mirror of
https://github.com/diegosouzapw/OmniRoute.git
synced 2026-07-26 09:52:11 +03:00
1bda6c15dc
* fix(install): add pnpm-workspace.yaml allowBuilds + pnpm.json for pnpm 11+ pnpm 11 introduced ERR_PNPM_IGNORED_BUILDS for native addon packages. Without explicit allowBuilds approval, these packages silently skip build scripts and OmniRoute fails to start with missing native modules. Changes: - pnpm-workspace.yaml: Set allowBuilds=true for all 13 native addon packages (@parcel/watcher, @swc/core, better-sqlite3, core-js, esbuild, keytar, koffi, libxmljs2, onnxruntime-node, protobufjs, sharp, tls-client-node, unrs-resolver) - pnpm.json: Migrate onlyBuiltDependencies from package.json (deprecated field) to the new pnpm.json config file per pnpm 11 spec. Tested on: pnpm 11.9.0, Node 24, Windows 11. Fixes: pnpm install ERR_PNPM_IGNORED_BUILDS on fresh clone with pnpm 11. * chore(release): open v3.8.44 development cycle * test(security): parse Kimi Web URL host instead of substring match (CodeQL #689) (#5928) Alert js/incomplete-url-substring-sanitization: the Kimi Web executor test asserted result.url.includes("www.kimi.com"), which a hostile host like www.kimi.com.evil.net would also satisfy. Parse the URL and assert on the exact hostname (new URL(result.url).hostname === "www.kimi.com"), which is both a stronger check and clears the CodeQL warning. * refactor(translator): extract thinking-budget fitting from openai-to-claude (#5932) Extract the thinking-budget fitting cluster (fitThinkingToMaxTokens + private safeCapMaxOutputTokens + MIN_* constants) verbatim into the pure leaf openai-to-claude/thinkingBudget.ts. Host re-exports fitThinkingToMaxTokens so external importers keep working and imports it back for internal use. Host 822 -> 738 LOC (under the 800 cap). No behavior change: byte-identical bodies, public export set unchanged. Adds a split-guard test; all consumer tests stay green (translator-openai-to-claude, strip-empty, minimax-m3, passthrough). * chore(release): pipeline hardening — test-masking pre-flight gate + contributors/uncovered helpers (#5926) * chore(ci): add test-masking PR-context gate to release-green pre-flight Reproduce check:test-masking (vs origin/main) inside validate-release-green so non-allowlisted net-assert reductions surface in the local pre-flight instead of in a ~40-min CI layer on the release PR. run() now merges a per-gate opts.env so GITHUB_BASE_REF reaches the child. HARD gate; skipped under --quick. Context: v3.8.43 release cost 3 CI round-trips for PR-context gates (test-masking, file-size, pr-evidence) that check:release-green did not reproduce locally. * chore(release): add contributors generator + uncovered-commit reconciliation helpers - scripts/release/gen-contributors.mjs: reproducible `### 🙌 Contributors` table for a CHANGELOG version (parenthetical-group parser → accurate per-PR attribution, noise-handle denylist). v3.8.43 shipped without the section (a real miss) because it was hand-built. npm run release:contributors <version> [--inject]. - scripts/release/list-uncovered-commits.mjs: lists commits since the last tag with no CHANGELOG bullet (v3.8.43 had 123/176 uncovered at reconciliation start). Advisory, maintainer-side. npm run release:uncovered. - 20 unit tests (parenthetical attribution, noise exclusion, idempotent injection, coverage window). * chore(quality): absorb web-cookie-providers-new file-size drift from #5928 (base-red on release/v3.8.44) * refactor(translator): split openai-responses request translator into pure leaves (#5940) Extract the shared pure primitives and the chat->Responses direction out of the 894-line openai-responses.ts request translator: - openai-responses/helpers.ts: pure primitives (toRecord/toString/clampCallId/ normalizeVerbosity/etc + markers/regexes/JsonRecord), zero host imports - openai-responses/toResponses.ts: openaiToOpenAIResponsesRequest (chat->Responses), imports the helpers leaf Host keeps openaiResponsesToOpenAIRequest (Responses->chat, imported by production) plus both register() directions, and re-exports openaiToOpenAIResponsesRequest so external importers (tests) keep working. Host 894 -> 529 LOC (under the 800 cap). Verbatim bodies (multiset check: leaf A 54/54, leaf B 294 lines, fn1 intact), public export set unchanged, leaves never import the host (no cycle). Adds a split-guard test; all consumer tests stay green (responses-translation-fixes 37, verbosity 4, reasoning-effort 4, orphaned-tool-filter 8, empty-tool-name-loop 8, headroom-responses-format 3). * chore(ci): pr-evidence FAIL output tells you to push (body edit does not re-run the gate) (#5944) ci.yml ignores the 'edited' event, so adding the Evidence block to the PR body after a push does not re-run check:pr-evidence — you need another commit. The FAIL report now says so, at the exact place someone sees the red check. + 5 unit tests (classification + hint-on-fail / no-hint-on-pass). Decided against a separate edited-triggered workflow: pr-evidence is not a required check (no ruleset gates it; release PRs merge UNSTABLE, not BLOCKED), so the gap is cosmetic and the generate-release skill already puts Evidence in the body before the first push. * fix(providers): Perplexity Web emits real tool_calls in streaming mode (mirror chatgpt-web toolMode) (#5927) (#5937) Perplexity Web (Pro/Max) only converted <tool>{...}</tool> text into OpenAI tool_calls for non-streaming requests (hasTools && !stream). Streaming requests -- the default for agentic coding clients -- got the raw <tool> text as plain delta.content and never emitted a tool_calls SSE delta, so clients could not execute tools. Reuses the provider-agnostic buildToolModeResponse()/ toolCompletionToSseStream() helpers already shipped for chatgpt-web (#5240): when tools are requested, buffer the full completion and convert it into either a JSON completion or a terminal SSE replay carrying delta.tool_calls + finish_reason: tool_calls, regardless of the caller's stream flag. Extended buildToolModeResponse()'s idSeed to be caller-supplied (default 'cgpt', perplexity-web passes 'pplx') so tool_call ids stay provider-specific without duplicating the helper. Non-tool streaming is unchanged (still lives token-by-token via buildStreamingResponse). * fix(discovery): resolve duplicate /v1 paths and redirect aborts (#5904) Integrated into release/v3.8.44. Thanks @hamsa0x7 for diagnosing the doubled /v1 discovery path and the REDIRECT_BLOCKED probe-loop abort (#5899). De-scoped to the discovery fix (the #5903 session-affinity work is handled by #5943) and added Rule #18 regression guards. * docs(changelog): record #5926 + #5944 (release-pipeline hardening) under v3.8.44 Maintenance (#5952) * docs(claude): add Hard Rule #22 — cross-session safety (git stash + in-flight PRs) (#5955) Integrated into release/v3.8.44 — Hard Rule #22 (cross-session safety). * refactor(translator): extract pure helpers from response/openai-responses (#5949) Extract the 5 stateless helpers (normalizeToolName, stripEmptyOptionalToolArgs, normalizeOutputIndex, normalizeUpstreamFailure, extractResponsesReasoningSummaryText) verbatim into the pure leaf openai-responses/pureHelpers.ts (no stream state, no host import). Host imports them back and re-exports normalizeUpstreamFailure for external importers (tests). Host 1091 -> 1001 LOC. The stateful streaming core stays in the host (out of scope). Byte-identical bodies (multiset 73/73), no cycle. Adds a split-guard; consumer tests stay green (responses-translation-fixes 37, combo-param-validation-fallback-4519 5). * docs(compression): document upstream sync policy for RTK/Caveman engines (#5830) (#5948) Integrated into release/v3.8.44 — docs-only upstream sync policy for RTK/Caveman engines (closes #5830). All 7 checks green. * fix(sse): strip ANSI/VT100 codes from gemini-cli stream frames (#5934) Integrated into release/v3.8.44 — ReDoS-safe ANSI/VT100 strip for gemini-cli stream frames (port of upstream #2273, thanks @anki1kr). PR test green (5/5), file-size gate OK. * fix(translator): strict Anthropic content-block compliance in antigravity→openai request (#5935) Integrated into release/v3.8.44 — strict Anthropic content-block compliance in antigravity→openai (port upstream #2296). PR test green (9/9). UNSTABLE red is the pre-existing environmental setup-claude base-red (opencode-plugin dist not built in fast-path), not a regression from this PR. * fix(mcp): auto-recover stale streamable HTTP sessions on initialize (#5957) Integrated into release/v3.8.44 — MCP stale streamable-HTTP session auto-recovery (thanks @Chewji9875). * fix(providers): validate v0 Platform API keys via chats endpoint (#5954) Integrated into release/v3.8.44 — v0-vercel Platform API key validation (thanks @vittoroliveira-dev). * fix(api): relax provider-scoped chat completion validation (#5907) Integrated into release/v3.8.44 — relaxed provider-scoped chat validation + regression test (thanks @nickwizard). * fix(providers): strip /v1 unconditionally to avoid /v1/v1/models fetch error (#5899) (#5920) Integrated into release/v3.8.44 — unconditional /v1 strip in both models-discovery paths + regression test (thanks @anki1kr). * fix(resilience): per-window is_exhausted + honor quota-exhaustion preflight for priority combos (#5923) (#5941) Integrated into release/v3.8.44. * fix(resilience): honor active codex session affinity over per-request reset-aware re-scoring (#5903) (#5943) Integrated into release/v3.8.44. * fix(thinking): only inject redacted_thinking replay block when tool_use present and thinking enabled (#5945) (#5953) Integrated into release/v3.8.44. * feat(providers): add ClinePass API-key provider (#5942) Integrated into release/v3.8.44 — ClinePass API-key (BYOK) provider (port upstream 9router#2304, co-authored @adentdk). Validated locally: 16 clinepass tests green; fixed the APIKEY count 158→159 + translate-path golden snapshot (clinepass is a genuine new provider). Remaining UNSTABLE red is the pre-existing environmental setup-claude base-red (opencode-plugin dist not built in fast-path). Supersedes stub #5541. * feat(api): add /v1/ocr endpoint (Mistral OCR) + Mistral moderation (#5950) Integrated into release/v3.8.44 — /v1/ocr endpoint (Mistral OCR) + Mistral moderation (port upstream 9router#2064, co-authored @waguriagentic). Validated locally: 14 ocr-route tests + moderation/servicekind/endpoint-category suites green (CORS→Zod→handler + no-stack-leak assertion). Reds are inherited DRIFT only: cognitive-complexity ratchet (none from OCR files — pre-existing cycle drift, rebaselined at release) + environmental setup-claude base-red. * fix(codex): convert chat json schema to responses text format (#5933) Integrated into release/v3.8.44 — converts Chat Completions json_schema response_format → Responses API text.format on the Codex path, and preserves existing text.format through verbosity normalization. Base redirected main→release; the openai-responses.ts split that landed this cycle was reconciled by re-applying the delta onto openai-responses/toResponses.ts. Validated locally: 48 translator-openai-responses-req + 8 codex-verbosity tests green. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(providers): add Claude Sonnet 5 support across the model pipeline (#5833) Integrated into release/v3.8.44 — wires claude-sonnet-5 end-to-end (registries, modelSpecs, pricing ×3, cost, Sonnet-family fallback, 1M-ctx, static models). Reconciled the add/add overlap with the already-merged #5796 (kept the PR's superset test with the family-fallback assertion). Validated locally: kiro-sonnet-5 + catalog + pricing/modelSpecs/fallback suites all green. Thanks @ggiak! Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(relay): gate bifrost auto routing by provider manifest (#5870) Integrated into release/v3.8.44 — gates Bifrost auto-routing by the provider plugin manifest (only manifest-eligible providers reach the sidecar; ineligible/unknown fall back to the TS path with explicit reasons). Superset of #5869 (carries the full manifest + registry + docs). Resolved an integration-test conflict in favor of the release (which already subsumes this PR's readiness/removeDirWithRetry improvements). Validated locally: 4 provider-plugin-manifest + 11 relay-routing-backend tests green. Thanks @KooshaPari! Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * refactor(translator): extract pure message helpers from openai-to-kiro (852→751) (#5947) * refactor(translator): extract pure message helpers from openai-to-kiro Extract the pure tool/message helpers (parseToolInput, normalizeKiroToolSchema, serializeToolResultContent) verbatim into the leaf openai-to-kiro/messageHelpers.ts. The host imports them back for convertMessages. They were module-private, so the public export set is unchanged (no re-export needed). Host 852 -> 751 LOC. Byte-identical bodies (multiset 99/99), leaf has zero imports (no cycle). Adds a split-guard; consumer tests stay green (translator-openai-to-kiro 33, translator-ai-sdk-image-parts 3). * chore: re-trigger CI (stuck runner on 2/2 shard) * refactor(executors): extract pure prompt + composer helpers from cursor (#5960) Extract two pure clusters from the cursor executor into sibling leaves: - cursor/prompt.ts: isRecordLike + toolChoiceDirectiveLine + buildCursorOutputConstraints - cursor/composer.ts: composer thinking-as-content decoding (isComposerModel, visibleComposerContentFromThinking, composerReasoningRemainder + markers) Host imports both back for internal use and re-exports the 3 composer helpers for external importers (tests). Host 1576 -> 1451 LOC. Byte-identical bodies (verbatim multiset prompt 65/65, composer 32/32), leaves have zero imports (no cycle). Adds a split-guard; consumer tests stay green (cursor-composer-thinking, cursor-streaming, cursor-agent-tool-calls, translator-openai-to-cursor, cursor-agent-system-prompt). * refactor(executors): extract pure SSE-collect parsing from antigravity (#5962) Extract the pure SSE-payload -> collected-stream parser (AntigravityCollectedStream, stripZeroWidth, parseAntigravityTextualToolCall, addAntigravityTextualToolCall, processAntigravitySSEPayload/Text, flushAntigravitySSEText) verbatim into the leaf antigravity/sseCollect.ts. Host imports the helpers it uses and re-exports processAntigravitySSEPayload for external importers (tests). Host 1812 -> 1671 LOC. Byte-identical bodies (verbatim multiset 135/135), leaf does not import the host (no cycle). Credit/quota state, auth, and HTTP dispatch untouched. Adds a split-guard; consumer tests stay green (executor-agy 8, executor-antigravity 26, antigravity-sse-collect-socket-release, copilot-agent-antigravity-parity 6). * refactor(executors): extract pure model maps + resolvers from chatgpt-web (#5967) Extract the static model maps (MODEL_MAP, MODEL_FORCED_EFFORT, THINKING_CAPABLE_SLUGS) and the pure thinking-effort resolvers (isThinkingCapableModel, normalizeThinkingEffort, resolveThinkingEffort, ResolvedChatGptModel, resolveChatGptModel) verbatim into the pure leaf chatgpt-web/models.ts. Host imports the two resolvers it uses back. Host 3205 -> 3076 LOC. Byte-identical bodies (verbatim multiset 120/120), leaf has zero imports (no cycle). Auth/PoW/session/HTTP dispatch and all module caches untouched. Adds a split-guard; consumer tests stay green (chatgpt-web 86, chatgpt-web-tools-5240 4, chatgpt-web-sha3-boringssl-5531 5). * refactor(executors): decompose grok-web into pure tool/markup leaves (#5994) Extract the pure OpenAI<->Grok tool-translation, native-tool mapping, markup cleanup, and NDJSON stream types out of the 1872-line grok-web executor into 4 sibling leaves: - grok-web/types.ts: GrokStreamResponse/GrokStreamEvent (stream types) - grok-web/tool-bridge.ts: OpenAI<->Grok tool translation + registry + classifiers - grok-web/native-tools.ts: native-tool selection/scoring + native->OpenAI mapping - grok-web/text-cleanup.ts: Grok markup stripping + GrokMarkupFilter Layered, acyclic: types <- tool-bridge <- native-tools; text-cleanup <- types; host imports the leaves. All symbols module-private (no host re-export). Host 1872 -> 887 LOC. Byte-identical bodies (verbatim per-leaf), no cycle, all new leaves <= 800 cap (tool-bridge split at line 753 to stay under). Auth/cookie/TLS/HTTP dispatch untouched. Adds a split-guard; consumer tests stay green (grok-web 62, grok-cli-oauth 15, grok-cli-strip-params 2). * refactor(executors): extract pure quota parsing from codex (#5999) Extract the pure Codex quota-snapshot parsing + reset/cooldown scheduling (CodexQuotaSnapshot, parseCodexQuotaHeaders, getCodexResetTime, getCodexDualWindowCooldownMs) verbatim into the leaf codex/quota.ts. Host re-exports the 4 symbols so handlers/chatCore/codexQuota.ts + tests keep resolving. Host 1539 -> 1427 LOC. Byte-identical bodies (verbatim 98/98), leaf has zero imports (only Date, no cycle). WS transport, auth, HTTP dispatch untouched. Adds a split-guard; consumer tests stay green (executor-codex 40, codex-quota-fetcher 7, chatcore-codex-quota 5). * refactor(executors): extract pure stream formatters from deepseek-web (#6000) Extract the pure content/citation formatters (isThinkingModel, isSearchModel, cleanDeepSeekToken, formatStreamContent, DeepSeekSearchResult, appendSearchCitations) verbatim into the leaf deepseek-web/stream-format.ts. Host imports the 5 it uses back into transformSSE/collectSSEContent (cleanDeepSeekToken stays internal to the leaf). Host 1147 -> 1108 LOC. Byte-identical bodies (verbatim 34/34), leaf has zero imports (no cycle), all module-private (no re-export). PoW/auth/token-cache/HTTP dispatch untouched. Adds a split-guard; consumer tests stay green (deepseek-web 35, deepseek-web-rolling-window-2942 5, deepseek-web-tools-execute 3). * refactor(api): add validatedJsonBody helper (salvage #5075) (#5931) Fuses JSON body parsing + Zod validation into a single call that returns either type-narrowed data or a ready-to-return 400 NextResponse with the standard error envelope. Salvaged as the Tier 1 portable helper from the closed refactor PR #5075; the bulk route migration is intentionally not ported. Adds a focused 6-case regression test. Co-authored-by: KooshaPari <KooshaPari@users.noreply.github.com> * feat(qoder): drive PAT auth via qodercli, add dashboard quota, fix connection display (#5816) Integrated into release/v3.8.44 — Qoder PAT auth via qodercli binary + dashboard quota + dual-auth connection fix. Thanks @AgentKiller45 (co-author @judy459)! Validated locally (release-green on its own merits): lint 0, typecheck:core 0, 104 qoder/usage/UI tests green, file-size gate OK (owner-approved qoderCli.ts baseline-freeze 666→989), env-doc-sync fixed (documented QODER_CLI_CONFIG_DIR). The 2 remaining CI reds are INHERITED base-reds, not caused by this PR: (1) LEDGER-4 minimax-m3 supportsVision (minimax-m3 base + cline-pass/minimax-m3 from the already-merged #5942); (2) mutation-test-coverage missing 3 tests in stryker.conf (#5903/#5942/#5923). Both cleaned up separately. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix(providers): minimax-m3 supportsVision (LEDGER-4) + stryker tap.testFiles drift (#6012) Release-green cleanup — clears LEDGER-4 minimax-m3 supportsVision + stryker tap.testFiles drift base-reds. Validated locally. * fix(registry): flag cline-pass/minimax-m3 as multimodal (supportsVision) (#6003) The cline-pass provider's minimax-m3 entry was missing supportsVision, breaking the LEDGER-4 registry-consistency test (all minimax-m3 entries must set supportsVision to match lite.ts — minimax-m3 is multimodal). Every other minimax-m3 registry entry (trae, bazaarlink, cline, ollama-cloud, ...) already sets it. This was a base-red on release/v3.8.44 inherited by every open PR. Validated by the existing failing-then-passing guard tests/unit/review-reviews-v3814-fixes.test.ts (LEDGER-4). * refactor(executors): extract pure payload construction from claude-web (#6006) Extract the pure Claude-web payload types + transforms + default tools/style (ClaudeWebRequestPayload, ClaudeWebStreamChunk, DEFAULT_CLAUDE_MODEL, generateMessageUUIDs, getDefaultTools, getDefaultPersonalizedStyle, transformToClaude, transformFromClaude) verbatim into the leaf claude-web/payload.ts. Host imports the 3 it uses back (ClaudeWebRequestPayload type + the two transforms). Host 1056 -> 835 LOC. Byte-identical bodies (verbatim 149/149), leaf imports only randomUUID (no host import, no cycle), all module-private (no re-export). Cookie/auth/ Turnstile/TLS/HTTP dispatch untouched. Adds a split-guard; consumer tests stay green (claude-web 13, claude-web-auto-refresh 6). * refactor(executors): extract pure upstream-header helpers from base (#6008) Extract the pure upstream-header helpers (mergeUpstreamExtraHeaders, getCustomUserAgent, setUserAgentHeader, applyConfiguredUserAgent, isOpenAICompatibleEndpoint, stripStainlessHeadersForOpenAICompat) verbatim into the leaf base/headers.ts. base.ts is imported by ~18 executors, so the host re-exports all 6 to keep those import paths intact; it also imports the 4 it uses internally in the BaseExecutor class. The trivial JsonRecord type alias is redefined locally in the leaf to avoid a base<->leaf cycle. Host 1539 -> 1451 LOC. Byte-identical bodies (verbatim 78/78), leaf does not import the host (no cycle). typecheck:core validates all base importers still resolve via the re-export. Adds a split-guard; consumer tests stay green (executor-base-utils 22, executor-default-base 49, executor-strip-stainless-openai-compat 6, plus executor sanity via typecheck). * refactor(executors): extract pure wire protocol from perplexity-web (#6014) Extract the pure Perplexity wire protocol (consts, SSE stream types, SSE parsing, OpenAI<->Perplexity message translation, request/query builders, content extraction, sseChunk) verbatim into the leaf perplexity-web/protocol.ts. Host imports back the 10 symbols it uses; everything module-private (no re-export). Session cache, TLS fetch, auth, and the executor class stay in the host. Host 1028 -> 534 LOC. Byte-identical bodies (verbatim), leaf imports only randomUUID (no host import, no cycle). Adds a split-guard; consumer tests stay green (perplexity-web 26, streaming-tools-5927 2, tls-client 6, key-validation-models 2). * refactor(executors): extract pure URL normalizers from default (#6015) Extract the pure per-provider chat-URL normalizers (normalizeBailianMessagesUrl, normalizeDataRobotChatUrl, normalizeAzureAiChatUrl, normalizeWatsonxChatUrl, normalizeOciChatUrl, normalizeSapChatUrl, normalizeXiaomiMimoChatUrl, normalizeOpenAIChatUrl, getOpenRouterConnectionPreset) verbatim into the leaf default/urlNormalizers.ts. Host imports them back into buildUrl/transformRequest; the now-dead build*ChatUrl/normalizeBaseUrl imports move to the leaf. All module-private (no re-export). Host 864 -> 815 LOC (shrunk below its frozen baseline). Byte-identical bodies (verbatim 45/45), leaf does not import the host (no cycle). buildHeaders/execute/auth untouched. Adds a split-guard; consumer tests stay green (executor-default-base 49, anthropic-compatible-bearer 3, strip-client-metadata 3). * feat(webfetch): support self-hosted FireCrawl instances (#5793) Integrated into release/v3.8.44 — self-hosted FireCrawl support (FIRECRAWL_BASE_URL/FIRECRAWL_TIMEOUT_MS). Re-cut clean onto the release tip (branch was fossilized from a pre-v3.8.40 snapshot). Validated: 4 firecrawl tests green, env-doc-sync + docs-sync pass. UNSTABLE red is the inherited environmental setup-claude base-red. * feat(xai): register XaiExecutor with reasoning-effort suffix parsing (#5800) Integrated into release/v3.8.44 — XaiExecutor with reasoning-effort suffix parsing. Re-cut clean onto the release tip (branch was fossilized). Validated: 6 xai-executor tests green, provider-consistency OK, typecheck:core 0 errors, env-doc-sync in sync. UNSTABLE red is the inherited environmental setup-claude base-red. * feat(discovery): Phase 2 — reporter, /api/discovery/* routes (strict loopback-only) + dashboard UI (#5939) * feat(discovery): Phase 2 reporter — discoveryResults DB module + service wiring Adds src/lib/db/discoveryResults.ts (CRUD over the discovery_results table from migration 074) and wires the opt-in discovery service to persist and read findings through it: persistDiscoveryResult / getDiscoveryResults / getDiscoveryResultById / markVerified / deleteDiscoveryResult, with (provider, method, endpoint) upsert de-duplication. Re-exported from localDb. The service stays opt-in / default-off. The /api/discovery/* routes and the dashboard UI tab are intentionally deferred to Phase 2b — they need the local-only enforcement model (Hard Rules #15/#17 territory) decided first. TDD: tests/unit/db/discovery-results.test.ts (8 cases, DB + service delegation), isolated DATA_DIR with resetDbInstance cleanup. * feat(discovery): Phase 2b — /api/discovery/* routes (strict loopback-only) Adds the discovery HTTP surface on top of the reporter DB module: GET /api/discovery/results list findings (optional ?providerId) GET /api/discovery/results/:id one finding (404 if absent) DELETE /api/discovery/results/:id delete a finding POST /api/discovery/scan scan a provider + persist findings POST /api/discovery/verify/:id mark a finding verified Authorization: strict loopback-only. "/api/discovery/" is added to LOCAL_ONLY_API_PREFIXES so the central authz pipeline (proxy.ts → runAuthzPipeline → managementPolicy) rejects non-loopback callers with a 403 LOCAL_ONLY before any handler runs. It is deliberately NOT in LOCAL_ONLY_MANAGE_SCOPE_BYPASS_PREFIXES — no remote manage-scope bypass — because POST /scan issues outbound probes to provider endpoints (SSRF-adjacent) and must never be tunnel-reachable. Handlers also call requireManagementAuth (defense in depth) and return sanitized errors via createErrorResponse. Tests: - tests/unit/authz/discovery-routes-local-only.test.ts (8) — security guard: isLocalOnlyPath true + not manage-scope-bypassable for all four paths. - tests/unit/api/discovery-routes.test.ts (6) — handler integration over an isolated DATA_DIR: list/filter, by-id 200/404/400, scan persist + 400 on empty/malformed body, verify 200/404, delete 200/404, no stack-trace leak. * feat(discovery): Phase 2c — dashboard UI tab (Tools → Discovery) Adds the /dashboard/discovery page (DiscoveryPageClient) that consumes the Phase 2b /api/discovery/* routes: scan a provider, list findings, verify or delete them. Registered in the sidebar under the Tools group (icon travel_explore) and given a "discovery" i18n namespace + sidebar keys in en.json (other locales fall back to en via next-intl until synced — the locale files are in a pre-existing coverage deficit unrelated to this change). Registers the UI test path in vitest.config.ts (advisory ui suite). Tests: src/app/(dashboard)/dashboard/discovery/__tests__/DiscoveryPageClient.test.tsx (3 cases: loads+renders results, empty state, fetches /api/discovery/results on mount; stable useTranslations mock to avoid the fetch-loop). NOTE: the ui vitest suite cannot run in this workspace — @testing-library/dom (a @testing-library/ react peer dep) is absent from node_modules, which fails ALL existing ui tests equally; the test runs in CI. Component verified locally via typecheck + lint. * test(discovery): register discovery-routes-local-only in stryker tap.testFiles The mutation-test-coverage gate (--strict) flags any unit test covering a mutated module that isn't listed in stryker.conf.json tap.testFiles. This PR's tests/unit/authz/discovery-routes-local-only.test.ts covers src/server/authz/ routeGuard.ts (a mutated module, which this PR edits by adding the /api/discovery/ local-only prefix), so it must be registered for its mutant kills to count. No behavior change. * refactor(discovery): split DiscoveryPageClient to satisfy max-lines-per-function The complexity ratchet (max-lines-per-function: 80) flagged the single 184-line DiscoveryPageClient function (+1 over baseline). Extract the data layer into two hooks (useDiscoveryResults for list/loading/feedback, useDiscoveryActions for scan/verify/delete), a shared callApi helper, and two presentational sub-components (DiscoveryScanForm, DiscoveryResultCard). Every function is now under the 80-line ceiling; complexity gate back to baseline 1995. No behavior change — same exported component, same endpoints, same props. * test(sidebar): include discovery in omni-proxy item-order snapshot Adding the Discovery item to the Tools group (this PR's sidebar entry) extends the ordered omni-proxy section list. Update the exact-match deepEqual snapshot in sidebar-visibility.test.ts to include "discovery" in its position (after traffic-inspector). The assertion stays exact — this reflects the intentional new item, it does not weaken the check. * docs(changelog): restore release bullets eaten by merge auto-resolve; re-add discovery bullet additively * chore(quality): bump testFrozen for translator-openai-responses-req.test.ts (1097 -> 1172) Base-red inherited from #5933, which grew the test file to 1171 lines (Hard Rule #18 regression tests) without adjusting the frozen cap. The release tip itself fails check:file-size; this unblocks every PR into release/v3.8.44. File untouched by this PR. * chore(quality): restore stryker tap.testFiles entries eaten by merge auto-resolve The merge of origin/release/v3.8.44 silently dropped the 3 entries added on the release side (#5903, clinepass, #5923). Took the release version verbatim and re-added only this PR's entry (discovery-routes-local-only) in alphabetical order. check:mutation-test-coverage green locally. * chore(quality): reconcile inherited v3.8.44 merge-burst drift + include discovery in tools-group order test - complexity 1995->2003 and cognitive 856->859: both measure IDENTICAL on the pristine release tip (3a3d618fe) and this PR's merged HEAD — the PR is complexity-net-zero; drift is from the 2026-07-02 merge burst (notes added to both baselines, same family as prior reconciliations). - sidebar-tools-group.test.ts: append 'discovery' to the expected TOOLS_GROUP order — the intentional new sidebar item this PR adds (same expected-value update already made in sidebar-visibility.test.ts). * feat(providers): custom icon URL for compatible provider nodes (#5815) Integrated into release/v3.8.44 — custom icon URL for compatible provider nodes (DB migration 113 + nodes.ts + Zod schema + API routes + catalog + ProviderIcon UI). Re-cut onto the release tip (branch was fossilized ~13 real files); reconciled icon_url into the release's evolved nodes.ts/routes via 3-way. Validated: 14 backend + 5 frontend(vitest) + 24 page-utils tests green, typecheck:core 0, provider-consistency OK, file-size/env-doc-sync pass. UNSTABLE red is the inherited environmental setup-claude base-red. * feat(api): add /v1/audio/translations endpoint (#5809) Integrated into release/v3.8.44 — /v1/audio/translations endpoint (Whisper-style audio translation) + audioTranslation handler + translation providers in audioRegistry. Re-cut clean onto the release tip (branch was fossilized). Validated: 8 route tests (incl. no-stack-leak), typecheck:core 0, route-guard-membership OK, docs gates pass. UNSTABLE red is the inherited environmental setup-claude base-red. * feat(dashboard): wildcard-CORS runtime warning + CORS security doc (#5602) (#5759) Integrated into release/v3.8.44 — wildcard-CORS runtime warning banner + docs/security/CORS.md security guide (#5602). Re-cut clean onto the release tip (branch was fossilized). Validated: 20+9 backend + 2 banner(vitest) tests green, typecheck:core 0, docs-sync/symbols/fabricated/doc-links pass. UNSTABLE red is the inherited environmental setup-claude base-red. * refactor(executors): extract pure JSONL stream translation from huggingchat (#6016) Extract the pure JSONL->OpenAI-SSE translation (sseChunk, parseJsonlLine, streamJsonlToOpenAi, readJsonlResponse) verbatim into the leaf huggingchat/jsonlStream.ts. They consume a passed-in ReadableStream (no fetch/network/state). Host imports back the two it uses; all module-private (no re-export). Host 812 -> 594 LOC. Byte-identical bodies (verbatim), leaf has zero imports (no cycle). Cookie/auth/multipart/execute untouched. Adds a split-guard; consumer tests stay green (executor-huggingchat 6, huggingchat-model-catalog 3). * refactor(executors): extract pure Meta AI response parser from muse-spark-web (#6017) Extract the pure Meta AI SSE/JSON response parsing + content/reasoning/error extraction (parseMetaSseFrames, readMetaJsonPayloads, collect*/extract*/classify* helpers, parseMetaAiResponseText, isRecord, the reasoning/renderer key arrays, MetaSseFrame/ ParsedMetaAiResponse types) verbatim into the leaf muse-spark-web/response-parser.ts. Host imports back the 3 it uses; all module-private (no re-export). Host 1301 -> 925 LOC. Byte-identical bodies (verbatim), leaf has zero imports (no cycle). Conversation cache, cookie/auth, fetch, executor class untouched. Adds a split-guard; consumer tests stay green (muse-spark-cookie-copy-5449 2, muse-spark-web-continuation 6). * refactor(executors): extract pure EventStream framing from kiro (#6018) Extract the pure AWS EventStream binary framing (ByteQueue, CRC32 table + crc32, TEXT_ENCODER/TEXT_DECODER, KIRO_VERIFY_FULL_CRC, parseEventFrame, EventFrame type) verbatim into the self-contained leaf kiro/eventstream.ts (local JsonRecord alias to avoid a cycle). Host imports back the 3 it uses (ByteQueue, TEXT_ENCODER, parseEventFrame). Host 943 -> 758 LOC. Byte-identical bodies (verbatim 145/145), leaf has zero host imports (no cycle). Auth/token-refresh/streaming-state/executor class untouched; the test-imported flushBufferedToolArgs/resolveKiroRegion/kiroRuntimeHost stay exported on the host. Adds a split-guard; consumer tests stay green (executor-kiro 9, kiro-tool-args-streaming 7, kiro-iam-region 10). * refactor(executors): extract challenge solver from duckduckgo-web (#6020) Extract the DuckDuckGo anti-abuse challenge solver + FE signals (CHALLENGE_STUBS, countHtmlElements, buildHtmlLookup, sha256Base64, solveDuckDuckGoChallenge, makeDuckDuckGoFeSignals) verbatim into the leaf duckduckgo-web/challenge.ts. The vm sandbox + 5s timeout (SECURITY note) are preserved. Host imports back the two it uses. Host 924 -> 788 LOC. Byte-identical bodies (verbatim 132/132), leaf does not import the host (no cycle). The now-dead createHash/parse5 host imports are removed; vm stays (still used in host). Auth/cookie/warm/seed/executor untouched. Adds a split-guard; consumer tests stay green (duckduckgo-web-executor 15, duckduckgo-domain-4037 8). * test(cli): deflake setup-claude.test.ts — silence console to stop stdout/report interleaving (#5959) (#6019) Integrated into release/v3.8.44. Deflakes tests/unit/cli/setup-claude.test.ts (#5959) — verified in CI: setup-claude now passes in Unit Tests fast-path (2/2). Merged with --admin over two PRE-EXISTING base-reds proven independent of this test-only change (this PR only touches setup-claude.test.ts + CHANGELOG): - Fast Quality Gates → check:test-discovery: tests/unit/executors/{firecrawl-fetch,xai-executor}.test.ts are orphaned on release/v3.8.44 (added by #5793/#5800); the shard glob 'tests/unit/{api,...,ui}/**' omits 'executors'. Both blobs exist on the pristine base. - Unit Tests fast-path (2/2): tests/unit/settings-i18n-keys.test.ts → 'direct translation calls have English messages' fails on the pristine base too (unrelated i18n base-red). * fix(cli): stabilize setup-claude.test.ts flake — inject dry-run log sink (#6021) * fix(cli): stabilize setup-claude.test.ts flake — inject dry-run log sink (#5959) Root cause (isolated empirically, 5/10 fail on the pristine base): the dry-run path of syncClaudeProfilesFromModels console.log's a multi-byte box-drawing heading ("── [dry-run] … ──"). Under the node:test runner that write lands on the test child's stdout and corrupts the runner's V8-serialized event stream ~50% of the time ("Unable to deserialize cloned data due to invalid or unsupported version"), killing the file at the first logging test. ASCII-only logging never reproduced it (0/20); the unicode heading alone reproduced it (10/20). Fix: syncClaudeProfilesFromModels accepts an injectable log sink (opts.log, CLI default unchanged: console.log). The dry-run test injects a collector — keeping unicode off the child's stdout — and gains assertions on the dry-run report (path + parsed settings content), which FAIL on the old code (log ignored) and PASS on the new one. Validation: 0/30 failures post-fix vs 5/10 pre-fix on the same tree. Baselines: complexity 2003->2006 and cognitive 859->860 are inherited post-3a3d618fe release drift — measured identical on the pristine base with and without this change (notes added in both files). * test(ci): collect the orphaned tests/unit/executors/ directory (base-red unblock) #5800 created tests/unit/executors/ outside every unit-runner brace glob, so its 2 test files (firecrawl-fetch, xai-executor) never ran anywhere and check:test-discovery flags them as NEW orphans on the pristine base, red-flagging every PR into release/v3.8.44. Added 'executors' to the runner globs in package.json (7 scripts), ci.yml unit shards, quality.yml TIA glob, build-test-impact-map.mjs, and the test-discovery gate's COLLECTORS (the gate enforces those stay in sync). Both files pass when actually collected (10/10); cli+executors under suite flags: 99/99. * chore(quality): complexity baseline 2006 -> 2007 (CI-observed value) The GitHub fast-gates runner measures 2007 where local measures 2006 — the same local-vs-CI off-by-one documented in the 2026-06-26 note. Pin the CI-observed value so the gate is deterministic where it runs. * fix(i18n): add the 6 missing en.json keys flagged by settings-i18n-keys (base-red unblock) providers.iconUrlLabel/iconUrlHint (referenced by AddCompatibleProviderModal and EditCompatibleNodeModal) and settings.authz.cors.wildcard.title/desc (the #5602 CORS_ALLOW_ALL banner in AuthzSection) shipped without their en.json messages — 'direct translation calls have English messages' fails on the pristine release tip, red-flagging every PR. git log -S proves the keys never existed (not a merge-eat). Scanner test: 10/10 green. * refactor(executors): extract reasoning-effort (base) + tool-normalization (codex) leaves (#6030) Two pure-leaf follow-ups closing the Block H tail: - base/reasoningEffort.ts: provider-aware reasoning_effort sanitation (MISTRAL/GITHUB reject patterns, supportsMaxEffortForProvider, sanitizeReasoningEffortForProvider). Deps are config/services only (PROVIDER_CLAUDE, isClaudeCodeCompatible, supportsClaudeMaxEffort/supportsXHighEffort) so the leaf never imports the host — no cycle. base.ts re-exports sanitizeReasoningEffortForProvider for its external importers (mimoThinking + tests). base.ts 1466 -> 1312 LOC. - codex/tools.ts: Responses-API tool normalization (CODEX_HOSTED_TOOL_TYPES hosted-tool passthrough, isCodexFreePlan gating, normalizeCodexTools). Self-contained (console.debug only). codex.ts re-exports isCodexFreePlan + normalizeCodexTools for external importers (tests + provider services). codex.ts 1430 -> 1268 LOC. Byte-identical bodies (verbatim: base 100/100, codex 126/126); both leaves have zero host imports. Adds two split-guards asserting the leaf owns the symbol and both import paths resolve to the same function. Consumer tests stay green (base-executor-sanitize-effort 34, executor-codex 40, mimoThinking 9, codex-free-plan-image-generation 3, issue-fixes 6). * test(ci): move orphaned executor tests to top-level so a runner collects them (#6027) Integrated into release/v3.8.44 — collect orphaned executor tests (check:test-discovery base-red). * test(cli): deflake cli-setup-opencode.test.ts — silence console (#5959-class landmine) (#6033) The command under test prints CLI progress with multi-byte glyphs (printSuccess "✔" in the happy paths, printError "✖" in the dist-missing path that test 4 exercises) via console.log. Under the node:test runner those child-stdout writes interleave with the V8-serialized report frames and can corrupt the stream — the exact #5959 mechanism proven for setup-claude.test.ts; this file's ✖ line was already visible entangled in red CI runs. No test here asserts on stdout, so silence console.log/info/ warn for the file (same pattern as #6019/#6021, restored in after()). Validation: pre-fix the ✖/✔ lines reach stdout every run (grep-able); post-fix stdout is clean, 4/4 tests green, 0/20 failures across 20 runs. * feat(agy): support Google Cloud project ID settings (#5905) * feat(agy): support Antigravity project ID settings * refactor(agy): collapse Antigravity family project gate --------- Co-authored-by: Nikolay Alafuzov <alafuzov_nn@rusklimat.ru> * feat(proxy): add Webshare proxy pool import and sync (#5993) * feat(proxy): add Webshare proxy pool import and sync Adds Webshare (https://proxy.webshare.io) as a fourth source in the free-proxy provider framework alongside 1proxy, Proxifly, and IPLocate. WebshareProvider paginates the account's `/api/v2/proxy/list/` endpoint (Authorization: Token <key>), upserts proxies into the shared `free_proxies` table via the existing db/freeProxies.ts helpers, and tombstones proxies the account no longer lists (recycled/retired IDs) while never touching rows already promoted into the live proxy pool. Unlike the other sources, Webshare is a paid per-account list, so it is gated on FREE_PROXY_WEBSHARE_API_KEY rather than a plain on/off flag. No DB migration needed — reuses the existing free_proxies table and proxy_registry-on-promote path. Co-authored-by: ricatix <d.enistraju155@gmail.com> Inspired-by: https://github.com/decolua/9router/pull/1176 * chore(changelog): restore release entries + add webshare bullet --------- Co-authored-by: ricatix <d.enistraju155@gmail.com> * feat(api-keys): add per-key device/connection tracking (#5998) * feat(api-keys): add per-key device/connection tracking Tracks distinct client devices (SHA-256 fingerprint of IP + User-Agent) seen with each API key, with a 30-minute TTL and per-key/global caps. The tracker is in-memory only (module-scoped Map, same pattern as sessionManager.ts — no global.* singleton) and never stores the raw IP: it is masked before being written. Hooked into open-sse/handlers/chatCore.ts (the real chat entry) rather than the legacy src/sse/handlers path. New GET /api/keys/[id]/devices management route exposes masked device details for a key, and the API Keys dashboard tab gets a "Devices" count badge alongside the existing Sessions badge. This is a new granularity distinct from the existing maxSessions cap (src/lib/db/apiKeys.ts), which limits concurrent sticky-routing sessions rather than tracking device identity. Co-authored-by: Muhammad Mugni Hadi <mugni@rukita.co> Inspired-by: https://github.com/decolua/9router/pull/931 * chore(changelog): restore release entries + add api-keys device-tracking bullet --------- Co-authored-by: Muhammad Mugni Hadi <mugni@rukita.co> * fix(providers): only apply openai-family model inference fallback when no cataloged provider serves the id (#5852) (#5938) resolveModelByProviderInference() in open-sse/services/model.ts had an unconditional /^gpt-/i heuristic that hijacked any model id starting with gpt-/o1/o3 into provider openai, even when the id is cataloged under other providers. This broke bare (non-combo) requests for open-weight models like gpt-oss-120b (served by fireworks/cerebras/scaleway/byteplus/sambanova/ heroku), which don't exist on openai's catalog, producing a 404 with no fallback. Gate the heuristic on providers.length === 0 so it only fires for genuinely uncataloged openai-family ids, letting cataloged ids fall through to the existing single-candidate / ambiguous-candidate resolution paths. Regression guard: tests/unit/gptoss-provider-inference-5852.test.ts * fix(cc-compatible): send SSE accept for streamed requests (#5958) Integrated into release/v3.8.44 — SSE Accept header for streamed cc-compatible requests (thanks @rdself). * fix: deepseek-web reliability — auto-refresh on 401/403, refresh v2.0.0 client headers, fix token-kind bulk import (#5988) Integrated into release/v3.8.44 — deepseek-web auto-refresh + v2.0.0 headers + token-kind bulk import (thanks @backryun). * feat(providers): support Vercel AI Gateway embeddings and images (#5968) * feat(providers): support Vercel AI Gateway embeddings and images Extends the existing vercel-ai-gateway (alias vag) provider — currently chat-only — with embeddings and image generation support, since the gateway's OpenAI-compatible /v1 API also exposes /embeddings and /images/generations. Adds entries to EMBEDDING_PROVIDERS (embeddingRegistry.ts) and IMAGE_PROVIDERS (imageRegistry.ts) modeled on the existing openai entries. Out of scope for this PR (tracked as follow-ups): the /v1/credits usage reader, retry:{429:2} tuning, and claude->reasoning_effort mapping. Co-authored-by: Ngô Tấn Tài <tantai@newnol.io.vn> Inspired-by: https://github.com/decolua/9router/pull/1704 * chore(changelog): restore release entries + add vercel-gateway media bullet --------- Co-authored-by: Ngô Tấn Tài <tantai@newnol.io.vn> * feat(cli-tools): add Crush CLI tool to the dashboard (#5970) * feat(cli-tools): add Crush CLI tool to the dashboard Add a `crush` entry to the dashboard CLI-Tools catalog and a new `/api/cli-tools/crush-settings` route (GET/POST/DELETE), cloned from the `pi` tool's route as a template. OmniRoute already ships a `crush` CLI command path (bin/cli/commands/setup-crush.mjs) but the dashboard catalog had no matching entry. The new route writes the real Crush config shape (providers.omniroute as an openai-compat provider block) to the same canonical config path (~/.config/crush/crush.json) that setup-crush.mjs's resolveCrushTarget() already writes to, so the dashboard and the CLI command agree on one location. Adds CLI_TOOL_RUNTIME_CONFIG.crush for detection/status, and bumps EXPECTED_CODE_COUNT (18 -> 19) plus the catalog-count/schema tests that enumerate the full tool list. Co-authored-by: dopaemon <polarisdp@gmail.com> Inspired-by: https://github.com/decolua/9router/pull/1233 * chore(changelog): restore release entries + add crush cli bullet --------- Co-authored-by: dopaemon <polarisdp@gmail.com> * feat(dashboard): suggest HuggingFace Hub media models (#5990) * feat(dashboard): suggest HuggingFace Hub media models MVP scope: - imageRegistry.ts: add an image kind entry for the huggingface provider (HF Inference API text-to-image), with a dedicated "huggingface-image" format since the endpoint returns raw image bytes rather than JSON. - New handler open-sse/handlers/imageGeneration/providers/huggingface.ts, wired into imageGeneration.ts's format dispatch. - New pure helper module open-sse/services/hfModelSuggestions.ts: maps a dashboard media kind to an HF Hub pipeline_tag and sorts/limits raw HF Hub search results (unit-tested directly). - New route GET /api/v1/providers/suggested-models proxies the public HF Hub models search API server-side (Zod-validated query, buildErrorBody on every error path, no HF token exposed client-side — this project has no server-side HF search token config, so it calls unauthenticated). - UI: ImageExampleCard now fetches suggested HF Hub models for the huggingface provider and merges them into the model picker as a selectable chip row, alongside the existing static provider models list. - i18n: adds media.suggestedModels to en.json only. Co-authored-by: yicone <yicone@gmail.com> Inspired-by: https://github.com/decolua/9router/pull/1633 * chore(changelog): restore release entries + add hf-hub media suggest bullet --------- Co-authored-by: yicone <yicone@gmail.com> * feat(dashboard): collapse and sort provider quota rows by remaining (#5977) * feat(dashboard): collapse and sort provider quota rows by remaining Sort the expanded quota list by remaining percentage (highest first) and collapse it to the first 3 rows by default, with a "Show N more" / "Show less" toggle when a connection reports more than 3 quotas. This keeps the most at-risk quotas out of view below a long list of healthy ones. Extracts the sort/slice logic into pure helpers (sortQuotasByRemaining, getVisibleQuotas) exported from QuotaCardExpanded.tsx and unit-tests them directly. Co-authored-by: CườngNH <j2.cuong@gmail.com> Inspired-by: https://github.com/decolua/9router/pull/1919 * chore(changelog): restore release entries + add quota collapse/sort bullet --------- Co-authored-by: CườngNH <j2.cuong@gmail.com> * feat(providers): refresh The Old LLM (Free) model catalog (#5181) * feat(dashboard): add tool-source diagnostics settings toggle (#5978) * feat(dashboard): add tool-source diagnostics settings toggle Adds a Settings > Advanced card (cloned from DebugModeCard) that lets operators flip the existing `logToolSources` flag from the UI instead of editing the DB row directly. The backend gate (chatCore.ts) and DB default were already present but had no toggle. Also adds `logToolSources` to the /api/settings Zod PATCH schema (it is `.strict()`, so the key was previously rejected) and en-only i18n strings. Co-authored-by: DuyPrX <93126969+DuyPrX@users.noreply.github.com> Inspired-by: https://github.com/decolua/9router/pull/1825 * chore(changelog): restore release entries + add tool-source toggle bullet --------- Co-authored-by: DuyPrX <93126969+DuyPrX@users.noreply.github.com> * feat(oauth): import Codex connection from a raw ChatGPT access token (#5995) * feat(oauth): import Codex connection from a raw ChatGPT access token OmniRoute's only Codex import path (/api/oauth/codex/import) required both access_token and refresh_token, leaving no import path for a user who only has a bare ChatGPT website access token (no refresh token). - src/lib/db/providers.ts: createProviderConnection gains an explicit authType "access_token" branch — intentionally never deduped (no stable long-lived identity to match on) — and derives the connection name from email/name the same way "oauth" does. - src/lib/oauth/services/codexImport.ts: export extractCodexAccountInfo so the new import path reuses the existing JWT decode instead of duplicating one. - New route POST /api/oauth/codex/import-token (Zod-validated body { accessToken, name? }); errors routed through buildErrorBody / sanitizeErrorMessage. The executor's refreshCredentials() already degrades safely to null when there is no refresh token, forcing re-auth on expiry instead of a refresh exchange. - OAuthModal.tsx: the callback-URL manual-paste path for codex now detects an eyJ-prefixed pasted token and posts it to the new endpoint, mirroring the existing grok-cli raw-token paste pattern. Co-authored-by: ryanngit <74137224+ryanngit@users.noreply.github.com> Inspired-by: https://github.com/decolua/9router/pull/1290 * chore(changelog): restore release entries + add codex token-import bullet --------- Co-authored-by: ryanngit <74137224+ryanngit@users.noreply.github.com> * fix(resilience): parse Retry-After from 429 JSON body for cooldown (#5974) Integrated into release/v3.8.44 — parse Retry-After from 429 JSON body for cooldown (incl. #6013 retry-after-json extraction by @KooshaPari). * fix(embeddings): forward connection-level proxy to embedding requests (#5975) Integrated into release/v3.8.44 — forward connection-level proxy to embedding requests. * fix(api): guard shared API client against non-JSON error responses (#5973) Integrated into release/v3.8.44 — guard shared API client against non-JSON error responses. * feat(dashboard): surface Codex banked reset credits per account (#5199) * feat(providers): add NVIDIA NIM image generation (#5971) * feat(providers): add NVIDIA NIM image generation NVIDIA already exists as a chat provider (integrate.api.nvidia.com, OpenAI-compatible) but image generation is served on a different host (ai.api.nvidia.com/v1/genai/<model>) with a native NIM body shape, so it gets a dedicated `nvidia-nim` image format and handler rather than reusing the OpenAI image path. Adds the 4 FLUX models (flux.1-dev, flux.1-schnell, flux.1-kontext-dev, flux.2-klein-4b) to IMAGE_PROVIDERS, plus handleNvidiaNimImageGeneration() which shapes the per-model NIM request body (flux.1-dev's mode/cfg_scale and 768-1344px/64px-increment dimension validation, flux.1-kontext-dev's required input image + aspect_ratio, schnell/klein's optional array-form edit image) and normalizes the NIM response (artifacts[]/images[]/data[]/ single-value shapes) into the OpenAI `{created, data}` shape. Co-authored-by: eng2007 <aleksey.semenov@gmail.com> Inspired-by: https://github.com/decolua/9router/pull/1195 * chore(changelog): restore release entries + add nvidia-nim image bullet --------- Co-authored-by: eng2007 <aleksey.semenov@gmail.com> * feat(providers): add Augment (Auggie CLI) local provider (#5972) * feat(providers): add Augment (Auggie CLI) local provider Adds a new local, no-auth provider that spawns the user's local `auggie` CLI (`auggie --print --quiet --model <m> --`) and pipes a flattened prompt via stdin, wrapping stdout as an OpenAI-compatible SSE stream or a single chat.completion JSON body depending on the request's `stream` flag. Auth is delegated entirely to `auggie login` outside OmniRoute — the connection is registered `noAuth: true` and `refreshCredentials()` is a no-op, matching the existing `NOAUTH_PROVIDERS` credential-less flow (synthetic connection, no DB row required). An optional connection row is still admitted via `FREE_APIKEY_PROVIDER_IDS` for display/priority tracking, consistent with `opencode`. The dashboard "Test Connection" flow spawns `auggie --version` to confirm the CLI is installed and runnable, since there is no API key to validate upstream. Security hardening (spawn is an untrusted-input sink): - Command injection: spawn no longer passes `shell: true` on Windows. The binary is resolved to a concrete path/name and argv is handed straight to the OS loader, so no cmd.exe metacharacter interpretation is possible. - Argument injection (flag smuggling): `model` is validated against the registry allowlist (`auggieProvider.models`) before any spawn — a model that is unknown or starts with "-" is rejected with a sanitized error and the subprocess is never started. A trailing `--` marks end-of-options in the argv as belt-and-suspenders. Co-authored-by: chamdanilukman <16629923+chamdanilukman@users.noreply.github.com> Inspired-by: https://github.com/decolua/9router/pull/1200 * test(golden): regenerate translate-path for auggie provider --------- Co-authored-by: chamdanilukman <16629923+chamdanilukman@users.noreply.github.com> * feat(providers): add ModelScope OpenAI-compatible provider (#5965) * feat(providers): add ModelScope OpenAI-compatible provider Ports ModelScope (Alibaba 魔搭) as a new API-key, OpenAI-compatible provider — upstream 9router PR #1764. The upstream PR hardcoded `https://api-inference.modelscope.ai/...` (`.ai` TLD); verified against ModelScope's own API-Inference docs and third-party integration guides that the real production domain is `api-inference.modelscope.cn` (`.cn` TLD) and shipped that instead. Also drops the PR's static 5-model snapshot in favor of `passthroughModels: true` with an empty seed list + `modelsUrl`, since ModelScope's open-model catalog moves fast. Updates the providers-constants-split characterization test's hardcoded APIKEY_PROVIDERS count (159 -> 160) to match the new entry. Co-authored-by: Umar Javed <114807145+tn5052@users.noreply.github.com> Inspired-by: https://github.com/decolua/9router/pull/1764 * chore(changelog): restore release entries + add modelscope bullet * test(golden): regenerate translate-path for modelscope provider --------- Co-authored-by: Umar Javed <114807145+tn5052@users.noreply.github.com> * feat(providers): add Qiniu OpenAI-compatible provider (#5966) * feat(providers): add Qiniu OpenAI-compatible provider Wires Qiniu (七牛云) AI inference gateway as a BYOK API-key provider. Qiniu proxies many upstream models (DeepSeek V3/V4, Claude, Kimi and more) behind a single key, so it ships with an empty static seed and relies on passthroughModels + the live /v1/models catalog instead of a single stale hardcoded model id. - metadata: src/shared/constants/providers/apikey/gateways.ts - registry entry: open-sse/config/providers/registry/qiniu/index.ts (format openai, executor default, bearer auth, baseUrl https://api.qnaigc.com/v1/chat/completions, modelsUrl https://api.qnaigc.com/v1/models) - added to NAMED_OPENAI_STYLE_PROVIDERS so model import serves the live catalog and falls back to the (empty) local catalog on error, same pattern as the existing dgrid/zenmux/orcarouter gateways - tests: tests/unit/qiniu-provider.test.ts (metadata, registry resolution, passthrough validation, live /v1/models fetch + fallback) Co-authored-by: JiangZhuo <jiangzhuo@qiniu.com> Inspired-by: https://github.com/decolua/9router/pull/911 * chore(changelog): restore release entries + add qiniu bullet * test(golden): regenerate translate-path for qiniu provider * test(providers): bump APIKEY count 160→161 for qiniu --------- Co-authored-by: JiangZhuo <jiangzhuo@qiniu.com> * feat(providers): add b.ai OpenAI-compatible provider (#5969) * feat(providers): add b.ai OpenAI-compatible provider Adds bai as a new OpenAI-compatible BYOK provider, distinct from the existing thebai/theb.ai provider, using passthrough model discovery (no hardcoded model list, live catalog served from api.b.ai/v1/models). Co-authored-by: Delynn Assistant <zhen@dkzhen.org> Inspired-by: https://github.com/decolua/9router/pull/963 * test(golden): regenerate translate-path for b.ai provider * test(providers): bump APIKEY count 161→162 for b.ai --------- Co-authored-by: Delynn Assistant <zhen@dkzhen.org> * feat(providers): add Nube.sh OpenAI-compatible provider (#5936) * feat(providers): add Nube.sh OpenAI-compatible provider Nube.sh is a live BYOK OpenAI-compatible gateway (LiteLLM proxy) at https://ai.nube.sh/api/v1, Bearer/API-key auth. Registered as an apikey inference-host with an OpenAI-format, default-executor registry entry. Its live model catalog is only reachable with a valid key (/api/v1/models returns 401 unauthenticated), so no model IDs are hardcoded — the entry uses passthroughModels + modelsUrl for live enumeration instead of shipping unverifiable IDs. Co-authored-by: whale9820 <87256750+whale9820@users.noreply.github.com> Inspired-by: https://github.com/decolua/9router/pull/2294 * test(golden): regenerate translate-path for nube provider * test(providers): bump APIKEY count 162→163 for nube --------- Co-authored-by: whale9820 <87256750+whale9820@users.noreply.github.com> * feat(providers): add Charm Hyper OpenAI-compatible provider (#5961) * feat(providers): add Charm Hyper OpenAI-compatible provider Registers Charm Hyper (hyper.charm.land) as a new API-key gateway provider: OpenAI-compatible chat completions format, bearer auth, free tier (100 monthly Hypercredits). Models are resolved via passthrough (modelsUrl + live /v1/models import) instead of a hardcoded upstream model list, since the specific model catalog is not publicly documented. Co-authored-by: whale <admin@dyntech.cc> Inspired-by: https://github.com/decolua/9router/pull/2006 * test(golden): regenerate translate-path for charm-hyper provider * test(providers): bump APIKEY count 163→164 for charm-hyper --------- Co-authored-by: whale <admin@dyntech.cc> * feat(providers): add SumoPod and X5Lab OpenAI-compatible providers (#5963) * feat(providers): add SumoPod and X5Lab OpenAI-compatible providers Both are OpenAI-compatible BYOK aggregator gateways, wired via the default executor with bearer API-key auth. Neither ships a hardcoded model list — both use passthroughModels with an empty seed list and a live /v1/models fetcher, so the catalog always reflects what each gateway actually serves instead of speculative model IDs. - SumoPod: https://ai.sumopod.com/v1/chat/completions (sk- keys) - X5Lab: https://api.x5lab.dev/v1/chat/completions (x5- keys) Regression guard: tests/unit/sumopod-x5lab-provider.test.ts. Co-authored-by: Rigel Ramadhani Waloni <rigel8911@gmail.com> Inspired-by: https://github.com/decolua/9router/pull/1288 * chore(changelog): restore release entries + add sumopod/x5lab bullet * test(golden): regenerate translate-path for sumopod + x5lab providers * test(providers): bump APIKEY count 164→166 for sumopod + x5lab --------- Co-authored-by: Rigel Ramadhani Waloni <rigel8911@gmail.com> * feat(server): support reverse-proxy basePath deployment (#5992) * feat(server): support reverse-proxy basePath deployment Adds OMNIROUTE_BASE_PATH (opt-in, empty by default) to next.config.mjs using Next.js's native basePath support so a deployment behind a reverse-proxy subpath (e.g. https://host/omniroute/) works without manual header stripping. Next.js strips the configured prefix from nextUrl.pathname before route classification, so classifyRoute() and isLocalOnlyPath() keep matching un-prefixed paths. The two hardcoded auth redirect targets in src/server/authz/pipeline.ts (root "/" -> "/dashboard" and unauthenticated dashboard -> "/login") now prefix with request.nextUrl.basePath so they stay inside the deployed subpath. Default empty basePath is a no-op for existing root-path deployments. Co-authored-by: zocomputer <help@zocomputer.com> Inspired-by: https://github.com/decolua/9router/pull/1810 * docs(env): document OMNIROUTE_BASE_PATH in .env.example + ENVIRONMENT.md; restore changelog * docs(env): document AUGGIE_BIN + CLI_AUGGIE_BIN (base-red from #5972 auggie) --------- Co-authored-by: zocomputer <help@zocomputer.com> * refactor(combo): extract buildTargetTimeoutRunner from handleComboChat (#6036) Bloco J (hot-path decomposition), Task 1. Extract the per-target-timeout dispatch wrapper (handleComboChat's handleSingleModelWithTimeout closure) verbatim into the leaf combo/targetTimeoutRunner.ts as a factory buildTargetTimeoutRunner({handleSingleModel, comboTargetTimeoutMs, log}). The per-model abort still comes from target.modelAbortSignal, so the outer request signal is intentionally not a dependency. Host call-sites unchanged. combo.ts shrinks ~60 LOC; leaf is 91 LOC (<800). Body byte-identical (verbatim), no cycle. This is the first slice toward extracting the shared attempt-loop/success/error handlers (Tasks 3-4) that de-duplicate handleComboChat and handleRoundRobinCombo. Adds a dedicated test (5) so the failover path can be mutated independently. Consumer tests stay green (combo-strategy-fallbacks 24, combo-499-abort 5, empty-content-failover 3, body-400-stop 1, priority-quota-exhaustion 2, rr-streaming-lock 1, rr-session-stickiness 2). Plan: _tasks/superpowers/plans/2026-07-03-blocoJ-combo-hotpath-decomposition.md * feat(cli-tools): add CodeWhale CLI tool (#5996) CodeWhale (https://github.com/Hmbown/CodeWhale) is the actively-maintained successor to DeepSeek TUI — same author, renamed project. Added as a dual entry alongside the existing "deepseek-tui" catalog entry (rather than a hard rename) so users who still run the old DeepSeek TUI binary keep a working dashboard card, while new users are steered to "codewhale". New /api/cli-tools/codewhale-settings route writes the primary ~/.codewhale/config.toml and keeps an existing legacy ~/.deepseek/config.toml in sync (read fallback + best-effort write sync), mirroring deepseek-tui-settings/route.ts. CLI_TOOLS and cliRuntime catalogs updated; catalog cardinality tests/constants bumped accordingly (18→19 visible code tools, 28→29 total). Inspired-by: https://github.com/decolua/9router/pull/1761 Co-authored-by: aristorinjuang <aristorinjuang@gmail.com> * feat(i18n): auto-detect browser language on first visit (#5979) * feat(i18n): auto-detect browser language on first visit Adds a pure detectBrowserLocale() matcher (exact match, zh-HK/zh-MO folded to zh-TW, language-prefix match, else null) plus a client-only LocaleAutoDetect component mounted once in the root layout. On first visit (no locale cookie set), it reads navigator.languages, computes a match against the supported locales, and persists it via the same cookie/localStorage writer LanguageSelector already used for manual selection (now extracted to shared/lib/persistLocale.ts) before refreshing the router. Co-authored-by: anmingwei <anmingwei@dobest.com> Inspired-by: https://github.com/decolua/9router/pull/1324 * chore(changelog): restore release entries + add browser-lang-detect bullet --------- Co-authored-by: anmingwei <anmingwei@dobest.com> * fix(dashboard): render Update-now API errors as text, not the raw envelope object (#5991) (#6028) Integrated into release/v3.8.44 — fix(dashboard) render Update-now API errors as text, not the raw envelope object (#5991). Merged with --admin: the fix is a one-line frontend change funneling the error body through the already-tested extractApiErrorMessage() helper, guarded by tests/unit/ui/home-update-error-render-5991.test.ts (3/3 pass, 3/3 fail on pre-fix source). The release branch is under a heavy parallel-merge storm (tip advanced ~6× mid-CI), so the branch is synced to the latest tip and landed atomically to avoid perpetual CONFLICTING; unit-shard reds seen earlier were pre-existing base-reds/flakes unrelated to this source-scan-only change. * feat(api): expose provider plugin manifest (#6001) * feat(api): expose provider plugin manifest * test(translator): split responses chat request coverage * test(mutation): register provider coverage tests * feat(api): expose provider plugin manifest * fix(ci): fail closed for prerelease latest promotion * chore(ci): reconcile provider manifest complexity gate * feat(api): expose provider plugin manifest * test(translator): split responses chat request coverage * test(mutation): register provider coverage tests * fix(ci): fail closed for prerelease latest promotion * chore: rebase onto release tip; drop out-of-scope translator test split + promote-script tweak Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * docs(changelog): add provider plugin manifest entry Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * chore(stryker): register account-fallback-retry-after-json test (base-red) Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> --------- Co-authored-by: kooshapari <kooshapari@users.noreply.github.com> Co-authored-by: Diego Rodrigues de Sa e Souza <diegosouza.pw@outlook.com> Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(providers): add CN sign-up geo-restriction notices for SenseNova & StepFun (#5462) * feat(sidecar): advertise provider manifest url (#6007) * feat(sidecar): advertise provider manifest url via X-OmniRoute-Provider-Manifest-Url header Re-cut onto release tip: manifest-url feature only (dropped stale-base noise). Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * docs(changelog): add sidecar manifest-url entry Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * chore(complexity): rebaseline 2009->2015 (inherited release-tip drift; feature adds 0) Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> --------- Co-authored-by: KooshaPari <KooshaPari@users.noreply.github.com> Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(autoCombo): latency/speed-optimized routing mode + omniroute_pick_fastest_model MCP tool (#6011) * feat(autoCombo): latency/speed-optimized routing mode + omniroute_pick_fastest_model MCP tool * test(translator): split responses chat request coverage * refactor(mcp): extract fastest-model tool modules * fix(i18n): cover provider icon and cors labels * test(mutation): register latency coverage files * test(ci): collect executor unit tests * refactor(ci): reduce latency path complexity * fix(mcp): include models catalog module * feat(autoCombo): latency/speed-optimized routing + omniroute_pick_fastest_model MCP tool Re-cut onto release tip: keep speed-routing + MCP tool + supporting catalog split; drop out-of-scope translator split, en.json/ci.yml/package.json orphans, and unrelated proxyFetch/responsesStreamHelpers/tokenLimitCounter refactors. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> --------- Co-authored-by: kooshapari <kooshapari@users.noreply.github.com> Co-authored-by: Diego Rodrigues de Sa e Souza <diegosouza.pw@outlook.com> Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * docs(changelog): restore #5181/#5199/#5462 feature bullets eaten by merge * feat(usage): on-demand period-scoped usage-data reset (re-cut onto release tip) (#5831) * chore(quality): rebaseline eslintWarnings 4199->4256 + cognitiveComplexity 860->861 (v3.8.44 cycle drift) Inherited v3.8.44 cycle drift measured on release tip72ee80649by the release-green pre-flight during the /review-prs fix-batch round. The Quality Ratchet does NOT run on PR->release fast-gates, so eslint warnings + cognitive complexity accrue unmeasured across the cycle. Cyclomatic complexity is already green (2012 < baseline 2015) and needs no bump. Each value carries a dated justification note; no production code touched. * feat(claude-code): opt-in auto-permission classifier compat mode (re-cut onto release tip) (#5810) * feat(providers): client-identity header profiles for compatible nodes (re-cut) + forbid cookie in custom headers (#5812) * docs(openapi): document 9 newly-added routes to restore coverage ratchet (v3.8.44) Documents the routes added this cycle that dropped openapiCoverage 36.9%->36.2% below the ratchet baseline: 2 public v1 endpoints (/v1/ocr Mistral-OCR-compatible, /v1/audio/translations Whisper-compatible) with full request/response specs, plus 7 dashboard/CLI-local routes marked x-internal:true (suggested-models, provider-plugin- manifest, keys/{id}/devices, settings/purge-usage-history, oauth/codex/import-token, cli-tools crush-settings + codewhale-settings). Coverage 36.2%->37.8% (207/547), above baseline 36.9. check:openapi-routes/security-tiers/fabricated-docs all pass. * refactor(sse): decompose handleComboChat auto-strategy region (Block J Task 2 — parseAutoConfig + resolveAutoStrategyOrder) (#6049) * refactor(sse): extract pure parseAutoConfig leaf from handleComboChat Block J Task 2 (safe slice): the auto-strategy config-resolution block in handleComboChat is a pure function of (combo, eligibleTargets) with no side effects, no early returns and no mutation. Extract it verbatim into open-sse/services/combo/autoConfig.ts::parseAutoConfig so the god-function shrinks and the derivation is independently unit-testable. Behavior is byte-identical (verbatim-audited); combo.ts 3309->3280 LOC. Adds tests/unit/combo-auto-config-split.test.ts (5 cases) pinning the strategy-precedence, candidate-pool, weights and fallback derivations. * refactor(sse): extract resolveAutoStrategyOrder leaf from handleComboChat Block J Task 2 (coupled slice): the ~215-line `if (strategy === "auto")` branch of handleComboChat is extracted into open-sse/services/combo/resolveAutoStrategy.ts::resolveAutoStrategyOrder. The branch is a control-flow region (mutates orderedTargets + autoUsedExplicitRouter, early-returns 429, side-effect _registerExecutionCandidates), so it is not a pure byte-identical move: the two `return unavailableResponse(...)` exits become `{ earlyResponse }` and the mutated locals are returned instead of closed over. Every other logic line is verbatim (semantic diff = only those wrappers + the deeper getLKGP import path). `buildAutoCandidates` lives in combo.ts, so it is injected via deps to keep the leaf acyclic (same DI pattern as buildTargetTimeoutRunner) — which also makes the branch independently testable. combo.ts 3280->3065 LOC. typecheck:core + check:cycles clean; dead host imports removed. 60/60 consumer tests (router-strategies / auto-combo-engine / combo-strategy-fallbacks / scoring-clamp / candidate-expansion / hidden-models) cover the routable path end-to-end; new tests/unit/combo-resolve-auto-strategy-split.test.ts pins the DI contract + the early-429 and default-ordering exits. * test(sse): point quota-bypass source scan at resolveAutoStrategy leaf The 'auto combo disables hard provider quota cutoffs when relay requests bypass' source scan asserted combo.ts contains the bypass logic (relayOptions?.bypassProviderQuotaPolicy === true + quotaPreflight enabled:false). That block was extracted verbatim into combo/resolveAutoStrategy.ts (Block J Task 2), so the scan now reads the leaf. Behavior unchanged. * fix(ci): release-green base-reds — #5695 test regex + file-size rebaseline (#6093) - tests/unit/ui/quick-start-api-keys-link-5695.test.ts: tolerate Prettier splitting <Link href=...> across lines (\s+) so the step1Desc regex matches the multi-line /dashboard/api-manager Link instead of skipping to step2's single-line /dashboard/providers Link. Code is correct; the test was brittle. - config/quality/file-size-baseline.json: rebaseline 5 files that grew via already-merged PRs on the release tip (ApiManagerPageClient 3017->3058, OAuthModal 969->989, cliRuntime 1090->1100, webProvidersA 805->809, deepseek-web.test 1081->1092). Dated note added; shrink tracked in #3501. * fix(translator): wrap Kiro system prompt in <system-reminder> (port from 9router#2306) (#6053) Kiro/CodeWhisperer has no system role, so system messages were normalized to a user turn with no wrapper — the full Claude Code system prompt then appeared as raw user text, polluting the model context. Wrap system-origin content in <system-reminder> tags before merging it into the Kiro user message. Real user turns are unaffected. Existing history-merge tests aligned to the wrapped value. Reported-by: VitzS7 (https://github.com/decolua/9router/issues/2306) * fix(translator): strip multipleOf from antigravity/gemini tool schemas (port from 9router#2309) (#6052) `multipleOf` is not part of the Gemini/antigravity OpenAPI 3.0 schema subset, so leaving it in function_declaration parameters triggered a hard upstream 400 ("Unknown name multipleOf"). Add it to GEMINI_UNSUPPORTED_SCHEMA_KEYS so it is stripped at every schema level; minimum/maximum stay (Gemini accepts them). Reported-by: abil0321 (https://github.com/decolua/9router/issues/2309) * fix(kimi-web, qwen-web): align model catalog with live /models + map scenario per model (#5915) * fix(kimi-web): align catalog with live models Update the kimi-web catalog and request scenario selection to match www.kimi.com's live GetAvailableModels response. * fix(qwen-web): stop aliasing qwen3-coder-plus Keep qwen3-coder-plus as its own model because it is present in the live Qwen web models catalog. * feat(minimax): extract M3 <think> to reasoning_content on OpenAI-format tiers (#6050) MiniMax M3 is registered with format:"openai" on 8 provider tiers (trae, huggingchat, bazaarlink, ollama-cloud, opencode, cline, opencode-zen, codebuddy-cn), where its raw <think>...</think> tags leaked directly into `content` instead of surfacing as a separate `reasoning_content` field. OmniRoute already has the extraction primitive (extractThinkingFromContent in responseSanitizer/reasoning.ts); it was just gated to deepseek-r1/r1-distill/qwq. Extend the allowlist (isTextualReasoningTagNativeRoute) with a minimax-m3-only pattern, excluding the two direct minimax/minimax-cn tiers, which stay on Anthropic's Messages format (targetFormat: "claude") and already surface reasoning natively. Inspired-by: https://github.com/decolua/9router/pull/2231 Co-authored-by: Diego Rodrigues de Sa e Souza <diegosouza.pw@gmail.com> Co-authored-by: zmf963 <19422469+zmf963@users.noreply.github.com> * fix: unwrap Cline response envelope (#6046) Co-authored-by: KooshaPari <KooshaPari@users.noreply.github.com> * refactor(sse): extract applyStrategyOrdering leaf from handleComboChat (Block J Task 3) (#6063) * refactor(sse): extract applyStrategyOrdering leaf from handleComboChat Block J Task 3: the ~177-line else-if chain covering every non-auto combo strategy (lkgp / strict-random / random / fill-first / p2c / least-used / cost-optimized / reset-aware / reset-window / context-optimized / headroom / quota-share) is extracted into open-sse/services/combo/applyStrategyOrdering.ts::applyStrategyOrdering. Each branch only reorders orderedTargets (no early returns, no other mutable state), so the extraction is a clean verbatim move returning the reordered list; the host replaces the chain with `else { orderedTargets = await applyStrategyOrdering(strategy, orderedTargets, deps); }`. Semantic diff vs the original chain = only the leading `if` (was `} else if`), the trailing return and the deeper getLKGP import path — no logic line changed. None of the 13 strategy helpers live in combo.ts, so no DI/cycle (unlike the auto branch). combo.ts 3065->2883 LOC (3309->2883 across Task 2+3). typecheck:core + check:cycles clean; 9 dead host imports removed (targetSorters block emptied). 47/47 consumer tests (router-strategies / combo-strategy-fallbacks / rr-session-stickiness / tag-routing) cover the DB-backed branches end-to-end; new tests/unit/combo-apply-strategy-ordering-split.test.ts pins random / fill-first / unknown exits. * test(sse): point #2359 modelStr-guard scans at applyStrategyOrdering leaf The LKGP fallback + non-auto strategy ordering (the two target.modelStr string- method call sites) were extracted verbatim from combo.ts into the applyStrategyOrdering leaf (Block J Task 3). The #2359 source scans now read the leaf that owns those usages; the guard and the no-unguarded-usage assertions are unchanged in intent. * chore(ci): scan combo strategy leaves in check:known-symbols Block J decomposed the combo dispatch: the `strategy === "..."` branches for the 12 non-auto strategies moved to combo/applyStrategyOrdering.ts and the auto branch to combo/resolveAutoStrategy.ts. The known-symbols gate previously scanned only combo.ts, so it would report those strategies as canonicalNotHandled. Scan all three dispatch files. Verified: 18/18 canonical strategies via dispatch. * fix(combo): fallback to sibling model on 500 for per-model-quota providers (#5976) * fix(combo): fallback to sibling model on 500 for per-model-quota providers Two issues prevented combo fallback when gemini/gemma-4-31b-it returned 500: 1. targetExhaustion: connection-level exhaustion marked the shared gemini connection as exhausted, skipping the sibling model (gemma-4-26b-a4b-it). Skip markConnectionLevelExhaustion for per-model-quota providers (gemini, github, passthrough, compatible) since a model-level 500 does not mean the connection is bad. 2. combo retry loop: the auth layer records a model lockout on 500, but the retry loop did not check isModelLocked before retrying — it retried the same locked model instead of falling back. Add isModelLocked guard before the transient-retry decision. * fix tests timeout * fix: clear quota fallback CI gates * quality-gate: extract test SSE stream helpers * drop scope creep * fix(combo): retry sibling models only on 500 errors * fix(combo): reconcile onto release/v3.8.44 — keep targetExhaustion 500 fix, drop slow integration test Reconciled by maintainer onto the current release tip: - kept the core fix (targetExhaustion.ts model-500 guard for per-model-quota providers + the isModelLocked retry early-return in combo.ts) and its unit test - dropped tests/integration/combo-concurrent-failure-recovery.test.ts + _sseTestHelpers.ts: they use Math.random()-based delays and 30s timeouts, run >3min and are flake-prone in the test:integration CI job; the unit test (tests/unit/combo/combo-target-exhaustion.test.ts, 21 cases) fully covers the fix - CHANGELOG entry added Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> --------- Co-authored-by: Koosha Pari <kooshapari@gmail.com> Co-authored-by: Diego Rodrigues de Sa e Souza <diegosouza.pw@outlook.com> Co-authored-by: hartmark <hartmark@users.noreply.github.com> Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(xai): surface Grok usage on quota dashboard via local usageHistory aggregation (#5806) xAI has no public per-account quota API (the billing console requires a session cookie, not an API key). Add getXaiUsage(connectionId), mirroring the existing Xiaomi MiMo self-track pattern: sum tokens routed to the connection from usage_history via getMonthlyProviderTokensForConnection and surface them as a cumulative, uncapped quota (unlimited: true, remaining: 100 — xAI has no fixed monthly cap). Register 'xai' in USAGE_FETCHER_PROVIDERS and wire a switch case in getUsageForProvider. Inspired-by: https://github.com/decolua/9router/pull/2150 Co-authored-by: ron <devestacion@gmail.com> * feat(services): add Mux managed embedded service (#6034) Adds Mux (coder/mux — local agent-orchestration daemon) as a fourth-tier embedded service built on the existing ServiceSupervisor framework, the same shape as 9Router and CLIProxyAPI: - Installer (src/lib/services/installers/mux.ts): npm install/update via runNpm (array args + env-based prefix, no shell interpolation), modeled on ninerouter.ts. Mux ships an npm package (`mux`) with a documented headless `mux server --host <host> --port <port>` mode, so no git-clone+build path was needed. - Registered in bootstrap.ts (SERVICES[] + buildSpawnArgsFactory). - DB seed migration 113 (version_manager row, not_installed/auto_start=0). - 7 API endpoints under /api/services/mux/ (install/start/stop/restart/ update/status/auto-start) plus the shared [name]/logs SSE endpoint, mirroring the cliproxy route shape and delegating errors through createErrorResponse(). - Dashboard tab (MuxServiceTab) reusing ServiceStatusCard, ServiceLifecycleButtons, AutoStartToggle, ServiceLogsPanel. - Docs: EMBEDDED-SERVICES.md (service table, architecture diagram, API reference, key-injection section), openapi.yaml, ENVIRONMENT.md, .env.example. Security: - Every /api/services/mux/* route is covered by the existing LOCAL_ONLY_API_PREFIXES "/api/services/" prefix (Hard Rule #17); added an explicit isLocalOnlyPath regression test for all 8 routes. - Mux binds to 127.0.0.1 explicitly (never 0.0.0.0) as defense-in-depth, since it orchestrates AI agents that can execute host commands. - The bearer token is generated the same way as 9Router's key (getOrCreateApiKey) and injected via MUX_SERVER_AUTH_TOKEN (mux's documented env form) rather than a CLI flag, so it never appears in `ps`/process listings. - No shell interpolation anywhere in the installer (Hard Rule #13): all npm/spawn args are static arrays; the install prefix and auth token travel via the env option. Inspired-by: https://github.com/decolua/9router/pull/1802 Co-authored-by: Ansh7473 <Ansh7473@users.noreply.github.com> * feat(services): promote Bifrost to embedded/supervised service (#5670) (#5817) Promotes Bifrost (@maximhq/bifrost — Go AI-gateway) from an env-only relay sidecar to a first-class embedded/supervised service, matching the existing cliproxy/9router model. Implements item #2 of #5670; the broader RouterBackend contract (items #1, #3-#5) stays out of scope. - Installer (npm-style, ninerouter model): install/update/getInstalledVersion/ getLatestVersion (1h cache)/resolveSpawnArgs (Go single-dash flags, pinned BIFROST_TRANSPORT_VERSION), needsApiKey=false - Bootstrap SERVICES entry (healthPath /v1/models) + spawn-args factory branch - Migration 113 seeds the version_manager row (not_installed, port 8080, auto_update=1, provider_expose=1) - 7 lifecycle API routes under /api/services/bifrost/ (verbatim from cliproxy, errors sanitized) — loopback-only via existing LOCAL_ONLY_API_PREFIXES - Shared [name]/logs branch for bifrost - Dashboard tab + registration in the services page shell - Relay auto-wiring: getBifrostRoutingConfig defaults BIFROST_BASE_URL to the supervised port when the instance is running; explicit env still wins; the env-only relay path (/v1/relay/.../bifrost) stays unchanged (compat layer) - Docs (EMBEDDED-SERVICES, openapi) + unit tests (installer/route-guard/routing, 19 tests) + RUN_SERVICES_INT-gated integration lifecycle Note: the actual Go-binary install/start/health path requires a documented VPS live-test before merge (Hard Rule #18 / spec section 7); the gated integration harness is the vehicle for that run. * fix(ci): document BIFROST_PORT to clear env-doc-sync base-red The Bifrost embedded-service merge referenced process.env.BIFROST_PORT (src/lib/services/bootstrap.ts, default 8080) without adding it to .env.example / ENVIRONMENT.md, so check:env-doc-sync failed on the release tip and reddened Fast Quality Gates for every open PR->release. Docs-only. * fix(providers): emulate OpenAI tool_calls in GitLab Duo executor (#6051) (#6111) Co-authored-by: felssxs <felssxs@users.noreply.github.com> * fix(providers): strip orphan tool_result on Antigravity MITM path (#6026) (#6115) * fix(registry): update grok-cli model context lengths (#5913) grok-build 128k→256k, grok-composer-2.5-fast 128k→200k to match actual Grok CLI /context capacities so context-aware routing stops filtering these models out. Registry-only. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(proxy): batch delete, auto-test, health scheduler + transitive alias fix (#5918) Proxy-registry batch management (batch-delete, auto-test, background health scheduler) + fix resolveProviderAlias to follow the alias chain transitively (oc -> opencode -> opencode-zen). Probe target now operator-configurable via PROXY_HEALTH_TEST_URL. Scope-creep files from the original branch dropped. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * feat(minimax): extract M3 reasoning_content on OpenAI-format tiers (#6073) MiniMax M3 leaks raw <think>...</think> into content on 8 OpenAI-format provider tiers; extract it into reasoning_content, leaving the direct minimax/minimax-cn (Claude-format) tiers untouched. Replacement for the stale #5804 branch. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix(ci): harden provider translate-path golden across CI runners (#6076) Normalize OS/arch-derived request headers (X-Stainless-Os/Arch, (OS;arch) UAs, and Antigravity's os.platform()-derived platform substring) in the golden so the test is runner-independent. Fixes the Mac-literal Antigravity UA that would have failed on Linux CI. Supersedes stale #6002. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * test(embeddings): pin seeded connection to direct egress in route-edge-coverage (#5975 collateral) #5975 made the embeddings service honor the connection-level proxy. The pre-existing route-edge-coverage embeddings edge-case tests seed an openai connection while the settings-proxy suite has left a provider-level proxy (provider.local:8080) in the shared DATA_DIR that resetStorage() does not clear — inert before #5975, but now the leaked proxy fast-fails the embedding upstream with PROXY_UNREACHABLE. These tests do not exercise proxying, so seedOpenAIConnection now pins the connection to proxyEnabled:false, making resolveProxyForConnection return a direct egress regardless of leaked global proxyConfig. No assertions weakened; 16/16 in the file pass. Regression surfaced by the concurrency=1 full-suite run; passes on #5975's parent, red after it. * fix(config): externalize ws for copilot-m365-web executor (#6130, closes #6062) Re-lands the #6098 ws-externalization fix onto release/v3.8.44 (it had merged to main by mistake and was reverted). Externalize ws/bufferutil/utf-8-validate so the copilot-m365-web WebSocket masking path works at runtime. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix(providers): update Perplexity Web models (#6106) Refresh the Perplexity Web model catalog + mode/model_preference mappings to the current live set. Regression guard: perplexity-web.test.ts. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix(providers): update Gemini Web cookies and models (#6095) Refresh Gemini Web cookie handling + model catalog. Regression guard: gemini-web.test.ts. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix(models): normalize GLM-5.2 provider context (#6091) Hosted GLM-5.2 provider aliases now respect their declared context caps instead of inheriting the native 1M; native/bare + verified OpenCode/ZenMux routes stay at 1M. Regression guards added. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix(combo): prefer known context capacity over unknown (#6088) When a combo filters a target for exceeding a known context limit, prefer remaining known-compatible targets over unknown-metadata ones. Regression guard: combo-context-window-filter.test.ts. Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com> * fix: keep Claude tool results adjacent (#6035) Reattach OpenAI tool_result adjacent to tool_use before Claude send (#6026). Integrated into release/v3.8.44. * fix(security): persist IP filter config + enforce it in the authz pipeline (#6131) (#6132) Integrated into release/v3.8.44 — IP filter persistence + authz-pipeline enforcement (closes #6131). HARD-neutro: validate-release-green on the merge shows the same 3 pre-existing base-reds as the release baseline (test-masking cycle-wide, unit red-herring, integration batch-E2E env); #6131's own tests + ip-filter/pipeline suites all green. * fix(codex): use access_token.exp instead of id_token.exp for import expiresAt (#6075) (#6084) Prefer access_token.exp over id_token.exp for Codex auth import (#6075). Integrated into release/v3.8.44. * fix(compression): send patch-only to PUT /api/settings/compression in CompressionHub (#6039) (#6077) Send patch-only to PUT /api/settings/compression in CompressionHub (#6039). Integrated into release/v3.8.44. * fix: reqId ReferenceError in safety-net redirect, dead code, filename typo (#6097) Fix reqId ReferenceError in safety-net combo redirect + dead-code + DESING→DESIGN rename. Integrated into release/v3.8.44. * fix(combo): expand fingerprint-based providers into per-fingerprint combo targets (#6082) Expand fingerprint-based providers into per-fingerprint combo targets. Integrated into release/v3.8.44. * fix(auth): persist quota preflight account lockouts (#6090) Persist quota preflight account lockouts until reset window. Integrated into release/v3.8.44. * fix(combos): expand OpenCode/MiMo fingerprint accounts in combo builder (#6087) (#6092) Expand OpenCode/MiMo fingerprint accounts in combo builder (#6087). Integrated into release/v3.8.44. * chore(quality): rebaseline v3.8.44 release-green drift (eslint/cognitive/cyclomatic/file-size) Measured on release tip32e4c906eduring the #6131/#5975 release-green pass: eslintWarnings 4256->4270 (+14), cognitiveComplexity 861->867 (+6), cyclomatic count 2015->2026 (+11), and testFrozen caps for models-catalog-route (1507->1600), perplexity-web (959->999), route-edge-coverage (1234->1241, my #5975 comment +7). Inherited cycle drift (the Quality Ratchet does not run on PR->release fast-gates); compression 'bun not found' is a local-env false and codeql is within baseline, so neither is rebaselined. No production code touched. * fix(accountFallback): persist per-account 429 cascade + classify 'Monthly usage limit. Resets in N days.' (#6061) Persist per-account 429 cascade + classify 'Monthly usage limit. Resets in N days'. Integrated into release/v3.8.44. * feat(build): backend-only fast build (skip the dashboard frontend) (#6119) Backend-only fast build (skip dashboard frontend). Integrated into release/v3.8.44. * fix(provider-limits): clear transient rate-limit state when quota recovers (#6128) Clear transient rate-limit state when quota recovers. Integrated into release/v3.8.44. * docs: Normalize mixed-language documentation content (#6105) Normalize mixed-language documentation to English. Integrated into release/v3.8.44. * chore docs * i18n(zh-CN): translate CHANGELOG entries and section headings (#6043) Adopt zh-CN as a translated locale: translate CHANGELOG + supporting docs. Integrated into release/v3.8.44. * chore(quality): rebaseline residual eslint + file-size drift (v3.8.44) Residual drift on release tip716041223(moving target): eslintWarnings 4270->4279 (+9 as the branch advanced past the prior rebaseline) and testFrozen/frozen file-size caps for providerLimits.ts (955->982), accountFallback.ts (1790->1864) and sse-auth.test.ts (1553->1600). All inherited from parallel-session merges (e.g. #6128); the two production god-files ideally warrant decomposition rather than a bump (tracked as debt). No production code touched. * fix(repo): remove Windows case-conflicting DESIGN duplicate (#6140) Remove stale root DESIGN.md (Windows case-conflict with design.md). Integrated into release/v3.8.44. * fix(provider-limits): close TOCTOU race in quota recovery clear (I2) (#6139) Close TOCTOU race in quota recovery clear via CAS primitive (I2 from #6128). Integrated into release/v3.8.44. * fix(glm): suppress </think> close marker leak in GLM Anthropic transport (#6133) Suppress </think> close-marker leak in GLM Anthropic transport. Integrated into release/v3.8.44. * fix(cli): give setup-claude a fallback profile generator like setup-codex (#6138) Give setup-claude a fallback profile generator like setup-codex. Integrated into release/v3.8.44. * fix(onboarding): route provider-details link by node id, not provider slug (#6145) (#6145) Route onboarding provider-details link by node id (#6145). Integrated into release/v3.8.44. * fix(translator): strip Responses-only truncation field before Chat Completions forwarding (#6109) Strip Responses-only truncation field before Chat Completions forwarding (#2311). Integrated into release/v3.8.44. * fix(mitm): guard against concurrent MITM server starts (#6107) Guard against concurrent MITM server starts (#2316). Integrated into release/v3.8.44. * feat(models): add claude-sonnet-5 to Antigravity catalog (#6103) Add claude-sonnet-5 to Antigravity catalog. Integrated into release/v3.8.44. * fix(providers): strip thinking param for minimax-m2.7 on NVIDIA NIM (#6102) Strip unsupported thinking param for minimax-m2.7 on NVIDIA NIM. Integrated into release/v3.8.44. * feat(providers): add Kenari OpenAI-compatible gateway (#6104) Add Kenari OpenAI-compatible gateway (BYOK). Integrated into release/v3.8.44. * feat(sse): per-request Auto-Combo controls (X-OmniRoute-Mode / X-OmniRoute-Budget) — closes #6023 #6024 #6025 (#6057) Per-request Auto-Combo controls (X-OmniRoute-Mode / X-OmniRoute-Budget). Integrated into release/v3.8.44. * feat(resilience): throttle concurrent upstream quota fetches — closes #6009 (#6058) Throttle concurrent upstream quota fetches (#6009). Integrated into release/v3.8.44. * fix(oauth): graceful 400 for keychain-import-only providers (zed) (#6041) (#6054) Graceful 400 for keychain-import-only providers on OAuth route (zed, #6041). Integrated into release/v3.8.44. * fix(dashboard): resolve broken Card import breaking next build (base-red from #6061) (#6155) * fix(dashboard): resolve broken Card import breaking next build (base-red from #6061) CoolingConnectionsPanel imported `Card` from `@/components/ui/card`, a path that does not exist in this repo (there is no shadcn-style `src/components/ui/`). The PR->release fast-gates do not run `next build`, so the broken import slipped in and `next build` failed with: Module not found: Can't resolve '@/components/ui/card' Fix: the <Card> here was only a styled container, so replace it with a <div> carrying the equivalent Tailwind classes (border/bg/padding + rounded-card shadow-sm). Also normalize the file from CRLF to LF (it shipped with CRLF). Adds a vitest/jsdom regression test (tests/unit/ui/CoolingConnectionsPanel.test.tsx) that fails-without-fix (Vite: 'Failed to resolve import @/components/ui/card') and passes with it, plus renders/empty-state coverage. Rule #18. * fix(dashboard): stop client CoolingConnectionsPanel dragging server DB barrel into browser bundle Second base-red from #6061, surfaced once the broken Card import was fixed: ./node_modules/ioredis/built/connectors/StandaloneConnector.js Module not found: Can't resolve 'net' Import trace: ioredis <- rateLimiter.ts <- apiKeys.ts <- @/lib/localDb <- CoolingConnectionsPanel.tsx (a "use client" component) The client panel imported `formatResetCountdown` from `@/lib/localDb` — the server-side DB re-export barrel — which transitively pulls better-sqlite3/ioredis (node:net) into the browser bundle. That violates the CLAUDE.md rule 'never barrel-import from localDb'. `formatResetCountdown` is a pure date-formatting function, so move its implementation to the client-safe `@/shared/utils/formatting` (alongside formatTime/formatDuration) and re-export it from db/providers/rateLimit.ts for the existing server callers + barrel. The panel now imports it directly from the shared util — no server code in the client bundle. Tests (Rule #18): - tests/unit/format-reset-countdown.test.ts (node:test, blocking test:unit) — pure-function coverage: null/past/invalid, s, m+s, h+m, ISO string. - tests/unit/ui/CoolingConnectionsPanel.test.tsx mock updated to the new module. * fix(release): v3.8.44 Phase-0 pre-flight — base-red sweep + ratchet absorption - fix(models): stop resolveProviderAlias at registered provider ids so oc/ reaches the no-auth opencode provider again (#2901 contract, regressed by #5918's transitive chain; transitivity kept across alias-only hops) - fix(auggie): handle async EPIPE 'error' events on child stdin so a fast-exiting CLI surfaces a sanitized error instead of crashing (both spawn sites); deflakes auggie-executor tests - test: align provider family count 166->167 (Kenari #6104), regenerate translate-path golden on Linux (+kenari), opencode quota scope provider->connection (#6061) - quality(test-masking): add _deletedWithReplacement allowlist support to check-test-masking.mjs (deletion exempt ONLY when the declared replacement test exists in HEAD; 5 new gate unit tests) + reduction allowlist entries for the verified #5958/#6088/#5816 migrations + targetExhaustion-> combo-target-exhaustion replacement (#5976, 21 cases/52 asserts vs 13/37) - quality(file-size): absorb v3.8.44 cycle drift (oauth route 960, providerLimits 998, chat 1662, auth 2426) with justification; #6158 will restore the oauth-route freeze - changelog: bullets for the above + the #6155 cooling-panel build fix * chore(release): v3.8.44 — 2026-07-04 Release reconciliation + close (generate-release Phases 0a/1): - CHANGELOG [3.8.44]: 21 PR refs added to existing bullets, 62 new bullets (incl. restoration of ~10 bullets erased by the stale-branch merge in1f6ec5bc8), 3 Maintenance rollups, #6061/#6130 credit fixes, 🙌 Contributors table (35 external contributors) — coverage 144/153 cycle commits by #ref - 42 docs/i18n CHANGELOG mirrors synced (EN content; i18n workflow translates) - README: What's New refreshed for v3.8.44 highlights - build scope: exclude electron/node_modules + electron/dist-electron + .build from tsconfig (local build-output leak poisoned next build with 8GB OOM — same class as the 2026-06-25 incident; scope 14765→5207, gate green) - quality: cyclomatic baseline 2026→2028 (+2 inherited end-of-cycle drift; verified the release-captain code fixes add 0 new violations) * fix(release): v3.8.44 one-pass release-PR CI sweep - fix(dashboard): /dashboard/system/proxy 500'd on EVERY render — #5918 put useProxyBatchOperations(load) before the const load declaration (TDZ ReferenceError, digest 539380095). Hook block moved after load; SSR renderToString regression test added (the exact crash mode). - fix(server): TRACE/TRACK/CONNECT crashed Next's middleware adapter (undici cannot represent them) into a raw 500 on every route — the raw HTTP method guard now answers 405 + Allow up-front (dast-smoke Schemathesis finding on /api/keys/{id}/devices); guard test added. - fix(api): restore Zod validation on the provider-scoped chat route via a .passthrough() schema preserving #5907's relaxed semantics (t06 gate). - docs(openapi): /api/keys/{id}/devices 401 now refs the management error envelope (Schemathesis schema-conformance). - quality: rebaseline i18nUiCoverage 77.5->76.8 (+~1352 new en.json UI keys from the cycle await the async translation workflow; v3.8.39 precedent). - CodeQL: dismissed 2 incomplete-url-substring FPs on unit-test asserts (v3.8.35 precedent) with Hard Rule #14 justifications. - changelog: bullets for the above + 42 i18n mirrors re-synced * fix(release): round-2 CI findings — LocaleAutoDetect refresh gating + ratchet tighten - fix(i18n): LocaleAutoDetect (#5979) refreshed the router on EVERY cookie-less first visit, even when the detected locale matched the server-rendered <html lang> — re-navigating mid-interaction (flaky e2e 'execution context destroyed' + visible flash for new visitors). Refresh now only fires when the locale actually differs; regression test added. - quality: tighten openapiCoverage.pct 36.9->39.3 (require-tighten gate on the release PR; value measured by the CI Quality Ratchet on00c55afcb) - quality(file-size): shrink the ProxyRegistryManager TDZ note to fit the 1117-line freeze (prettier reflow added a line at commit time) - changelog bullet + 42 i18n mirrors re-synced * test(release): collect the #6082 fingerprint-expansion ghost test check:test-discovery (Lint job, layered behind the round-1 t06 fix) flagged tests/e2e/fingerprint-expansion.test.ts as a NEW orphan — it is a node:test server-boot test that no runner collected, so it had never run. Moved to tests/integration/ (the collector for this shape), fixed the helper import, and verified it actually passes (3/3 on first-ever run). CHANGELOG ref updated. --------- Co-authored-by: Chirag Singhal <76880977+chirag127@users.noreply.github.com> Co-authored-by: Hamsa_M <116961508+hamsa0x7@users.noreply.github.com> Co-authored-by: Chewji <126886556+Chewji9875@users.noreply.github.com> Co-authored-by: Vittor Guilherme Borges de Oliveira <vittoroliveira.dev@gmail.com> Co-authored-by: nickwizard <35692452+nickwizard@users.noreply.github.com> Co-authored-by: Ankit <177378174+anki1kr@users.noreply.github.com> Co-authored-by: Fadhil Yusuf <33994304+yusufrahadika@users.noreply.github.com> Co-authored-by: Giorgos Giakoumettis <giorgos@yiakoumettis.gr> Co-authored-by: KooshaPari <42529354+KooshaPari@users.noreply.github.com> Co-authored-by: KooshaPari <KooshaPari@users.noreply.github.com> Co-authored-by: AgentKiller45 <jamalzzj45@gmail.com> Co-authored-by: Nikolay Alafuzov <alafuzov_nn@rusklimat.ru> Co-authored-by: ricatix <d.enistraju155@gmail.com> Co-authored-by: Muhammad Mugni Hadi <mugni@rukita.co> Co-authored-by: Randi <55005611+rdself@users.noreply.github.com> Co-authored-by: backryun <bakryun0718@proton.me> Co-authored-by: Ngô Tấn Tài <tantai@newnol.io.vn> Co-authored-by: dopaemon <polarisdp@gmail.com> Co-authored-by: yicone <yicone@gmail.com> Co-authored-by: CườngNH <j2.cuong@gmail.com> Co-authored-by: DuyPrX <93126969+DuyPrX@users.noreply.github.com> Co-authored-by: ryanngit <74137224+ryanngit@users.noreply.github.com> Co-authored-by: eng2007 <aleksey.semenov@gmail.com> Co-authored-by: chamdanilukman <16629923+chamdanilukman@users.noreply.github.com> Co-authored-by: Umar Javed <114807145+tn5052@users.noreply.github.com> Co-authored-by: JiangZhuo <jiangzhuo@qiniu.com> Co-authored-by: Delynn Assistant <zhen@dkzhen.org> Co-authored-by: whale9820 <87256750+whale9820@users.noreply.github.com> Co-authored-by: whale <admin@dyntech.cc> Co-authored-by: Rigel Ramadhani Waloni <rigel8911@gmail.com> Co-authored-by: zocomputer <help@zocomputer.com> Co-authored-by: aristorinjuang <aristorinjuang@gmail.com> Co-authored-by: anmingwei <anmingwei@dobest.com> Co-authored-by: janeza2 <49841619+janeza2@users.noreply.github.com> Co-authored-by: zmf963 <19422469+zmf963@users.noreply.github.com> Co-authored-by: Markus Hartung <mail@hartmark.se> Co-authored-by: Koosha Pari <kooshapari@gmail.com> Co-authored-by: hartmark <hartmark@users.noreply.github.com> Co-authored-by: ron <devestacion@gmail.com> Co-authored-by: Ansh7473 <Ansh7473@users.noreply.github.com> Co-authored-by: felssxs <felssxs@users.noreply.github.com> Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com> Co-authored-by: Arthur Bodera <abodera@gmail.com> Co-authored-by: Semianchuk Vitalii <fix20152@gmail.com> Co-authored-by: PizzaV <103120356+pizzav-xyz@users.noreply.github.com> Co-authored-by: NOXX - Commiter <artur1992123@mail.ru> Co-authored-by: Milan Soni <123074437+Iammilansoni@users.noreply.github.com> Co-authored-by: Devin <studyzy@gmail.com> Co-authored-by: Raxxoor <manker_lol@hotmail.com> Co-authored-by: derhornspieler <15236687+derhornspieler@users.noreply.github.com>
89 KiB
89 KiB
🚀 OmniRoute — The Free AI Gateway
Never stop coding. Connect every AI tool to 237 providers — 90+ free — through one endpoint.
Plug Claude Code, Codex, Cursor, Cline, Copilot & Antigravity into FREE Claude / GPT / Gemini. Auto-fallback.
RTK + Caveman compression saves 15–95% tokens. Never hit limits.
~1.6B documented free tokens/month — up to ~2.1B in your first month with signup credits — aggregated across the free tiers, plus a long tail of permanently-free, no-cap providers, and the compression above stretches every one further. (how we count →)
⭐ Star the repo if OMNIROUTE helped you save money and make your work easier.
💬 Join the community
Questions, provider tips, roadmap & support → Discord · Telegram · WhatsApp 🌍 Global / 🇧🇷 Brasil
🧩 Available
🚀 Quick Start • 🎯 Combos • 🌐 Providers • 🔌 CLI & MCP • 🗜️ Compression • 🌍 Website
💥 The Promise • 🤔 Why • 🏆 What Sets Apart • 🤖 Compatible CLIs • 🖥️ Where It Runs • 🔒 Private • 🎬 In Action • 📚 Explore More • 📧 Support
🌐 In 42+ languages
| 🇺🇸 | 🇧🇷 | 🇵🇹 | 🇪🇸 | 🇫🇷 | 🇮🇹 | 🇩🇪 | 🇳🇱 | 🇷🇺 | 🇺🇦 | 🇵🇱 | 🇨🇿 | 🇸🇰 | 🇷🇴 | 🇭🇺 |
| 🇧🇬 | 🇩🇰 | 🇫🇮 | 🇳🇴 | 🇸🇪 | 🇨🇳 | 🇹🇼 | 🇯🇵 | 🇰🇷 | 🇹🇭 | 🇻🇳 | 🇮🇩 | 🇲🇾 | 🇵🇭 | |
| 🇮🇳 | 🇮🇳 | 🇮🇳 | 🇮🇳 | 🇮🇳 | 🇮🇳 | 🇧🇩 | 🇵🇰 | 🇮🇷 | 🇸🇦 | 🇮🇱 | 🇹🇷 | 🇦🇿 | 🇹🇿 |
💰 ~1.6B Free Tokens / Month
Stacking free tiers by hand is painful — dozens of SDKs, dozens of rate limits, and no idea how much you actually have. OmniRoute aggregates the documented free tiers of 40+ provider pools / 500+ models into one honest number and shows it live on the dashboard (
/dashboard/free-tiers).
- ~1.6B free tokens / month (steady) — and up to ~2.1B in your first month with signup credits.
- Pool-deduped, honest — we count each shared free pool once, so the headline isn't inflated by rate-limit ceilings the way multi-billion competitor claims are. (Counting every rate limit 24/7 would read ~10B; we don't publish that.)
- Plus the un-countable — permanently-free, no-token-cap providers (SiliconFlow, Z.AI GLM-Flash, Kilo, OpenCode Zen…) and a $10 OpenRouter top-up that unlocks +24M/mo, both surfaced separately so they never inflate the headline.
- Per-model breakdown, live used / remaining for the current month, and a transparent terms flag per provider.
Preview mockup — a real screenshot lands once the
/dashboard/free-tierspage is validated. Full methodology (pool dedupe, credit tiers, provider terms): docs/reference/FREE_TIERS.md.
💥 The Promise
One endpoint. 237 providers. Never stop building — and let OmniRoute pick the cheapest one that works.
| 🚫 Never hit limits Auto-fallback across 237 providers in milliseconds. Quota out? Next provider takes over — zero downtime. |
💸 Save up to 95% tokens RTK + Caveman stacked compression cuts 15–95% of eligible tokens (~89% avg on tool-heavy sessions). |
🆓 $0 to start 90+ providers with a free tier, 11 free forever (Kiro, Qoder, Pollinations, LongCat…). No card needed. |
| 🔌 Every tool works 24+ coding agents — Claude Code, Codex, Cursor, Cline, Copilot, Antigravity — through one config. |
🧩 One endpoint OpenAI ↔ Claude ↔ Gemini ↔ Responses API translation. Point any tool at /v1 and it just works. |
🛡️ Production-grade Circuit breakers, TLS stealth, MCP (95 tools), A2A, memory, guardrails, evals. 21,000+ tests. |
🤔 Why OmniRoute?
Stop juggling 10 dashboards, dead API keys, and surprise bills.
| ❌ The daily pain | ✅ How OmniRoute fixes it |
|---|---|
| 📉 Subscription quota expires unused every month | Maximize subscriptions — track quota, use every token before reset |
| 🛑 Rate limits stop you mid-coding | 4-tier auto-fallback — Subscription → API → Cheap → Free, in milliseconds |
🔥 Tool outputs (git diff, grep, logs) burn tokens |
RTK + Caveman compression — save 15–95% eligible tokens per request |
| 💸 Expensive APIs ($20–50/mo per provider) | Cost-optimized routing — auto-route to the cheapest viable model |
| 🧰 Each AI tool wants its own setup | One endpoint, every tool, one dashboard |
| 🌍 AI blocked in your country | 3-level proxy + TLS fingerprint stealth — use AI from anywhere |
┌──────────────────────────────────────────────────────────┐
│ Your IDE / CLI (Claude Code, Cursor, Cline…) │
└─────────────────────────┬──────────────────────────────────┘
│ http://localhost:20128/v1
▼
┌──────────────────────────────────────────────────────────┐
│ OmniRoute — Smart Router │
│ RTK + Caveman compression · 17 routing strategies │
│ Circuit breakers · TLS stealth · MCP · A2A · Guardrails │
└─────────────────────────┬──────────────────────────────────┘
┌─────────────┬────┴────────┬─────────────┐
▼ Tier 1 ▼ Tier 2 ▼ Tier 3 ▼ Tier 4
SUBSCRIPTION API KEY CHEAP FREE
Claude Code, DeepSeek, GLM $0.5, Kiro, Qoder,
Codex, Copilot Groq, xAI MiniMax $0.2 Pollinations
quota out? ───▶ budget hit? ─▶ budget hit? ─▶ always on
🎯 Combos — The Flagship
A combo is a chain of models OmniRoute routes across automatically. Quota runs out, a provider fails, or costs spike — the combo silently slides to the next model. This is what makes OmniRoute unbreakable. 🛡️
⚡ Zero-config — just use auto
No combo to create. Set your model to auto (or a variant) and OmniRoute builds a virtual combo from your connected providers, scored live:
| Model ID | What it optimizes for |
|---|---|
auto |
🎯 Balanced default (LKGP — sticks to your last good provider) |
auto/coding |
🧑💻 Quality-first weights for code generation |
auto/fast |
⚡ Lowest latency first |
auto/cheap |
💰 Cheapest per token first |
auto/offline |
🔋 Most quota / rate-limit headroom first |
auto/smart |
🔭 Quality-first + 10% exploration to discover better models |
🔀 Or build your own — 17 routing strategies
All 17 strategies — mix & match per combo step:
| # | Strategy | What it does |
|---|---|---|
| 1 | priority |
First-target ordered list — drain each before the next 🥇 |
| 2 | fill-first |
Fill each target's quota fully before moving on |
| 3 | weighted |
Weighted random by per-target weight |
| 4 | round-robin |
Cycle through targets in order |
| 5 | p2c |
Power-of-two-choices random load balancing |
| 6 | least-used |
Pick the target with the lowest current load |
| 7 | random |
Uniform random pick (deduplicated) |
| 8 | strict-random |
Random without de-duplicating repeats 🎲 |
| 9 | cost-optimized |
Minimize $ per request from live catalog pricing 💸 |
| 10 | headroom |
Pick the target with the most remaining quota |
| 11 | reset-window |
Prefer the target whose quota window resets soonest |
| 12 | reset-aware |
Rank by quota reset time — short windows first 📊 |
| 13 | context-relay |
Hand off context across targets for long conversations 🧠 |
| 14 | context-optimized |
Pick the best fit for the current context size |
| 15 | lkgp |
Last-Known-Good Path — sticky to the last successful target |
| 16 | auto |
9-factor live scoring across every connection 🤖 |
| 17 | fusion |
Fan out to a panel of models + a judge synthesizes one answer 🧬 |
The Auto-Combo engine scores every candidate on 9 factors (health, quota, cost, latency, success rate, freshness…) — see docs/routing/AUTO-COMBO.md.
⚖️ Quota-Share — split one subscription across a team ✨ NEW
Running several keys against the same upstream account (one Codex Pro plan, one Kimi key, one GLM Coding seat)? A burst on one key can burn the whole 5-hour / hourly quota and lock everyone else out. Quota-Share distributes a provider's time-based quota fairly across the keys in a pool — and it's work-conserving, so an idle member's slice is lent out instead of wasted.
| Knob | What it controls |
|---|---|
| ⚖️ Allocation weight | each key's slice of the pool — e.g. 50 / 30 / 20 |
| 📐 Dimensions | track % · requests · tokens · $, per 5h / 7d / per-model window |
| 🚦 Policy | hard (block over share) · soft (deprioritize) · burst (use idle headroom) |
| 🧱 Cap | absolute ceiling per key, independent of mode |
Pool "team-codex" · 1 Codex Pro account · 3 keys · 5-hour window
├─ alice weight 50 ██████████░░░░░░░░░░ ≤ 50% of the shared 5h quota
├─ bob weight 30 ██████░░░░░░░░░░░░░░ ≤ 30%
└─ ci-bot weight 20 ████░░░░░░░░░░░░░░░░ ≤ 20%
Generous mode (<50% pool used) → idle shares are lent out
Strict mode (≥50% pool used) → each key held to its fair share
Enforced in the hot path before the request leaves OmniRoute, with per-(key, model) caps + session stickiness for prompt-cache integrity. 📖 Quota Sharing Engine
🧱 Resilience is built in (3 independent layers)
| Layer | Scope | What it does |
|---|---|---|
| 🔌 Circuit breaker | whole provider | Stops hammering a provider that's failing upstream; auto-probes to recover |
| 💤 Connection cooldown | one account / key | Skips a rate-limited key while other keys keep serving |
| 🎯 Model lockout | provider + model | Quarantines just one quota-limited model, not the whole connection |
Combo: "always-on" Strategy: priority
1. cc/claude-opus-4-7 ← subscription (use it fully)
2. cx/gpt-5.5 ← second subscription
3. glm/glm-5.1 ← cheap backup ($0.5/1M)
4. kr/claude-sonnet-4.5 ← FREE, unlimited (never fails)
Result: 4 layers of fallback = zero downtime
📖 Auto-Combo Engine · Resilience Guide
🏆 What Sets OmniRoute Apart
| Feature | OmniRoute | Other routers |
|---|---|---|
| 🌐 Providers | 237 | 20–100 |
| 🆓 Free providers | 90+ (11 free forever) | 1–5 |
| 🔀 Routing strategies | 17 (priority, weighted, cost-optimized, context-relay, fusion…) | 1–3 |
| 🗜️ Token compression | RTK + Caveman stacked (15–95%) | None / 20–40% |
| 🧰 Built-in MCP server | 95 tools, 3 transports, 30 scopes | Rare |
| 🤝 A2A agent protocol | 6 skills, JSON-RPC 2.0 | None |
| 🧠 Memory (FTS5 + vector) | Yes | Rare |
| 🛡️ Guardrails (PII, injection, vision) | Yes | Rare |
| ☁️ Cloud agents | Codex, Cursor, Devin, Jules | None |
| 🥷 TLS fingerprint stealth | JA3/JA4 via wreq-js | None |
| 🖥️ Multi-platform | Web · Desktop · Termux · PWA | Web only |
| 🌍 i18n | 42 locales | 0–4 |
📊 Detailed comparison vs LiteLLM, OpenRouter & Portkey → docs/comparison/OMNIROUTE_VS_ALTERNATIVES.md
✨ What's New
Recent highlights from v3.8.20 → v3.8.44. Full history in
CHANGELOG.md.
- 🗜️ Compression hardening — a default-on inflation guard (discard the stacked result and send the verbatim original whenever compression would grow the prompt), completed Caveman rule packs for German / French / Japanese (dedup + ultra) plus a new Chinese (文言 / wényán) input pack with zh-vs-ja auto-detection, and RTK filters for Gradle & .NET (
dotnet) build output. → Compression - 💸 Honest flat-rate cost — subscription / coding-plan providers (ChatGPT Web, grok-web, the Minimax / Kimi / GLM / Alibaba Coding plans, Xiaomi MiMo…) now read $0 in cost analytics instead of an inflated per-token estimate, while budget / quota / routing keep estimating unchanged. → API Reference
- ⚖️ Quota-Share routing — a dedicated combo strategy that spreads load across accounts by available quota: Deficit-Round-Robin scheduling, per-connection
max_concurrentwith cooldown-wait queueing, multi-window usage buckets (5h / 7d / per-model), per-(key, model) caps, session stickiness for prompt-cache integrity, and proactive saturation from upstream token-usage headers. → Resilience Guide - 🤖 One-command CLI/agent setup — a dedicated
setup-*command configures each coding tool to route through OmniRoute (Claude Code, Codex, Cline, Continue, Cursor, Roo Code, Kilo Code, Crush, Goose, Qwen Code, Aider, OpenCode);omniroute launch/omniroute launch-codexare zero-config launchers. → CLI Integrations - 🛰️ Remote mode — drive a remote OmniRoute from any machine with scoped access tokens (
omniroute connect/omniroute contexts/omniroute tokens), plus anomniroute login antigravityhelper that runs Google "native/desktop" OAuth on your own machine and pastes a credential blob into a remote/VPS install (where the loopback redirect is unreachable). → Remote Mode - 🧭 Smarter auto-routing — OpenRouter-style
auto/<category>:<tier>combos (e.g.auto/coding:fast,auto/reasoning:pro), a Fusion strategy (fan out to a panel of models in parallel, then synthesize via a judge), task-aware routing (best-fit connection per task type), per-requestX-Route-Modeloverride, live Arena-ELO + models.dev model intelligence, per-step account allowlists, provider-wildcard combo steps, nested combo-ref execution, sticky weighted selection,web_search-aware routing, and per-request Auto-Combo controls (X-OmniRoute-Modemode-preset override +X-OmniRoute-Budgethard USD cost ceiling, scoped to a single request). → Auto-Combo - 🗜️ Pluggable compression — an async pipeline of 10 composable engines with Compression Studios, an LLMLingua-2 ONNX engine and a heuristic/SLM two-tier Ultra, RTK, delegated Anthropic Context Editing, Output Styles (output-axis steering: terse-prose / less-code / terse-CJK), an adaptive context-budget dial (escalate only as far as needed to fit the context window), per-request
x-omniroute-compressioncontrol, an opt-in offline eval harness, one-click Headroom proxy lifecycle management from the dashboard (Docker sidecar supported), a synthetic compression playground (Play lanes + A/B Compare with USD-capped fidelity verdicts), an opt-in per-step fidelity gate that rejects a lossy engine before it degrades the prompt, a best-of-N candidate encoder (GCF vs TOON — keep whichever is shorter, with an A/B bytes/token table in the studio), CCR ranged/grep/stats retrieval (pull an exact byte/line slice or summary of a stored block instead of re-expanding it), a unified panel with named profiles + an active-profile selector, an opt-in per-engine pipeline circuit-breaker, an opt-in LLM-tier engine (a model pass for higher-ratio semantic compression), a read-lifecycle engine that collapses superseded file reads, usage-observed prefix freeze, a graduated CCR retrieval-feedback ramp, apreserveSystemPromptmode enum, and a drag-reorder pipeline editor in the studio. → Compression - 🕵️ Transparent MITM decrypt (TPROXY) — capture & translate traffic from CLIs that ignore proxy env vars, with a per-SNI certificate authority and a trust-store installer. → MITM/TPROXY
- 💸 Cost telemetry everywhere —
X-OmniRoute-*cost/usage headers on every endpoint (including media), a non-token cost engine, a cache-HITX-OmniRoute-Cost-Savedheader, and per-key USD spend quotas. → API Reference - 🧠 Memory you control — opt-in int8 vector quantization (Qdrant + sqlite-vec), opt-in typed memory decay (aged low-value memories fade on a per-type schedule), memory off by default, and a per-request
x-omniroute-no-memoryheader. → Memory - 🛡️ Security — a prompt-injection guard across every LLM route (backed by a red-team suite), plus a free DuckDuckGo last-resort web search. → Guardrails
- 🖼️ New endpoints —
/v1/ocr(Mistral OCR) and/v1/audio/translations(Whisper-style audio translation) round out the media API surface. → API Reference - 🌍 Deployment & ops — reverse-proxy
basePathdeployment (OMNIROUTE_BASE_PATH, e.g. serving OmniRoute under/omniroute/), browser-language auto-detect on first visit, and per-API-key device/connection tracking (IP+UA fingerprint, masked, in-memory only). → Environment - 🤝 More providers & agents — Cursor Cloud Agent (a 4th cloud agent), CodeBuddy CN (
copilot.tencent.com), a Google Flow video-generation provider, new gateways DGrid and Pioneer AI (Fastino Labs), inbound xAI Grok translators plus Grok Build (xAI) with an OAuth import-token flow, GPT-4 / GPT-4o-mini on the GitHub Copilot provider, multi-model Factory Droid, ZenMux Free (session-cookie free tier), Alibaba DashScope text-to-video (wan2.7-t2v), a refreshed 237-provider catalog (OrcaRouter, Wafer AI, OpenAdapter, dit.ai, TokenRouter, …), Vertex AI media generation (speech/transcription/music/video), a first-class Ollama local-provider card, the SenseNova free Token Plan (chat + text-to-image), one-click account import from CLIProxyAPI (~/.cli-proxy-api/), Claude Sonnet 5 wired end-to-end, a new provider wave (Kenari, SumoPod, X5Lab, Charm Hyper, Nube.sh, b.ai, Qiniu, ModelScope, Augment/Auggie CLI, ClinePass, NVIDIA NIM image generation), and Codex account import from a raw ChatGPT access token. → Providers - ⚡ Local performance & infra — a one-click local Redis launcher (
omniroute redis up, plus a dashboard Redis panel), one-click Cloudflare Workers and Deno Deploy relay deployers wired into the proxy pool, a relay-backend selector (OMNIROUTE_RELAY_BACKEND=ts|bifrost|auto) so/v1/relaystays the stable surface while choosing the fastest backend internally, Bifrost (Go AI-gateway) and Mux (agent-orchestration daemon) promoted to first-class embedded/supervised services alongside 9Router/CLIProxyAPI, and Webshare added as a paid fourth source in the free-proxy provider framework. → Embedded Services
🤖 Compatible CLIs & Coding Agents
One config —
http://localhost:20128/v1— and every AI IDE or CLI runs on free & low-cost models.
 Claude Code |
 Codex CLI |
 Cursor |
 Copilot |
 Continue |
|
 OpenCode |
 Kilo Code |
 Droid |
 OpenClaw |
 Kiro |
 Command |
＋ also works with · Cline · Antigravity · Windsurf · AMP · Hermes · Qwen CLI · Roo · Continue · any OpenAI-compatible tool
📖 Per-tool setup for all 24+ tools → docs/reference/CLI-TOOLS.md · 🧩 OpenCode plugin → @omniroute/opencode-provider
🌐 237 AI Providers — 90+ Free
The most complete catalog of any open-source router: 237 providers, 90+ with a free tier, 11 free forever.
🏢 Every major lab — through one endpoint
 OpenAI |
 Anthropic |
 Gemini |
 xAI Grok |
 DeepSeek |
 Mistral |
 Qwen |
 Meta Llama |
 Groq |
 NVIDIA |
 MiniMax |
 Cohere |
 Perplexity |
 HuggingFace |
 Together |
 Fireworks |
 Cloudflare |
 Baidu |
…and 220+ more — every icon resolves live from the dashboard's provider catalog. 📖 Provider Reference
🆓 Free Forever — $0, no card
 AgentRouter GPT-5, Claude, Gemini $100 free credits |
 Qoder AI Kimi-K2, DeepSeek-R1 Unlimited FREE |
 Pollinations GPT-5, Claude, Llama 4 No key needed |
 LongCat LongCat-2.0 10M tokens one-time (KYC) 🔑 |
Cloudflare AI 50+ models 10K neurons/day |
NVIDIA NIM 129 models ~40 RPM free |
 Cerebras Qwen3 235B 1M tokens/day |
📖 Full machine-readable catalog → docs/reference/PROVIDER_REFERENCE.md
🖥️ Where OmniRoute Runs — Anywhere
Same app, your machine, your rules. From a global npm install to your phone via Termux.
| Platform | Install | Highlights |
|---|---|---|
| 📦 npm (global) | npm install -g omniroute |
One command, any OS |
| 🐳 Docker | docker run … diegosouzapw/omniroute |
Multi-arch AMD64 + ARM64 |
| 🖥️ Desktop (Electron) | npm run electron:build |
Native window + system tray — Windows / macOS / Linux |
| 💪 ARM | native arm64 |
Raspberry Pi, ARM servers, Apple Silicon |
| 📱 Android (Termux) | pkg install nodejs && npx -y omniroute |
Runs on your phone, 24/7, no root |
| 📲 PWA | "Add to Home Screen" | Fullscreen, offline, installable from browser |
| 🧩 OpenCode plugin | @omniroute/opencode-provider |
Native OpenCode integration |
| 🛠️ From source | npm install && npm run dev |
Hack on it, contribute |
📖 Docker Guide · Desktop · Termux · PWA · OpenCode
🔒 Private & Local-First
Your keys, your machine, your data. OmniRoute is a local proxy — it never phones home.
- 🏠 Runs 100% on your hardware — npm, Docker, desktop, or your phone. No OmniRoute cloud sits in the request path.
- 🔐 Credentials encrypted at rest — API keys & OAuth tokens sealed with AES-256-GCM.
- 🚫 Zero telemetry by default — your prompts go only to the providers you choose, nowhere else.
- 🛡️ Hardened gateway — API-key scoping, IP filtering, rate limits, prompt-injection guard, loopback-only process routes.
- 📜 MIT licensed & fully open-source — audit every line, self-host forever.
📖 Authorization · Guardrails · Compliance
🔌 Full CLI + A2A & MCP
OmniRoute isn't just a server — it's a full command-line cockpit with 80+ commands, plus open agent protocols so an AI agent can drive OmniRoute by itself.
⌨️ A real CLI (not just start)
omniroute # serve gateway + dashboard (port 20128)
omniroute chat # interactive TUI chat client (slash: /model /combo /skill /memory)
omniroute setup # guided first-run wizard
omniroute doctor # diagnose providers, ports, native deps
🛰️ Remote mode — run the CLI here, OmniRoute on a VPS
OmniRoute on a server? Drive it from your laptop with the same CLI. Log in once with a scoped access token; every command then targets the remote.
omniroute connect 192.168.0.15 # password → scoped token, saved as a context
omniroute models list # ← runs against the REMOTE server
omniroute configure codex # ← picks a remote model, writes a local Codex profile
omniroute tokens create --name ci --scope read # mint narrower tokens for other machines
omniroute contexts use default # ← switch back to the local server
Tokens are scoped read / write / admin; process-spawning routes stay loopback-only.
📖 Remote Mode
providers · oauth · keys · combo · nodes · models · cache · compression · cost · usage · quota · health · resilience · telemetry · logs · audit · mcp · a2a · cloud · memory · skills · eval · tunnel · backup · sync · webhooks · policy · pricing · translator · simulate …
🤝 Connect an agent — and it controls OmniRoute itself
Expose OmniRoute over MCP or A2A and any capable agent gets the keys to the whole gateway — routing, providers, combos, cache, compression, memory — autonomously.
| Protocol | Endpoint | Use it for |
|---|---|---|
| 🧰 MCP (stdio) | omniroute --mcp |
Plug into Claude Desktop, Cursor, any MCP client |
| 🌊 MCP (HTTP) | http://localhost:20128/api/mcp/stream |
Remote MCP — 95 tools, 30 scopes, full audit trail |
| 📡 MCP (SSE) | http://localhost:20128/api/mcp/sse |
Streaming MCP transport |
| 🤝 A2A | http://localhost:20128/.well-known/agent.json |
Agent-to-agent, JSON-RPC 2.0 + SSE, 6 skills |
# Give Claude Code the full OmniRoute toolset over MCP:
claude mcp add-server omniroute --type http --url http://localhost:20128/api/mcp/stream
📖 MCP Server · A2A Server · Agent Protocols
🗜️ Save 15–95% Tokens — Automatically
Why use many tokens when few tokens do the trick? Every request passes through OmniRoute's compression pipeline transparently — no client changes. It's now a stack of 10 composable engines that run in order and mix & match per routing combo — building on ideas from RTK, Caveman (⭐ 78K+), LLMLingua-2, and Troglodita (PT-BR).
🧱 The 10-engine stack
Engines run in pipeline order; each is independently toggleable and configurable per combo:
| # | Engine | What it does |
|---|---|---|
| 1 | Session-Dedup | Drops content repeated across turns (content-addressed, cross-turn) |
| 2 | CCR | Archives large blocks behind retrieve markers, fetched on demand |
| 3 | RTK | Smart tool-result filtering, dedup & truncation (command-aware) |
| 4 | Headroom | Lossless tabular compaction of homogeneous JSON arrays (~30%+) |
| 5 | Relevance | Extractive sentence scoring against the last user query |
| 6 | Caveman | Rule-based prose compression (~65–75% on output) |
| 7 | LLMLingua-2 | ML semantic pruning via MobileBERT ONNX — code-safe, async |
| 8 | Lite | Whitespace + image-URL trimming (latency-light baseline) |
| 9 | Aggressive | Summarization + progressive aging of old turns |
| 10 | Ultra | Heuristic token pruning with an optional small-model (SLM) tier |
Code blocks, URLs and structured data are always preserved byte-perfect. One-click presets combine the engines:
| Mode | Savings | Best for |
|---|---|---|
| 🪶 Lite | ~15% | Always-on safe default |
| 🪨 Standard (Caveman) | ~30% | Daily coding |
| ⚡ Aggressive | ~50% | Long tool-heavy sessions |
| 🔥 Ultra | ~75% | Maximum savings |
| 🧰 RTK | 60–90% | Shell/test/build/git output |
| 🔗 Stacked (RTK → Caveman) | 78–95% | Mixed prompts + tool logs |
Real example — Standard mode:
Before (69 tokens): "The reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. When you pass an inline object as a prop, React's shallow comparison sees it as a different object every time, which triggers a re-render. I would recommend using useMemo to memoize the object."
After (19 tokens): "New object ref each render. Inline object prop = new ref = re-render. Wrap in useMemo."
Same answer. 72% fewer tokens. Zero accuracy loss. ✅
PT-BR example — Troglodita mode:
Antes (42 tokens): "O problema é que o componente está re-renderizando porque uma nova referência de objeto está sendo criada em cada ciclo de renderização. Eu recomendaria usar useMemo."
Depois (12 tokens): "Re-render: ref nova cada ciclo (objeto inline recriado). Usar
useMemo."Mesma resposta. ~70% menos tokens. Precisão técnica intacta. ✅
📖 How it works — pipeline, architecture & savings math
Client (10,000 tok) ──▶ OmniRoute Compression (10 engines) ──▶ Provider (~1,080 tok, up to 95% saved)
Default stacked combo runs RTK → Caveman. When both act on the same tool/context payload, savings compound:
combined = 1 − (1 − RTK) × (1 − Caveman_input)
average = 1 − (1 − 0.80) × (1 − 0.46) = 89.2%
range = 78.4 – 94.6%
Code blocks, URLs, JSON and structured data are always protected by the preservation engine.
🎚️ Beyond the engines — output styles, the adaptive dial & per-request control
The 10 engines above shrink what goes in. Three more layers shape how, when, and what comes out:
- 🪄 Output Styles (output-axis steering) — inject deterministic, cache-safe response-shaping instructions; combinable, each at
lite/full/ultraintensity. Adding a style is a one-line registry entry:- Terse prose — drop filler / articles / hedging; keep technical substance exact.
- Less code — "lazy senior dev" YAGNI: smallest working change, no unrequested scaffolding.
- Terse CJK (文言) — classical-Chinese ultra-terse style (locale-gated to
zh).
- 🎯 Adaptive context-budget (the dial) — instead of one on/off token threshold, escalate the cheapest, most-lossless engines only as far as needed to fit the model's context window. Policy:
reserve-output(default, model-aware) ·percentage·absolute. Mode:floor(guarantee fit) ·replace-autotrigger(your explicit choice wins) ·off(legacy threshold). - 🎛️ Where compression is decided (precedence, high → low) — per-request
x-omniroute-compressionheader › routing-combo override › active named profile › adaptive / auto-trigger › panel default › off. The applied plan echoes back in theX-OmniRoute-Compression: <mode>; source=<source>response header.
Auto-trigger by token threshold, flip on the adaptive dial, pin a named profile, set a one-off per request, or assign a pipeline per routing combo — whichever fits the workload. An opt-in offline eval harness (npm run eval:compression) scores fidelity vs. savings on a pinned corpus before you promote a change.
📖 COMPRESSION_GUIDE.md · RTK_COMPRESSION.md · COMPRESSION_ENGINES.md
⚡ Quick Start
1) Install & run
npm install -g omniroute
omniroute
Dashboard at http://localhost:20128 · API at http://localhost:20128/v1.
2) Connect a FREE provider (no signup)
Dashboard → Providers → connect Kiro AI (free Claude, ~50 credits/month per account) or OpenCode Free (no auth) → done.
3) Point your coding tool
Base URL: http://localhost:20128/v1
API Key: [copy from Dashboard → Endpoints]
Model: auto (zero-config smart routing — or any provider/model)
4) Verify it's working
curl http://localhost:20128/v1/models -H "Authorization: Bearer YOUR_KEY"
You should see your connected models listed. 🎉 That's it — start coding, and OmniRoute auto-routes & falls back for you.
If your client cannot send custom headers, OmniRoute also exposes tokenized compatibility aliases:
OpenAI catalog: http://localhost:20128/vscode/YOUR_KEY/
OpenAI models: http://localhost:20128/vscode/YOUR_KEY/models
OpenAI chat: http://localhost:20128/vscode/YOUR_KEY/chat/completions
OpenAI responses: http://localhost:20128/vscode/YOUR_KEY/responses
Ollama chat: http://localhost:20128/vscode/YOUR_KEY/api/chat
Ollama tags: http://localhost:20128/vscode/YOUR_KEY/api/tags
Use these only for clients that cannot attach Authorization: Bearer .... Header auth remains the preferred mode.
📦 More install methods — Docker, source, pnpm, Arch
🐳 Docker
docker run -d --name omniroute --restart unless-stopped --stop-timeout 40 \
-p 20128:20128 -v omniroute-data:/app/data diegosouzapw/omniroute:latest
🛠️ From source
cp .env.example .env && npm install
PORT=20128 npm run dev
📦 pnpm
pnpm add -g omniroute@latest --allow-build=better-sqlite3 --allow-build=@swc/core && omniroute
🐧 Arch Linux (AUR)
yay -S omniroute-bin && systemctl --user enable --now omniroute.service
🔧 Nix (Flake)
# Using Nix flakes
nix develop
npm run dev
# Or using devbox
devbox run npm run dev
📖 Docker Guide — Compose profiles, Caddy HTTPS, Cloudflare tunnels.
🦭 Podman
# 1. Build the image
podman build --target runner-base -t omniroute:base .
# 2. Fix data directory permissions for rootless Podman
mkdir -p data && podman unshare chown 1000:1000 ./data
# 3. Set runtime in .env, then run (see contrib/podman/ for Quadlet)
echo "CONTAINER_HOST=podman" >> .env
podman compose --profile base up -d
📖 Podman Guide — Quadlet setup, podman-compose, Quadlet.
⚡ Faster / leaner install (skip the native build)
The native SQLite engine (better-sqlite3) is an optional dependency, so a global
install never blocks on compiling from source: it uses a prebuilt binary when one matches
your platform/Node, and otherwise falls back transparently to a pure-JS engine
(node:sqlite on Node 22+, else the bundled sql.js WASM) — no build tools required.
To skip the post-install native warm-up entirely (CI, headless, or slow machines):
OMNIROUTE_SKIP_POSTINSTALL=1 npm install -g omniroute # CI=1 also skips it
For the fastest installs prefer pnpm (content-addressed store + hard links — see above).
For a dashboard-free, headless runtime use the Docker base profile (above) or the
Termux guide. The CLI and the web dashboard are served by the
same process on one port, so there is no separate CLI-only package today.
🎬 OmniRoute in Action
 🇧🇷 Português Guia completo |
 🇺🇸 English Complete walkthrough |
 🇷🇺 Русский Полное руководство |
🎬 Made a video about OmniRoute? Open an issue or discussion with the link — we'll feature it here.
📚 Explore More
💰 Pricing at a glance & the $0 Free Stack (11 providers)
| Tier | Example | Cost |
|---|---|---|
| 💳 Subscription | Claude Code Pro / Codex / Copilot | $10–200/mo |
| 🔑 API Key (free tiers) | NVIDIA NIM, Cerebras, Groq | FREE |
| 💰 Cheap | GLM-5 $0.5/1M · MiniMax M2.5 $0.3/1M | pennies |
| 🆓 Free Forever | Kiro, Qoder, Qwen, Pollinations, LongCat | $0 |
The $0 Free Stack — combine into one unbreakable combo:
| Provider | Prefix | Free models | Quota |
|---|---|---|---|
| Kiro | kr/ |
Claude Sonnet 4.5, Haiku 4.5, Opus 4.6 | 50 credits/mo |
| Qoder | if/ |
kimi-k2-thinking, qwen3-coder-plus, deepseek-r1 | ♾️ Unlimited |
| Qwen | qw/ |
qwen3-coder-plus/flash/next | ♾️ Unlimited |
| Pollinations | pol/ |
GPT-5, Claude, Gemini, DeepSeek, Llama 4 | No key needed |
| LongCat | lc/ |
LongCat-2.0 | 10M one-time (KYC) |
| Cloudflare AI | cf/ |
50+ models | 10K neurons/day |
| NVIDIA NIM | nvidia/ |
129 models | ~40 RPM |
| Cerebras | cerebras/ |
Qwen3 235B, GPT-OSS 120B | 1M tok/day |
💡 The dashboard "cost" is a savings tracker, not a bill — OmniRoute never charges you. A "$290 total cost" using free models means $290 saved.
📖 Complete free directory → docs/reference/FREE_TIERS.md — 25+ providers, quotas, base URLs.
🎯 Use Cases — ready-made combo playbooks
$0 forever:
1. kr/claude-sonnet-4.5 (Kiro — ~50 credits/mo per acct)
2. if/kimi-k2-thinking (Qoder — unlimited)
3. pol/gpt-5 (Pollinations — no key)
4. lc/LongCat-2.0 (10M one-time backup, KYC)
Compression: aggressive (~50%) → double your free quota · Cost: $0/mo
24/7 no interruptions: chain 2 subscriptions → cheap → free for 5 layers of fallback.
Blocked region: free providers + global/per-provider proxy → access AI from any country.
Max savings: subscription + cheap backup + ultra compression (~75%) → ~$150–300/mo saved for heavy users.
🌍 Bypass geo-blocks — 3-level proxy + stealth
🇷🇺 🇨🇳 🇮🇷 🇨🇺 🇹🇷 In a blocked region? OmniRoute's 3-level proxy (Global / Per-Provider / Per-Connection) proxies API requests, OAuth flows, connection tests, token refresh & model sync.
- Protocols: HTTP/HTTPS, SOCKS5, authenticated proxies
- 🆓 1proxy marketplace — hundreds of free validated proxies, quality scores, auto-rotation
- Anti-detection — TLS fingerprint spoofing (
wreq-js), CLI fingerprint matching, proxy IP preservation
✨ Full feature list — 30+ capabilities (memory, evals, observability)
Routing: 17 strategies · task-aware smart routing · thinking budget controls · wildcard routing · system prompt injection.
Compatibility: OpenAI ↔ Claude ↔ Gemini ↔ Responses API · auto OAuth refresh (PKCE, 8 providers) · multi-account round-robin · Batch + Files API · live OpenAPI 3.0.
Protocols: MCP (95 tools, 3 transports, 30 scopes) · A2A (JSON-RPC 2.0, SSE, 6 skills) · ACP · cloud agents (Codex, Cursor, Devin, Jules).
Plugins: custom plugin marketplace (system-configured registry URL with SSRF-guarded fetch) · install / enable / disable · Notion + Obsidian knowledge-base integrations (WebDAV file server, vault search, note CRUD).
Embedded services: one-click install & lifecycle management of local sidecar services (CLIProxy, NineRouter).
Quality & Ops: built-in Evals (golden-set: exact/contains/regex/custom) · guardrails (PII, injection, vision) · health dashboard · p50/p95/p99 telemetry · webhooks · compliance audit.
AI Agent Skills: drop-in markdown manifests — point any agent at a skills/*/SKILL.md manifest. 43 skills available.
📖 MCP Server · A2A Server · Resilience Guide · Features Gallery
📖 Setup, env vars & FAQ
| Env var | Default | Purpose |
|---|---|---|
PORT |
20128 |
API + dashboard port |
REQUIRE_API_KEY |
false |
Require API key for all requests |
DATA_DIR |
~/.omniroute |
Database & config storage |
Will I be charged by OmniRoute? No — it's free, open-source software on your machine. You only pay paid providers directly. OmniRoute has no billing system. Are FREE providers really unlimited? Mostly — Qoder, Pollinations, LongCat, and Cloudflare are free with no per-account credit cap. Kiro is free too but capped at ~50 credits/month per account. Stack multiple free providers in a combo and auto-fallback keeps you serving for $0. Will compression hurt quality? No — it only compresses the input; code, URLs, JSON are always protected. Does it work where AI is blocked? Yes — 3-level proxy + 1proxy marketplace reach all 237 providers.
🐛 Troubleshooting
| Problem | Quick fix |
|---|---|
| "Language model did not provide messages" | Provider quota exhausted → use a combo fallback |
| Rate limiting (429) | Add fallback: cc/claude → glm/glm-4.7 → if/kimi-k2-thinking |
| OAuth token expired | Auto-refreshed; if stuck, delete + re-auth in Providers |
unsupported_country_region_territory |
Configure proxy in Settings → Proxy |
| Docker SQLite locks | Use --stop-timeout 40 for clean WAL checkpoint |
| Node runtime errors | Use Node >=22.0.0 <23 or >=24.0.0 <27 |
🐛 Reporting a bug? Run npm run system-info and attach system-info.txt. 📖 docs/guides/TROUBLESHOOTING.md
📸 Dashboard screenshots
| Page | Screenshot | Page | Screenshot |
|---|---|---|---|
| Providers |  |
Combos |  |
| Analytics |  |
Health |  |
| Translator |  |
Settings |  |
| CLI Tools |  |
Usage Logs |  |
📧 Support & Community
💬 Chat with the community — Discord, Telegram & WhatsApp (🌍 / 🇧🇷) links are at the top of this README.
- 🌍 Website: omniroute.online
- 🐙 GitHub: github.com/diegosouzapw/OmniRoute
- 🐛 Issues: report a bug (attach
npm run system-infooutput) - 🤝 Contributing: see CONTRIBUTING.md or pick a
good first issue
🛠️ Tech Stack
- Runtime: Node.js 22.x or 24.x LTS (24 LTS recommended) —
>=22.0.0 <23 || >=24.0.0 <27 - Language: TypeScript 6.0 — 100% TypeScript across
src/andopen-sse/(zeroanyin core modules since v2.0) - Framework: Next.js 16 + React 19 + Tailwind CSS 4
- Database: better-sqlite3 (SQLite) + LowDB (JSON legacy) — domain state, proxy logs, MCP audit, routing decisions, memory, skills
- Schemas: Zod (MCP tool I/O validation, API contracts)
- Protocols: MCP (stdio/HTTP) + A2A v0.3 (JSON-RPC 2.0 + SSE)
- Streaming: Server-Sent Events (SSE) + WebSocket bridge (
/v1/ws) - Auth: OAuth 2.0 (PKCE) + JWT + API Keys + MCP Scoped Authorization
- Testing: Node.js test runner + Vitest (21,000+ test cases across 2,586 files — unit, integration, E2E, security, ecosystem)
- Platforms: Desktop (Electron), Android (Termux), PWA (any browser)
- CI/CD: GitHub Actions (auto npm publish + Docker Hub on release)
- Website: omniroute.online
- Package: npmjs.com/package/omniroute
- Docker: hub.docker.com/r/diegosouzapw/omniroute
- Resilience: Circuit breaker, exponential backoff, anti-thundering herd, TLS spoofing, auto-combo self-healing
📖 Documentation
📘 Getting Started
| Document | Description |
|---|---|
| User Guide | Providers, combos, CLI integration, deployment |
| Setup Guide | Full install methods, CLI tool configs, protocol setup, timeout tuning |
| CLI Tools Guide | Per-tool setup for Claude Code, Codex, Cursor, Cline, OpenClaw, Kilo, Copilot |
| Remote Mode | Drive a remote OmniRoute (VPS) from your laptop CLI via scoped access tokens |
| Claude Code Config | Point Claude Code at OmniRoute (local/remote) with launch + per-model profiles |
| Quick Start | 3-step install → connect → configure |
🔧 Operations & Deployment
| Document | Description |
|---|---|
| Docker Guide | Docker run, Compose profiles, Caddy HTTPS, tunnels, image tags |
| Podman Guide | Quadlet systemd integration, podman-compose, SELinux |
| VM Deployment | Complete guide: VM + nginx + Cloudflare setup |
| Fly.io Deployment | Deploy to Fly.io with persistent storage |
| Termux Guide | Run OmniRoute on Android via Termux |
| PWA Guide | Progressive Web App install, caching, architecture |
| Uninstall Guide | Clean removal for all install methods |
| Environment Config | Complete .env variables and references |
🧠 Features & Architecture
| Document | Description |
|---|---|
| Architecture | System architecture, data flow, and internals |
| Compression Guide | 7-option pipeline: off / lite / standard / aggressive / ultra / RTK / stacked |
| RTK Compression | Command-output compression, filters, trust, verify, raw-output recovery |
| Compression Engines | Caveman, RTK, stacked pipelines, dashboard/API/MCP surfaces |
| Compression Rules Format | JSON rule-pack schemas for Caveman and RTK filters |
| Compression Language Packs | Language detection and Caveman rule-pack authoring |
| Resilience Guide | Circuit breakers, cooldowns, queue, anti-thundering herd, TLS spoofing |
| Auto-Combo Engine | 9-factor scoring, mode packs, self-healing |
| Proxy Guide | 3-level proxy system, 1proxy marketplace, registry CRUD |
| Free Tiers | 25+ free API providers consolidated directory |
| Features Gallery | Visual dashboard tour with screenshots |
| Codebase Documentation | Beginner-friendly codebase walkthrough |
🤖 Protocols & APIs
| Document | Description |
|---|---|
| API Reference | All endpoints with examples |
| OpenAPI Spec | OpenAPI 3.0 specification |
| MCP Server | 95 MCP tools, IDE configs, Python/TS/Go clients |
| MCP Server Guide | MCP installation, transports, and tool reference |
| A2A Server | JSON-RPC 2.0 protocol, skills, streaming, task mgmt |
| A2A Server Guide | A2A agent card, tasks, skills, and streaming |
📋 Project & Quality
| Document | Description |
|---|---|
| Contributing | Development setup and guidelines |
| Changelog | Full per-version release history |
| Security Policy | Vulnerability reporting and security practices |
| i18n Guide | 40+ language support, translation workflow, RTL |
| Release Checklist | Pre-release validation steps |
| Coverage Plan | Test coverage strategy and 21,000+ test suite |
⭐ Top Contributors
OmniRoute is shaped by a passionate open-source community. These individuals have made exceptional contributions that directly impact the quality, stability, and reach of the project. Thank you.
 oyi77 🥇 189 commits • +155K lines Analytics engine, SQL aggregations, proxy marketplace, test coverage |
 Chris Staley 🥈 70 commits • +5.7K lines SSE stream hardening, Responses API, Gemini pagination, test regression fixes |
 zenobit 🥉 62 commits • +24K lines CI/CD pipeline, i18n for 33 languages, Void Linux package, platform fixes |
 R.D. & Randi 🏅 108 commits • +30K lines Endpoints page, tunnel integrations, Docker workflows, A2A status, compression UI |
 benzntech 🏅 22 commits • +7.5K lines Electron desktop app, auto-updater, release build workflows, cross-platform CI |
 herjarsa 🏅 21 commits • +6K lines Zero-latency combos, vision-bridge auto-routing, catalog context-length, resilience 429 hints |
🙏 These contributors' features, bug fixes, and infrastructure improvements are a core part of what makes OmniRoute reliable and feature-rich. Every pull request, every test case, and every i18n translation file matters. Open source is built by people like them.
👥 280+ Contributors
How to Contribute
- Fork the repository
- Create your feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
See CONTRIBUTING.md for detailed guidelines.
Releasing a New Version
# Create a release — npm publish happens automatically
gh release create v3.8.2 --title "v3.8.2" --generate-notes
📊 Stars
🌍 StarMapper
🙏 Acknowledgments
OmniRoute stands on the shoulders of giants. It started as a fork of 9router and a TypeScript port of the Go project CLIProxyAPI — and from there, every subsystem below was inspired by an open-source project that got there first. Each one shaped a concrete piece of OmniRoute. This is our thank-you to all of them. 🙏
⭐ star counts as of June 2026 — go give these projects a star.
🧬 Lineage & gateway
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| 9router · decolua | 19.0k | The original project this fork is built on — extended here with multi-modal APIs and a full TypeScript rewrite. |
| CLIProxyAPI · router-for-me | 38.8k | The Go implementation that inspired this JavaScript / TypeScript port. |
| LiteLLM · BerriAI | 52.1k | The AI gateway whose public pricing dataset feeds our cost-tracking sync and whose provider-normalization model informed our routing. |
🗜️ Context & token compression — engines
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| Caveman · JuliusBrussee | 78.2k | The viral "why use many token when few token do trick" project — its caveman-speak philosophy powers our standard compression mode and 30+ filler/condensation rules. |
| RTK – Rust Token Killer · rtk-ai | 67.3k | High-performance command-output compression — inspired our RTK engine, JSON filter DSL, raw-output recovery and the stacked RTK → Caveman pipeline. |
| headroom · headroomlabs-ai | 54.5k | Reversible context-compression (SmartCrusher) — inspired our headroom engine and the ccr retrieve-marker pattern. |
| LLMLingua · Microsoft | 6.4k | Prompt-compression research (LLMLingua / LLMLingua-2) — inspired our async, code-safe, fail-open llmlingua engine. |
| llmlingua-2-js · atjsh | 28 | The JS/ONNX port (MobileBERT / XLM-RoBERTa) used as the worker-thread backend for our LLMLingua engine. |
| Troglodita · Lenine Júnior | 16 | PT-BR token compression — powers our pt-BR language pack: pleonasm reduction and filler removal tuned for Brazilian-Portuguese grammar. |
| ponytail · DietrichGebert | 68.8k | The viral "lazy senior dev" YAGNI-coder skill — inspired our less-code Output Style: smallest-working-change steering that cuts generated code (the output-axis sibling to Caveman's terse prose). |
🧩 Compact formats, token research & code-aware tooling
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| TOON · toon-format | 24.7k | Token-Oriented Object Notation — its columnar, header-plus-rows model shaped our tabular compaction stage. |
| GCF – Graph Compact Format · Blackwell Systems | 14 | Schema-aware "JSON for LLMs" notation — co-inspired our lossless homogeneous-array compaction with [N rows] markers. |
| token-optimizer-mcp · ooples | 421 | Brotli/SQLite cache + per-session context-delta — inspired our session-dedup engine. |
| token-savior · Mibayy | 1.0k | Bash-output compaction + MCP profiles — inspired our compression bail-out discipline and MCP tool-manifest reduction. |
| token-saver · ppgranger | 110 | Content-aware, per-file-type output compression with failure-aware bail-out — validated our per-type dispatch and minimum-gain skip. |
| token-optimizer · alexgreensh | 1.5k | "Find the ghost tokens" — its offload + recoverable-handle pattern informed our CCR offload thinking. |
| TokenMizer · Shweta-Mishra-ai | 2 | A session-graph + cross-turn line-dedup blueprint that informed our session-dedup design. |
| OmniCompress · jessefreitas | 2 | Rust columnar-JSON + content-addressed retrieve + cross-message dedup — validated our headroom/ccr/session-dedup engine design and the cache-stable "compressed form is position-independent" invariant. |
| mcp-compressor · Atlassian Labs | 89 | MCP tool-schema/description compression — informed our MCP tool-manifest cardinality reduction. |
| RepoMapper · pdavis68 | 181 | Aider-style repo-map ranking — informed our repo-map / retrieval-ranking exploration. |
| quiet-shell-mcp · mrsimpson | 4 | Declarative shell-output reduction over MCP — validated our declarative bash-output compaction. |
| ts-morph · David Sherret | 6.1k | TypeScript Compiler API toolkit — inspired our parser-based comment removal that preserves string, template and regex literals. |
🧠 Memory & RAG
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| Mem0 · mem0ai | 59.8k | Universal memory layer — its proxy-as-write/read-boundary model shaped our memory architecture. |
| Letta (MemGPT) · letta-ai | 23.6k | Stateful agents with tiered memory — inspired our Context Control & Recovery (CCR) tiered model. |
| WFGY · onestardao | 1.8k | The ProblemMap taxonomy of 16 recurring RAG/LLM failure modes — the shared vocabulary in our troubleshooting guide. |
🛰️ Traffic inspection, MITM & transparent proxy
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| llm-interceptor · chouzz | 48 | MITM interception/analysis of coding-assistant ↔ LLM traffic — our Traffic Inspector ports its SSE merge, conversation normalization, host passthrough and secret masking (MIT). |
| ProxyBridge · InterceptSuite | 5.3k | Transparent per-process proxy routing — inspired our crash-safe MITM teardown, socket idle-timeouts, /proc process attribution and TPROXY capture. |
📚 Model data, observability & UI
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| models.dev · SST / OpenCode | 5.6k | Open database of AI model specs, pricing and capabilities — synced natively into our model catalog. |
| React Flow / xyflow · xyflow | 37.4k | The node-based graph library powering our real-time Compression Studio and Combo/Routing Studio. |
| LangGraph · LangChain | 36.1k | LangGraph Studio's live workflow-graph visualization inspired our Studios' real-time cascade view. |
| Langfuse · Langfuse | 30.1k | Its trace → span → generation observability model shaped our Compression Studio waterfall. |
| Kiali · Kiali | 3.6k | Istio service-mesh observability — inspired our circuit-breaker badges and error-edge visuals in the Routing/Combo Studio. |
| lobe-icons · LobeHub | 2.2k | AI/LLM brand logos that render the provider icons across our dashboard. |
🛡️ Security
| Project | ⭐ | How it inspired OmniRoute |
|---|---|---|
| awesome-secure-defaults · tldrsec | 708 | A curated list of secure-by-default libraries that guides our security choices (Helmet.js, DOMPurify, ssrf-req-filter, safe-regex, Google Tink). |
❤️ Support
OmniRoute is free and open source, built and maintained in the open. If it saves you time or money, consider supporting development:
- ⭐ Star the repo — it genuinely helps visibility
- 💖 GitHub Sponsors — fund ongoing maintenance and new providers
- 🐛 Report bugs and share feedback in Discussions
📄 License
MIT License - see LICENSE for details.
⬆ Back to top · Built with ❤️ for the open-source AI community.
OmniRoute v3.8.43 · Node ≥22.0.0 · MIT License · omniroute.online