drive the auto-close vs manual-copy UI decision (no behavior change for the success path). Adds a regression test (`tests/unit/ui/oauth-callback-postmessage-scope.test.tsx`) that mounts the page with a stubbed cross-origin opener and asserts no `postMessage` call ever uses `"*"` and every call lands on an allowlisted origin. The test failed against the pre-fix code (red), passes after the fix (green) — TDD per CLAUDE.md hard rule cheaper-inference#18. Partial port of upstream decolua/9router#998: the upstream PR also re-enabled TLS verification on a DNS-bypass fetch in `open-sse/utils/proxyFetch.js`; that part is N/A here because OmniRoute's `proxyFetch.ts` never disabled TLS verification (no `rejectUnauthorized: false` anywhere in the file). Inspired-by: decolua/9router#998 Co-authored-by: aeonframework <aeon@aeonframework.dev> * fix(sse): default combo per-target timeout to 120s for fast failover (cheaper-inference#4365) Combo per-target timeout inherited the full FETCH_TIMEOUT_MS (600s) when a combo did not set its own targetTimeoutMs, so a single hung/slow target stalled the whole combo for up to 10 minutes before falling through to the next model. Introduce DEFAULT_COMBO_TARGET_TIMEOUT_MS (120s) as the unset-default in resolveComboTargetTimeoutMs (new 3rd arg) and wire it in phaseComboSetup. The upstream ceiling (600s) and per-combo opt-out (targetTimeoutMs, up to the ceiling) are preserved; single non-combo requests are unchanged. For streaming requests this only bounds time-to-first-headers, so token generation is not cut short. TDD: failing-then-passing unit test in tests/unit/combo-config.test.ts. * refactor(combo): de-dup exhausted-target skip predicate across both dispatchers (cheaper-inference#4362) Primeiro incremento da de-dup dos 2 dispatchers de combo (handleComboChat + handleRoundRobinCombo). O bloco de pre-check cheaper-inference#1731/#1731v2 (skip de target já exhausted no provider/connection) era BYTE-IDÊNTICO nos dois (mesmas condições, mesmas mensagens), diferindo só na tag de log e no control-flow. - comboPredicates.ts: getExhaustedTargetSkipReason(target, exhaustedProviders, exhaustedConnections) — predicate PURO que retorna a mensagem de skip (ou null); cada dispatcher mantém seu próprio log-tag + control-flow (return null / continue) + fallbackCount. No mutate do stryker (cobertura de mutação). - combo.ts: −20 linhas (os 2 blocos viram 1 chamada cada). - 7 testes de caracterização travam condições + strings exatas. Comportamento preservado: 376/376 testes combo (357 caracterização + 7 novos), integração sse-correctness 5/5, typecheck 0, complexity neutro (1895), file-size encolhe. Próximo incremento: de-dup do error-handling/exhausted-tracking (handleTargetError). * refactor(combo): de-dup upstream-error exhaustion classification across both dispatchers (cheaper-inference#4366) Segundo incremento da de-dup dos 2 dispatchers (handleTargetError). Após cada erro de target, ambos rodavam um bloco quase-idêntico que marca o provider exhausted (cheaper-inference#1731), a conexão connection-errored (#1731v2) ou o provider transiently rate-limited. - combo/targetExhaustion.ts: applyComboTargetExhaustion(target, opts) — atualiza os 3 Sets de exhaustion e retorna providerExhausted. As MUTAÇÕES de Set (que dirigem o skip de targets, lidas por getExhaustedTargetSkipReason) são BYTE-IDÊNTICAS nos dois; as diferenças reais viram parâmetros: tag, allAccountsRateLimited (termo extra do RR, false no handleComboChat), exhaustedLogLevel (info no handleComboChat, debug no RR). Connection-level extraído p/ markConnectionLevelExhaustion (privado, <15 complexity). - combo.ts: −73 linhas; 4 imports órfãos removidos. - 7 testes de caracterização travam as mutações + o return. ÚNICA mudança de comportamento: o WORDING das mensagens de log do RR ganha o sufixo 'on remaining targets' (cosmético; mesmo #code, mesmas mutações, mesmos níveis de log). 376/376 combo (caracterização preservada), integração sse 5/5, typecheck 0, complexity neutro (1895), file-size encolhe. * refactor(chatCore): extract checkHeapPressureGuard leaf (god-file decomposition start) (cheaper-inference#4371) Primeiro incremento da decomposição do chatCore.ts (5127 LOC, hot-path mais quente). O guard de memória do topo do handleChatCore (rejeita 503 quando o heap V8 passa o threshold de shed) vira um leaf testável, co-locado com o threshold em heapPressure.ts. - heapPressure.ts: checkHeapPressureGuard(heapUsedMb, thresholdMb) — retorna o result 503 pronto ou null. Byte-idêntico ao guard inline (mesmo check, mesma 503, mesmo warn). A figura de heap fica em telemetria INTERNA, nunca no response do cliente (Hard Rule cheaper-inference#12). - chatCore.ts: o bloco inline (~22 ln) vira 3 linhas; import órfão de HEAP_PRESSURE_THRESHOLD_MB trocado por checkHeapPressureGuard. - 3 testes novos (incl. assert Rule cheaper-inference#12: o MB medido não vaza no payload). complexity-baseline 1895->1896: drift de base pós-cheaper-inference#4338 (medido com minhas mudanças stashed = 1896); esta mudança é complexity-NEUTRA (helper complexity 2, handleChatCore só perde código). 190/190 chatcore tests, typecheck 0, file-size encolhe. * Localize CLI and stabilize fetch, memory, and coverage handling (cheaper-inference#4383) en-only i18n, fetch-start-timeout hardening, EngineConfigPage icon fix, CI build-artifact-reuse overhaul. Memory production hunk dropped as a no-op (tests kept). Thanks @JxnLexn. * test(combo): reset circuit breakers between stream-readiness cases (restore green) (cheaper-inference#4396) The combo-dispatch cases in combo-stream-readiness-fallback.test.ts deliberately fail `glm` (zombie streams / repeated 504s), which legitimately trips the per-provider circuit breaker. That OPEN state is a module-level singleton, so it leaked into the next test and combo.ts then SKIPPED `glm/*` targets entirely ("Skipping … circuit breaker OPEN"). That made "combo does not retry stream readiness timeouts on the same model" never attempt glm/zombie — expected ['glm/zombie','openai/gpt-5.4-mini'] but got ['openai/gpt-5.4-mini']. This was a pre-existing red on release/v3.8.31 (present at the cycle-open tip), order-dependent: the test passes in isolation, fails after the preceding cases. Add a test.beforeEach(resetAllCircuitBreakers) so each scenario starts from a clean breaker slate. Test-isolation only — the breaker behavior is correct and no production code or assertion changes. Full combo suite: 390/390 green. * fix(cli): decline confirm() cleanly on non-interactive stdin + document contexts workflow The `contexts remove` command already has `--yes` to skip confirmation, but when run without it under a non-interactive stdin (pipe, CI, EOF) the [y/N] prompt could never be answered — the readline question stayed pending and Node warned about an "unsettled top-level await" at exit. confirm() now detects `!process.stdin.isTTY` and declines cleanly (returns false), pointing at `--yes` for non-interactive use. Exported confirm() for a regression test. Docs: REMOTE-MODE.md gains a full "Managing contexts" section (list/current/use to switch between remote and local, add/show/rename, remove with --yes, export/import), fixes a `context current` -> `contexts current` typo, and the README remote-mode snippet now shows switching back to local. Verified against the live CLI: command signatures, --yes, and the non-TTY decline path all behave as documented. Tests: cli-contexts.test.ts asserts confirm() declines on non-TTY stdin (RED before, GREEN after). All docs gates (fabricated/links/symbols) pass. * ci(t11): bump any-budget for executors/base.ts (2 false-positive "any" strings) Unblocks the Fast Quality Gates on release/v3.8.31: `check:any-budget:t11` was red on `open-sse/executors/base.ts` for ALL PRs (pre-existing base drift, unrelated to this branch). The checker counts `\bany\b` after stripping comments but NOT strings, and the native-Claude tool_choice logic uses the API value `"any"` in two string literals (`tb.tool_choice === "any"`, `.type === "any"`). There are zero actual TypeScript `any` types in the file — budget set to the matched count, mirroring the existing cursor.ts false-positive entry right below it. * perf: combos UI split + next config + 1-click redis + bifrost sidecar (cheaper-inference#3932) (cheaper-inference#4381) Combos UI split + next.config perf + 1-click local Redis launcher + bifrost relay. Review fixes (co-author): --rm/--restart conflict, error sanitization, UI/route names, dead-guard re-doc, bifrost Zod, IPv6 + CLI test fixes. Thanks @KooshaPari. * fix(plugin): prefix OC static-catalog combo+raw keys with providerId (cheaper-inference#4384) OC parses model ids on '/'; combo keys now carry 'omniroute/' (was 'combo/'). Live-validated against the VPS via OpenCode. Thanks @herjarsa. * docs(env): document TAILSCALE_AUTHKEY (env/docs contract drift on .31) Second pre-existing base-drift fix needed to get Fast Quality Gates green: the `repository contract is in sync` test (check:env-doc-sync) was red on ALL .31 PRs. PR cheaper-inference#4343 added a code reference to `process.env.TAILSCALE_AUTHKEY` (src/lib/tailscaleTunnel.ts) for non-interactive `tailscale up`, but never added the var to .env.example / docs/reference/ENVIRONMENT.md — the contract requires code vars to appear in both. Add the (commented) entry to .env.example next to TAILSCALE_BIN and a row to ENVIRONMENT.md. Verified: `node scripts/check/check-env-doc-sync.mjs` → in sync. Unrelated to this branch's confirm()/docs change; surfaced because touching a quality script triggers the TIA fail-safe full unit suite. * chore(release): v3.8.31 — 2026-06-20 Finalize the v3.8.31 release: reconcile the CHANGELOG (full commit-to-bullet coverage, 26 bullets across Features/Fixed/Security/Maintenance), refresh the README What's New section, back-fill the .30/.31 sections into the 41 i18n CHANGELOG mirrors, and add TAILSCALE_AUTHKEY to the env contract (.env.example + ENVIRONMENT.md). * chore(release): align mitm-hosts test comment with main to clear merge conflict The same CodeQL false-positive fix landed twice — cheaper-inference#4386 on release/v3.8.31 and cheaper-inference#4387 directly on main — with only the explanatory comment differing (the assertion is byte-identical). Adopt main's comment wording on the release branch so the release PR merges without a comment-only conflict. * test(translator): align stale openai->gemini remote-URL tests with cheaper-inference#4373 Third pre-existing base-drift fix to get Fast Quality Gates green on .31. PR cheaper-inference#4373 ('Gemini accepts HTTP/HTTPS image URLs', port of 9router#344) intentionally changed convertOpenAIContentToParts so remote http(s) image URLs pass through as a native `fileData: { fileUri }` part instead of the old cheaper-inference#2807 drop+warn — and added its own test (gemini-helper-http-image-url-port344.test.ts) for the new behavior. But it left three tests in translator-openai-to-gemini.test.ts asserting the OLD drop+warn contract, so they fail deterministically (verified: the file fails in isolation on clean .31). These only surface under the TIA fail-safe FULL suite, which a quality- script touch triggers. Align the three stale tests to the real, intended behavior (captured by running the function): remote URLs -> fileData.fileUri (mimeType image/*), still never inlineData (the sync path cannot fetch+encode). This is test-vs-code alignment to a deliberate, separately-tested change — not a weakened assertion. 40/40 in the file; 94/94 across the translator + env-doc combo that previously failed. * chore(quality): reconcile complexity baseline 1896->1900 (/review-prs v3.8.31 batch) (cheaper-inference#4410) * fix(release): reconcile full-CI drift for v3.8.31 (gemini tests cheaper-inference#4373, any-budget cheaper-inference#4389, masking allowlist cheaper-inference#4384) The release PR's full CI surfaced cumulative cycle drift the per-PR fast gates skip: - tests/unit/translator-openai-to-gemini.test.ts: realign 3 cases to cheaper-inference#4373's HTTP/HTTPS-URL fileData pass-through (they asserted the old warn-and-drop). - scripts/check/check-t11-any-budget.mjs: base.ts budget 0→2 — cheaper-inference#4389 compares tool_choice against the string literal "any" (not a TS any type). - config/quality/test-masking-allowlist.json: allowlist cheaper-inference#4384's opencode combos net-assert reduction (obsolete combo/ namespace removed). No production behavior change. * chore(release): re-trigger full CI for v3.8.31 finalization Force a fresh pull_request CI run on the head carrying the cycle-drift fixes (gemini cheaper-inference#4373 tests, any-budget cheaper-inference#4389, masking allowlist cheaper-inference#4384) — the prior synchronize event did not spawn a ci.yml run. * chore: re-trigger CI (no Actions runs registered for prior push) --------- Co-authored-by: Xiangzhe <32761048+xz-dev@users.noreply.github.com> Co-authored-by: hydraromania <252583922+hydraromania@users.noreply.github.com> Co-authored-by: Bian-Sh <24520547+Bian-Sh@users.noreply.github.com> Co-authored-by: thaitryhand <248103256+thaitryhand@users.noreply.github.com> Co-authored-by: xxy9468615 <63351664+xxy9468615@users.noreply.github.com> Co-authored-by: ipeterpetrus <93033698+ipeterpetrus@users.noreply.github.com> Co-authored-by: DNNYF <74033321+DNNYF@users.noreply.github.com> Co-authored-by: mugnimaestra <13349159+mugnimaestra@users.noreply.github.com> Co-authored-by: ntdung6868 <103993527+ntdung6868@users.noreply.github.com> Co-authored-by: nguyenvanhuy0612 <57367674+nguyenvanhuy0612@users.noreply.github.com> Co-authored-by: codename-zen <263238141+codename-zen@users.noreply.github.com> Co-authored-by: Anton <39598727+NomenAK@users.noreply.github.com> Co-authored-by: Ibrahim Ryan <ryan@nuevanext.com> Co-authored-by: anuragg-saxenaa <anuragg.saxenaa@gmail.com> Co-authored-by: aeonframework <aeon@aeonframework.dev> Co-authored-by: Jan Leon <Jan.gaschler@gmail.com> Co-authored-by: KooshaPari <42529354+KooshaPari@users.noreply.github.com> Co-authored-by: Hernan Javier Ardila Sanchez <hjasgr@gmail.com>
* fix(streaming): cheaper-inference#1211 greedy strip omniModel tags to prevent literal \n\n artifacts - Changed regex quantifier from ? to * in combo.ts, comboAgentMiddleware.ts, and contextHandoff.ts to greedily strip all JSON-escaped newline sequences surrounding <omniModel> tags in SSE streaming chunks - Added \r to the character class for cross-platform robustness - Fixed Playwright strict-mode violation in combo-unification.spec.ts - Bumped OpenAPI version and CHANGELOG to 3.6.6 * fix: 3 bugs found during issue triage (cheaper-inference#1175, cheaper-inference#1187/cheaper-inference#1218, cheaper-inference#1202) - fix(gemini): strip VS Code JSON Schema extensions from tool schemas (cheaper-inference#1175) Add enumDescriptions, markdownDescription, markdownEnumDescriptions, enumItemLabels and tags to UNSUPPORTED_SCHEMA_CONSTRAINTS so the Gemini sanitizer removes them before forwarding. GitHub Copilot injects these non-standard fields into tool definitions, causing Gemini to reject with 'Unknown name enumDescriptions at functionDeclarations[n].parameters'. - fix(health-check): unwrap proxy config object before passing to getAccessToken (cheaper-inference#1187 cheaper-inference#1218) resolveProxyForConnection() returns { proxy, level, levelId } but the health check loop was passing the full wrapper to getAccessToken(), which expects the inner config object (.host, .port etc). The proxy dispatcher validated .host on the wrapper (undefined) and threw 'Context proxy host is required', silently marking every connection as unhealthy every sweep. Fix mirrors the pattern already used in chatHelpers.ts: proxyResult?.proxy || null. - fix(ui): debounce models.dev sync interval slider to save only on release (cheaper-inference#1202) The slider's onChange fired updateInterval() on every drag tick, sending a PATCH per pixel of movement. Rapid API responses overwrote UI state mid-drag. Introduce draftIntervalHours for smooth visual feedback; the PATCH fires on onMouseUp / onBlur once the user releases the control. * fix(providers): update Xiaomi MiMo token-plan endpoints (cheaper-inference#1238) Integrated into release/v3.6.6 * fix(cc-compatible): trim beta flags and preserve cache passthrough (cheaper-inference#1230) Integrated into release/v3.6.6 * feat(memory+skills): full-featured memory & skills systems with tests (cheaper-inference#1228) Integrated into release/v3.6.6 * fix: forward client x-initiator header to GitHub Copilot upstream (cheaper-inference#1227) Integrated into release/v3.6.6 * feat(bailian-quota): add Alibaba Coding Plan quota monitoring (cheaper-inference#1235) * fix: resolve v3.6.6 backlog bugs (cheaper-inference#1206, cheaper-inference#1211, cheaper-inference#1220, cheaper-inference#1231) - fix(core): cheaper-inference#1206 inject startup guard against app/ and src/app/ conflict - fix(health): cheaper-inference#1220 add HEALTHCHECK_STAGGER_MS to prevent token refresh bursting - fix(proxy): cheaper-inference#1231 prioritize HTTP 429 over quota body heuristics - fix(sse): cheaper-inference#1211 strip leading double-newlines in responses API stream * fix(tests): resolve memory migration and skills route pagination bugs from PR overlaps * docs: Update CHANGELOG.md with v3.6.6 features (cheaper-inference#1182, cheaper-inference#1165, cheaper-inference#1177) * chore(release): bump version to 3.6.6 Update package versions for the electron app and open-sse package. Sync llm.txt metadata and feature headings with the 3.6.6 release. * feat(core): harden outbound provider calls and add cooldown retries Add guarded outbound fetch helpers with private/local URL blocking, controlled retries, timeout normalization, and route-level status propagation for provider validation and model discovery. Introduce cooldown-aware chat retries with configurable requestRetry and maxRetryIntervalSec settings, model-scoped cooldown responses, and improved rate-limit learning from headers and error bodies so short upstream lockouts can recover automatically. Also align Antigravity and Codex header handling, require API keys for Pollinations, validate web runtime env at startup, restore sanitized Gemini tool names in translated responses, and inject a synthetic Claude text block when upstream SSE completes empty. * feat(models): add glmt preset and hybrid token counting Introduce GLM Thinking as a first-class provider preset with shared GLM model metadata, pricing, usage sync, dashboard support, and provider request defaults for higher token budgets and longer timeouts. Use provider-side /messages/count_tokens when a Claude-compatible upstream supports it, while preserving estimated fallback behavior for missing models, missing credentials, and upstream failures. Also add startup seeding for default model aliases and normalize common cross-proxy model dialects so canonical slashful model ids do not get misrouted during resolution. * feat(api): add sync tokens and v1 websocket bridge Add dedicated sync token storage, issuance, revocation, and bundle download routes backed by stable config bundle versioning and ETag support. Expose the v1 websocket handshake route and custom Next server bridge so OpenAI-compatible websocket traffic can be upgraded and proxied through the dashboard and API bridge. Expand compliance auditing with structured metadata, pagination, request context, auth and provider credential events, and SSRF-blocked validation logging. * docs: Update all documentation for v3.6.6 - CHANGELOG: Add WebSocket bridge, GLM Thinking preset, safe outbound fetch/SSRF guard, cooldown-aware retries, compliance audit v2, model alias seeding, and all Internal Improvements for the 3 new commits - README: Expand v3.6.x highlights table with 10 new features; add SafeOutboundFetch, CooldownAwareRetry, SSRF guard, TPS metric, sync tokens, WebSocket bridge to Resilience/Observability/Deployment tables - ARCHITECTURE: Bump date; add new modules to executive summary, API routes, SSE core services, Auth/Security section; add SSRF/Outbound guard failure mode (section 6); expand module mapping - ENVIRONMENT: Add OMNIROUTE_CRYPT_KEY/OMNIROUTE_API_KEY_BASE64 legacy aliases, OUTBOUND_SSRF_GUARD_ENABLED, CODEX_CLIENT_VERSION, and REQUEST_RETRY/MAX_RETRY_INTERVAL_SEC cooldown retry settings - FEATURES: Add 6 new feature sections — V1 WebSocket Bridge, Sync Tokens & Config Bundle, GLM Thinking Preset, Safe Outbound Fetch & SSRF Guard, Cooldown-Aware Retries, Compliance Audit v2 * fix: use api64 for proxy test (cheaper-inference#1255) Integrated into release/v3.6.6 — IPv6 proxy test fix * fix(page): update custom models section to include all providers cheaper-inference#1200 (cheaper-inference#1256) Integrated into release/v3.6.6 — Gemini custom model picker fix * fix: provide default client_id fallbacks to prevent broken OAuth requests (cheaper-inference#1246) Integrated into release/v3.6.6 — OAuth client_id default fallbacks * fix: translate max_tokens/max_completion_tokens → max_output_tokens in Chat→Responses translator (cheaper-inference#1245) Integrated into release/v3.6.6 — max_tokens → max_output_tokens Responses API translation + unit tests * feat(oauth): support cursor-agent CLI as Cursor credential source (cheaper-inference#1258) Integrated into release/v3.6.6 — cursor-agent CLI credential source support * fix(cc-compatible): restore upstream SSE and correct stream/combo timeout behavior (cheaper-inference#1257) Integrated into release/v3.6.6 — CC-compatible upstream SSE restore + stream timeout fix + README table repair * fix(cli-tools): resolve API key resolution and model mapping bugs in CLI tools (cheaper-inference#1263) Integrated into release/v3.6.6 * feat(cli-tools): add Qwen Code CLI integration (cheaper-inference#1266) Integrated into release/v3.6.6 * fix(i18n): add missing zh-CN translations and fix logger imports (cheaper-inference#1269) Integrated into release/v3.6.6 * fix(i18n): add Chinese i18n support to dashboard components (cheaper-inference#1274) Integrated into release/v3.6.6 * feat: update Pollinations to require API key, remove free tier flag (cheaper-inference#1177) * feat: friendly error messages for crypto/encryption failures (cheaper-inference#1165) * feat: add TPS (tokens per second) metric column to request logs (cheaper-inference#1182) * feat: merge custom/imported models into filter list for all providers (cheaper-inference#1191) * feat(fallback): Fix provider-profile-driven lockouts (cheaper-inference#1267) This integrates rdself's unify-provider-profile-locks PR manually to handle structural conflicts. * fix(claude): proper Anthropic SDK integration (cheaper-inference#1271) * fix(healthcheck): use correct proxy wrapper format for getAccessToken (cheaper-inference#1272) * chore(release): v3.6.6 — skills registry stability fix + final integration * fix(auth): harden bootstrap auth and memory dashboard behavior Restrict unauthenticated writes to /api/settings/require-login to the initial bootstrap window while keeping read-only checks public. This prevents post-setup config changes without blocking first-run login setup, and the onboarding flow now logs in immediately after setting the password. Restore memory API filtering and pagination behavior by supporting q searches, honoring offset-based requests, and avoiding unrelated fallback results when FTS misses. Update dashboard stats fallback to use the response totals consistently. Package the MCP server with explicit file entries and add regression tests for bootstrap auth and memory route behavior * fix(codex): remove max_output_tokens from body for compatibility * chore(release): v3.6.6 — include PR 1274 fixes in changelog * chore: exclude additional build artifacts and internal directories from npm package distribution * fix: update Gemini OAuth test to match registry defaults + codex UI improvements * fix: restore .mjs refs for scripts/ in test imports after ts migration * fix: restore next.config.mjs ref in dev-origins test * fix: implement db migration safety checks and codex config format * fix: disable mass-migration abort during unit tests based on auto-backup flag * fix: update script regex in auto-update tests to use .mjs * feat: Add Perplexity Web (Session) provider (cheaper-inference#1289) Integrated into release/v3.6.6 * fix(cli): resolve codex routing config parsing, standardize select model button positioning, and clarify oauth documentation * docs(changelog): record recent cli, provider, and test updates Document the latest fixes for Codex routing configuration parsing and Lobehub provider icon fallback behavior. Add the note that the remaining JavaScript test files were migrated to TypeScript ES modules to reflect the completed test stack transition. * chore(release): merge cheaper-inference#1286 minor improvements manually to avoid testing conflict * chore(test): rename perplexity-web.test.mjs to .ts to maintain 100% TS codebase * chore(docs): update CHANGELOG.md for perplexity-web provider * fix(security): resolve CodeQL incomplete URL substring sanitization via URL parsing in test mocks * fix: integrate compressContext() into chatCore.ts request pipeline Proactively compress oversized contexts before sending to upstream providers, preventing context_length_exceeded errors. Compression triggers at 85% of model's context limit using the existing 3-layer compressContext() function. - Import compressContext, estimateTokens, getTokenLimit from contextManager - Add compression check after translation, before executor dispatch - Estimate tokens and compare against 85% threshold of model's context limit - Apply 3-layer compression (trim tools, compress thinking, purify history) - Log compression events with before/after token counts and layers applied - Audit compression events for observability - Add unit tests verifying integration behavior Closes cheaper-inference#1290 * fix(tests): align reasoning expectations with GLM thinking structure * fix: prevent orphaned tool_result messages in purifyHistory() When purifyHistory() drops oldest messages to fit context window, it can split tool_use/tool_result pairs — keeping the tool_result but dropping the tool_use that initiated it. This causes upstream providers to reject the request with format errors. Add fixToolPairs() that runs after each purification pass to remove: - OpenAI format: orphaned role='tool' messages without matching tool_calls ID - Claude format: orphaned tool_result content blocks without matching tool_use ID Closes cheaper-inference#1291 * fix(tests): supply tool_use in mock so it is not dropped * chore: convert remaining test to TypeScript * fix(tests): restore compatibility with compressContext threshold test after tsx migration * docs: finalize v3.6.6 release documentation * fix(core): finalize provider removal, type issues, and codex API key config * fix(dashboard): render Web/Cookie, Search, Audio provider sections and fix TypeScript errors * fix: increase MCP web_search timeout to 60s (cheaper-inference#1278) * fix: route combo testing properly for embedding models (cheaper-inference#1260) * fix: accumulate excluded accounts in combo fallback loop (cheaper-inference#1233) * fix: strip leading whitespace and newlines from first streaming chunk (cheaper-inference#1211) * docs: clarify VPS and Docker settings for OAuth credentials (cheaper-inference#1204) * fix: return real retry-after for pipeline gates (cheaper-inference#1301) Integrated into release/v3.6.6 — returns real Retry-After values from pipeline gates * feat: streaming semantic cache, Cursor auto-version detection, and call-log enhancements (cheaper-inference#1296) Integrated into release/v3.6.6 — streaming semantic cache, Cursor auto-version detection, call-log cache_source tracking * feat(api): support more OpenAI types (image, embeddings, audio-transcriptions, audio-speech) (cheaper-inference#1297) Integrated into release/v3.6.6 — adds embeddings, audio-transcriptions, audio-speech, and images-generations support for custom OpenAI-compatible providers, plus Pollinations image registry * deps: bump hono from 4.12.12 to 4.12.14 (cheaper-inference#1302) Integrated into release/v3.6.6 * deps: bump hono from 4.12.12 to 4.12.14 (cheaper-inference#1306) Integrated into release/v3.6.6 * chore: stabilization fixes for v3.6.6 (cheaper-inference#1298, cheaper-inference#1254, cheaper-inference#59, CI) * fix(providers): match correct endpoint for Xiaomi MiMo, strip routing prefix for custom openai endpoints (cheaper-inference#1303, cheaper-inference#1261) * feat(storage): add database backup cleanup controls * chore(release): v3.6.6 — Final Stabilization Push * Backport call log storage refactor to release/v3.6.6 (cheaper-inference#1307) Integrated into release/v3.6.6 * deps: update dompurify to 3.4.0 to resolve CVE-XYZ (cheaper-inference#60) * test: disable sqlite auto backup in CI to resolve E2E timeout (#24481475058) * chore(docs): sync CHANGELOG for v3.6.6 with missing features and fixes * chore(release): prep v3.6.6 infrastructure and type safety fixes - Migrated legacy .mjs scripts to .ts (bin, prepublish, policies) - Resolved pre-commit strict lint (t11 budget) errors in combo.ts - Explicitly typed all TS bindings in pack-artifact policies - Updated package.json commands to run Node via tsx/esm internally - Hardened CI/CD with explicit node version 22.22.2 checks - Completed stage validations for v3.6.6 final release * chore: fix TS build errors and e2e timeouts in CI - Migrate nodeRuntimeSupport to TS interfaces avoiding implicit any - Increase visibility timeouts in skills-marketplace E2E test to 15s to bypass CI flakiness - Complete migration of .mjs scripts to .ts ensuring type safety * chore(release): sync package version 3.6.6 across workspaces * test(e2e): universally increase UI component visibility timeouts from 5s to 15s to bypass CI starvation * chore(build): inject baseUrl, paths, and types:node into MITM tsconfig within prepublish hook to fix missing types in CI check --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com> Co-authored-by: Jack <5443152+hijak@users.noreply.github.com> Co-authored-by: Randi <55005611+rdself@users.noreply.github.com> Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com> Co-authored-by: Samuel Cedric <ceds.sam@gmail.com> Co-authored-by: Max Garmash <max@37bytes.com> Co-authored-by: Markus Hartung <mail@hartmark.se> Co-authored-by: Gi99lin <74502520+Gi99lin@users.noreply.github.com> Co-authored-by: Payne <baboialex95@gmail.com> Co-authored-by: Benson K B <bensonkbmca@gmail.com> Co-authored-by: clousky2020 <33016567+clousky2020@users.noreply.github.com> Co-authored-by: Ravi Tharuma <25951435+RaviTharuma@users.noreply.github.com> Co-authored-by: oyi77 <oyi77@users.noreply.github.com> Co-authored-by: Hdsje <vovan877@gmail.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: xiaoge1688 <moyekongling@gmail.com>
Summary
Fixes two critical bugs that prevented the Dashboard CLI Tools feature from writing correct settings to
~/.claude/settings.json(and other CLI tool configs).Bug #1:
keyIdstripped by Zod validationAffected files:
src/app/api/cli-tools/claude-settings/route.tssrc/app/api/cli-tools/codex-settings/route.tssrc/app/api/cli-tools/droid-settings/route.tssrc/app/api/cli-tools/cline-settings/route.tssrc/app/api/cli-tools/openclaw-settings/route.tssrc/app/api/cli-tools/kilo-settings/route.tsProblem: Zod's
safeParse()stripped thekeyIdfield before the backend could read it, so the real API key was never resolved from the database and written to the config file.Fix: Extract
keyIdfromrawBodybefore validation runs.Bug #2: Key matching logic compared full key vs masked key
Affected file:
src/app/(dashboard)/dashboard/cli-tools/components/ClaudeToolCard.tsxProblem: Frontend compared full key (
sk-31c4ea6a04196891-...) against masked key (sk-31c4****dfa6). They never matched, so the dropdown never showed the correct selected key, and model mappings got scrambled.Fix: Mask the token from file before comparing against the masked API keys list.
Testing
~/.claude/settings.jsonwhen clicking Apply