OmniRoute v3.8.54
- Providers: official "Sign in with ChatGPT" provider (
chatgpt, PKCE + live model discovery); MiniMax Code (managed gateway + dashboard device-code sign-in); RenderHour (images, video, speech); MegaBrain (dual auth); EmpirioLabs, ParalonCloud, deAPI, Replicate chat, Cerebrium, Hetzner Inference and Grokified; nativeopencode-serverexecutor; Microsoft 365 Copilot work/school sign-in; GitHub Copilot embeddings; Gemini Lyria 3 music models; custom JSON API provider nodes (request template + response path). - Routing and combos: endpoint-specific combo routes (chat, search/fetch, embeddings, rerank, images, video, music, OCR, audio) with capability-aware search/fetch combos; new
jevandexpiry-firststrategies; preferred and reserve accounts for Auto combos; opt-in preferred-account response-start failover; opt-in deadline-failure sub-budget that ends a combo with a clean 504. - Search: Keenable web search and fetch provider (keyless by default) and
codex-searchthrough connected Codex accounts. - API keys and usage:
/v1/me/statusreports today's/this week's usage and every per-key limit;/v1/me/sessionsself-service session listing; rolling 5-hour USD budget per key; reusable key policy templates;POST /api/keys/import; usage attributed to coding-agent sessions and projects; provider-reported cache-read and 5-minute/1-hour cache-write tiers carried through analytics. - Agentic pipelines and streaming: repeated-tool-round and consecutive-failure detection across native protocols; opt-in context compaction with per-model budgets;
STREAM_READINESS_TIMEOUTnow says which failure shape occurred; large request bodies toapi.anthropic.comare gzip-compressed like Claude Code. - Privacy and media: opt-in reversible PII token vault (default off, requires
PII_REDACTION_ENABLED); consented, tenant-bound Video Bridge frame drill-down handles with dashboard controls; opt-in image normalization before dispatch. - Dashboard: "New models" inbox across connected providers; Costs > Pricing model comparison across providers and a "Configured only" filter; Quota Tracker order synced across browsers; Cmd+K ranked by match quality; Codex personal plan vs Business seat; "Quota check" toggle for compatible nodes.
- Distribution and release engineering: Docker images also published to the Cheaper Inference organization (
cheaperinference/omniroute) and GHCR, Electron auto-update served fromcheaper-inference/OmniRoute; derived counts refreshed only at release (npm run facts,release:refresh-derived); timestamped migration names; the omni-queue merge queue (shadow mode); complete European Portuguese (PT-PT) catalog and the 66 locales synced at release.
📊 Release by the numbers
| 👥 People who contributed | 121 |
| 📝 Commits in the cycle | 920 |
| 🔀 Pull requests referenced | 466 |
| 📋 Changelog entries | 1,024 |
| 🙌 Contributors credited in entries | 117 |
| 🤖 Automated dependency commits | 10 |
Entries by type
| Type | Count |
|---|---|
| 🐛 Fixes | 626 |
| ✨ Features | 156 |
| 📚 Docs | 73 |
| 🧪 Tests | 70 |
| 🧹 Chore | 35 |
| ♻️ Refactor | 25 |
| ⚡ Performance | 24 |
| ⚙️ CI | 14 |
| 📦 Dependencies | 1 |
🏆 Top 25 contributors this cycle
By commits in 6cc6965075..a9539085bd, author identities consolidated via .mailmap and the merged PR's GitHub login. Bots excluded.
| # | Contributor | Commits |
|---|---|---|
| 🥇 | diegosouzapw | 586 |
| 🥈 | serenity (@serenity-X) | 61 |
| 🥉 | Minxi Hou (@HouMinXi) | 35 |
| 4 | Bl0ck (@Bl0ck154) | 32 |
| 5 | Aldo (@aldoeliacim) | 21 |
| 6 | 小妍儿 ✨ (@xiaoyaner0201) | 15 |
| 7 | Rafa Martins (@rafacpti23) | 12 |
| 8 | Kiaro-Sama (@KiaroSama) | 9 |
| 9 | Andres Felipe Diago Matta (@AndresDiagoM) | 8 |
| 10 | Fouad Salkini (@fouadSalkini) | 8 |
| 11 | shin (@jjxyxsjr) | 7 |
| 12 | Dizzle (@maxmad64bis) | 7 |
| 13 | Praveen K Palaniswamy (@yourspraveen) | 6 |
| 14 | Ankit (@anki1kr) | 5 |
| 15 | Isaias Amaral (@bufftop25) | 5 |
| 16 | Gaurav Dubey (@gaurav0107) | 5 |
| 17 | Federico Mangini | 4 |
| 18 | RayJW | 4 |
| 19 | Felipe Fidelix (@fidelix) | 3 |
| 20 | Innokentiy Solntsev (@insoln) | 3 |
| 21 | Jan Leon (@JxnLexn) | 3 |
| 22 | Tiangao (@tiangao88) | 3 |
| 23 | Cristiano Martini Bergamo (@cmbergamo) | 2 |
| 24 | Tomas Fecko (@feci) | 2 |
| 25 | GodsQuantum | 2 |
✨ New Features
- feat(video): connect consented Video Bridge frame drill-down to the request path — already-derived JPEG frames are published through a shared tenant-bound lifecycle only with explicit operator and authenticated per-part consent (raw videos stay out of retention); JSON/SSE responses return bounded opaque handles, Video settings expose independent retention/remote-read controls and aggregate-only usage, and authenticated preview/standard/detail reads, isolated idempotent deletion, cancellation, TTL/quota cleanup and actual-content variant provenance are supported with bounded queries, sanitized errors and the 32 MiB serialized response limit. Defaults remain off; real-model promotion and deployment require separate evidence (#16074).
- fix(providers): treat TypeSafe (Jev) as a decisions-only provider: a chat request routed to
typesafenow fails fast with HTTP 400 pointing atPOST /v1/systemoneinstead of falling through to the default executor (which targets the OpenAI endpoint with the TypeSafe key), imported or discovered TypeSafe models are never classified as chat models, and Jev aliases get an input-only default price (#14498, #13987) — thanks @seanford - feat(systemone):
POST /v1/systemonecan serveollama-local/<model>decisions models on the local Ollama of the existingollama-localconnection (its base URL, defaulthttp://localhost:11434/v1, never a request field), with Ollama's own limits checked before any upstream call (2–26 candidates per choice/score question, at most 64 questions, no chat sampling fields, raw base64 images only, 64 KiB / 32 MiB body cap); opt-in through theSYSTEMONE_OLLAMA_LOCAL_BACKEND_ENABLEDfeature flag (default off) (#14667) — thanks @PixmaNts - feat(api-keys):
/v1/me/statusnow reports today's and this week's USD, request and token usage, alimits[]list covering every per-key limit (daily/weekly USD usage limits, budgets, token limits, key quota) measured on its own enforcement window, and masked account labels plus cache-firstaccountQuotaslimited to the connections the key can reach and the providers the admin chose to share. The endpoint also acceptsx-api-key. A per-keyanthropicRateLimitHeadersmode (auto/forward/strip) controls whether pooled accounts'anthropic-ratelimit-*andanthropic-organization-idheaders reach the client;autoforwards them only for keys pinned to one shared connection. The API manager gains the matching controls and a per-key details page with usage, limits, account quotas and editors for token limits and key quota (#14771). — thanks @fouadSalkini - feat(usage): attribute proxied requests to coding-agent sessions and projects for team usage reporting, recording session token counters, priced cost, working directory, and project repo (#14833) — thanks @fouadSalkini
- feat(api): add
/v1/me/sessionsself-service API key endpoints to list and inspect agent sessions with project, client, date, and sorting filters (#14840) — thanks @fouadSalkini - feat(api): per-key
anthropicRateLimitHeaderspolicy to forward or strip upstream Anthropic account-identity and quota headers on both streaming and non-streaming responses, preserving legacy forwarding by default (#14862) — thanks @fouadSalkini - feat(chatgpt): Add the official "Sign in with ChatGPT" provider (
chatgpt): dynamic registration with PKCE, live account model discovery, inference through the public Responses API and refresh withearliest_refresh_at— separate fromcodexandchatgpt-web. Routing fails closed to the registration that advertises the model; the ChatGPT catalog keeps its own owner and never inherits Cursorauto*aliases. (#15201 — thanks @JxnLexn) - feat(routing): new
jevcombo strategy — you order the providers, one TypeSafe System Onechoicecall admits the winner plus up to 4 fallbacks (options at or above 0.15 confidence) for each request; a missing key, timeout or low confidence falls open to your own order, live session pins skip the call, and the combo editor shows a privacy note because the latest user message is sent to TypeSafe. The TypeSafe call is opt-in through theCOMBO_JEV_STRATEGY_ENABLEDfeature flag (default off): a stored TypeSafe key alone never sends traffic, and with the flag off a jev combo keeps your order (#15279) — thanks @miguelenes - feat(sse): Agentic pipelines detect repeated tool rounds with unchanged, call-matched results and consecutive failed rounds across native protocols, then return control to the planner. Incomplete or ambiguous parallel rounds do not falsely stop execution (#15420) (#19318) — thanks @potatosips.
- feat(sse): Agentic pipelines support opt-in context compaction with per-model character budgets for the planner and each executor fallback, preserved instructions and active tool boundaries, and pre-dispatch oversize rejection. Exact incremental size accounting avoids repeated whole-history serialization; structured screenshot results remain intact (#15421) (#19317) — thanks @potatosips.
- feat(intelligence): the
models_dev_tierintelligence source is now persisted (rebuilt after every capabilities sync and on demand viaPOST /api/intelligence/sync {"rebuildTier": true}),user_overrideentries get a management-authedPOST/DELETE /api/intelligence/overridesAPI, andGET /api/intelligence/syncreports per-source health insourcesDetail(#15906 — thanks @pizzav-xyz) - Add an explicit per-account Codex quota query beside Quota, available even when quota details are hidden, without consuming reset credits. Distinguish current, partial, stale and failed observations; prevent duplicate requests and reconcile the account list without claiming that a cooldown was released. Backend recovery diagnostics and automatic recovery validation remain separate work for #16161. (#16191)
- feat(dashboard): set a model's context window when adding a custom model and from the catalog model list of native providers, so a model whose window the catalog does not know yet no longer falls back to 128k; a context-only override no longer turns a catalog model into a custom one (#16178)
- feat(usage): report provider-reported cache-read tokens in usage analytics and retained daily summaries, with partial-coverage indicators for legacy or filtered history while preserving billing calculations (#16452) — thanks @KiaroSama
- feat(proxy): report per-proxy and aggregate application-payload upload and download traffic, including retry attempts and incomplete transfers, without presenting payload counts as wire traffic or billing measurements (#16455) — thanks @KiaroSama
- feat(gemini-web): add opt-in
gemini-4-argonmodel support to thegemini-webprovider registry with UI mode selection and verification descriptor; off by default until its selectors are checked live, enable withOMNIROUTE_GEMINI_WEB_ARGON_ENABLED=1(#16573) — thanks @SkiiOS - Use registered local-provider URLs for model discovery and synchronized embedding routes, and align Docker Model Runner's default with its engine-neutral OpenAI endpoint. Docker Model Runner embedding requests now accept keyless connections while preserving configured credentials and per-model capability checks. (#16670)
- feat(systemone):
POST /v1/systemonecan serve TypeSafe Jev and Cloudflare Clef / Clef-flash on Cloudflare Workers AI through the existingcloudflare-aiconnection (cloudflare-ai/jev,cloudflare-ai/clef,cloudflare-ai/clef-flash, aliascf/), as a newsystemOneRegistrybackend, opt-in through theSYSTEMONE_CLOUDFLARE_BACKEND_ENABLEDfeature flag (default off) (#16676) — thanks @xiaoyaner0201 - feat(codex): add global and per-connection Ultrafast defaults scoped to GPT-6 Astra and GPT-6.1 Sol, while preserving explicit request and API-key precedence. (#16678)
- feat(mitm): Antigravity MITM can map Claude Sonnet 5.5 / Opus 5.5 (low/medium/high tiers) to OmniRoute routes (#16709) — thanks @lmortezal
- feat(images): Antigravity image edits accept multiple reference images (data URL or raw base64; remote URLs stay rejected), each forwarded as its own inline part; invalid references fail fast with a 400 and text-only answers surface as an error (#16731) — thanks @yheiakadylan
- feat(resilience): dashboard editor for operator error cooldown rules, optional status/match conditions, and an opt-in per-provider catch-all cooldown that never shortens a longer upstream Retry-After/reset hint (#16734) — thanks @zyizyiz
- feat(dashboard): "New models" inbox — model sync records each provider's listing (first sync seeds a baseline) and the Providers toolbar shows newly added models with an unread badge, grouped by provider, with mark-as-read (#16864) — thanks @ErfanBagheri404
- feat(providers):
opencode-serverprovider — a native executor for the opencode CLI's local headless server (opencode serve) that translates OpenAI chat/Responses requests to opencode sessions/parts with streaming, tool-call envelope support; the server is started from the dashboard (local-only/api/services/opencode-server/{status,start,stop}) or withopencode serve, never from a chat request; only the server URL is a dashboard setting, the CLI binary is env-only (OPENCODE_SERVER_BIN) and the spawned server gets a minimal environment; it speaks the opencode v1 server API, and an opencode 2.x server is detected and answered with a clear "unsupported opencode server version" error instead of being treated as healthy (#16995) — thanks @cmbergamo - feat(providers): Added EmpirioLabs AI (
empiriolabs) as an OpenAI-compatible API-key provider (https://api.empiriolabs.ai/v1), with live model discovery limited to chat models and API key checks against the authenticated/v1/account/usageendpoint (#17209, #19316) — thanks @Adam-Dalloul - feat(providers): add the MiniMax Code provider — OmniRoute can now route chat completions through the managed MiniMax subscription gateway under the
minimax-code/mmcodealias. The gateway path reads the local refresh token of the user's own MiniMax Code profile (~/.minimax/auth/prod/<region>/mcode-public/auth.json) and rewrites the auth files of that profile when the account rotates the credential, so the CLI and the local agent pick the rotated credential up on their next session; OmniRoute does not persist a separate provider credential in its connection database. The transport speaks the native tool protocol, so a request that carries client tools comes back astool_callsfor the client to execute, and a refusal of the gateway keeps its own status instead of being renamed into an outage of the model (#17286) — thanks @megamen32 - feat(oauth): the MiniMax Code provider can now be signed in from the dashboard — the provider page offers the MiniMax account device-code flow (RFC 8628 + S256-PKCE), the panel shows the verification URL and the code to confirm, and the credential that comes back is written into the local MiniMax Code profile, so the CLI and the local agent pick it up on their next session; the dashboard connection row stores no access or refresh token; the provider itself stays a no-auth provider: the dashboard connection is the account entry, the local profile is the credential (#17286) — thanks @megamen32
- fix(sse): refuse an oversized MiniMax Code request with a request-scoped 413 — a request that did not fit the ceiling of the managed gateway was refused with a status that the combo layer renamed into a 502, and a 502 is in the model-lockout list, so one request that was too large took the MiniMax leg out of every combo for minutes and the next request answered 503; the refusal is now a 413 marked
invalid_request_error, which is a property of the request and never of the model (#17286) — thanks @megamen32 - fix(sse): the MiniMax Code executor answers 499 while the instance is draining — a graceful shutdown tears the transport down under the requests that are still in flight, so the failure is evidence about our own shutdown and not about the health of the provider, but a 502 said the opposite: it is in the model-lockout list and in the provider-breaker list, and it is not exempt from the connection cooldown, so the account was written as
unavailablewith arateLimitedUntilinto theprovider_connectionstable — a state that outlives the restart; the answer during a drain is now 499, the status an aborted caller already gets, which carriesclient_disconnectedand is exempt from that cooldown (#17286) — thanks @megamen32 - feat(usage): Preserve provider-reported 5-minute and 1-hour cache-write tiers through response translation, call logs, and Cache Health diagnostics; keep legacy unknown tiers explicit and existing quota deductions unchanged; translated Claude and OpenAI usage now also carries
usage.cache_creation(additive) (#17307) — thanks @xiaoyaner0201 - feat(combo): new
expiry-firstcombo routing strategy (the 21st) — orders targets by the provider-level expiry-first salvage score, using the quota window of the model family each target spends, with a relative tie band and priority order kept for targets without quota telemetry (#17829) — thanks @HouMinXi - feat(search): Keenable web search and fetch provider (
keenable-search, aliaskeenable): keyless by default, with an optional API key that only lifts the per-IP rate limits. Serves/v1/search(web only, fallback-only so it never displaces a configured paid provider) and/v1/web/fetch(Markdown only, capped at 100,000 characters) (#17920) — thanks @ilya-bogin-keenable - feat(search): Add
codex-search(aliascodex_search) for web/news search through connected Codex (ChatGPT) accounts. A request that namescodex-searchuses it directly; automatic selection is opt-in behind theCODEX_SEARCH_AUTO_FALLBACK_ENABLEDflag (default off), which keeps configured search providers first and tries Codex before keyless fallbacks while preserving other providers' order, API-key connection scopes and REST provider blocks. Parse bounded SSE responses into source-grounded rows and log upstream tokens/search counts. Search failures only apply an in-memory search-specific backoff, capped at 60 seconds; they never change shared chat-account health. Searches consume subscription quota, use a 60-second default timeout unless explicitly configured otherwise, and treat time/domain constraints as hints. A search-locked account is skipped in favour of another provider rather than another Codex account. (#17933) — thanks @xiaoyaner0201 - Add
POST /api/keys/importto register caller-supplied API keys with create-key policies, atomic duplicate checks, non-mutating UUID retries, management authentication, and masked responses without restoring unrelated configuration (#18010) — thanks @xiaoyaner0201. - Dots Studio tool-call markup:
<dots_function_call>text blocks (structured<invoke>or pure JSON) are converted into real tool calls in the Claude stream translator and the OpenAI response sanitizer, accepting both OpenAI-shaped and Anthropic-shaped client tools (including tools declared without a schema). On the Claude stream an unfinished block, or an opener split mid-tag across chunks, is held back until it completes; on the OpenAI leg the scrub covers non-streaming responses only (streamed OpenAI deltas are passed through unchanged) (#18024). — thanks @pizzav-xyz - feat(providers): RenderHour provider for images, video and speech — sign in with a RenderHour subscription (OAuth PKCE, paid in plan credits) or paste an
rh_…API key (paid from the dollar balance); models asrenderhour/<model>(#18103 — thanks @sha367) - feat(providers): MegaBrain as a dual-auth provider (
megabrain/mb): sign in with a MegaBrain account through Kilo's device-code flow onapi.getmegabrain.comor pastemb-v1-…API keys; requests carryX-KILOCODE-FEATURE: omnirouteso they run on the account's plan (session and weekly limits, shown in Provider Limits), and a 403 for a model the account cannot pay for (plan_tier_required,extra_usage_off) locks only that model instead of banning the connection (#18129) — thanks @sha367 - feat(providers): add an Import auth modal to the Antigravity CLI (
agy) provider page: loopback-only 1-click import of the local CLI login (AGY_TOKEN_FILE, Linux keyring,~/.gemini/antigravity-cli/and AgentControl profiles), single token file/paste and bulk JSON import (#18181) — thanks @ziuus - feat(plugin): render picker names as
Provider - Modelwith a leading free icon, the free-tier budget kept as a· 1M tokens/monthsuffix, the editor of an explicitEditor: Modelname in trailing parentheses, andcombo - slugfor combos. Behavior change:providerTagno longer removes the provider prefix (the provider is now always shown);providerTag: falseonly drops the editor parenthesis, and configs that matched entries by their old exact name need updating (#18230) — thanks @maxmad64bis - feat(providers): GitHub Copilot embeddings (
github/text-embedding-3-small,github/text-embedding-ada-002) through the Copilot API with the exchanged Copilot token, refreshed before dispatch (#18291) — thanks @AppleVibe, @tmaarcxs - feat(providers): the Grok CLI catalog now lists
grok-4.7-build-fast(Grok 4.7 on Grok Build's faster lane, 500k window, low–xhigh efforts) andgrok-code-fast-1, both accepted by the Grok Build proxy but missing from the static catalog that the pickers fall back to before live discovery runs (#18318) — thanks @cristianob - feat(dashboard): the "Import only free models" option now also works on OpenAI/Anthropic-compatible nodes. It keeps only models that the endpoint's own
/modelslist marks as free and that have no paid signal; models without free evidence are never imported as free (#18320) — thanks @IlyaKhalafi / @maryasov - fix(providers): positive discovery prices and explicit paid flags override free catalog entries and Free badges, and remain available through named-provider discovery, cache, and model sync (Refs #17028). (#18320)
- fix(providers): retire five unconditionally free Ollama Cloud entries after checking current model prices, document seven free offers with their eligibility limits, and expose Requesty's NVIDIA trial training restriction (Refs #17028). (#18320)
- fix(providers): stop advertising OpenCode-only free offers as generally available through the proxy, including cached Free badges, while keeping the provider and normal imports available and leaving uncertain pricing unclassified. Existing free-only connections retain their preference and an explicit opt-out control (Refs #17028). (#18320)
- fix(providers): remove static free grants for retired Ollama DeepSeek V4 Flash, Ollama GLM5.1 and Tencent Hunyuan Pro, while preserving normal model imports (Refs #17028). (#18320)
- fix(providers): remove static free grants for AINative’s retired
nous-coderalias and Enterprise-tierdeepseek-r1, preserving normal imports (Refs #17028). (#18320) - feat(dashboard): the Quota Tracker provider order now follows the operator across browsers (saved as
settings.quotaProviderOrderthroughPATCH /api/settings, with local storage kept as the cache), and the Full layout's provider sections get the same drag grips and "Restore default order" action as the overview; display-only, connection priority is never touched (#18344) — thanks @ape1121 - feat(usage): The Cursor Quota card shows per-model API tokens (input, output, cache read/write) and the cost Cursor billed for the current cycle, fetched fail-open so the existing % bars render unchanged when that call fails (#18346) — thanks @AMSeify
- feat(dashboard): "Quota check" toggle for OpenAI- and Anthropic-compatible nodes reads a standard
/creditsbalance into Provider Limits;anthropic-compatible-*keys now use the configurable quota fetcher, which goes through the SSRF-guarded outbound fetch (#18350) — thanks @MWGE22629 - feat(audio): Cloudflare Workers AI Deepgram Nova-3 transcription (
cf/@cf/deepgram/nova-3) with opt-in speaker diarization andverbose_jsonwords, per-speaker segments andnum_speakers(#18352) — thanks @LeandroTechify - feat(providers): custom JSON API provider nodes — an OpenAI-compatible node with
apiType: "custom-json"fills a JSON request template ({{prompt}},{{messages}},{{model}},{{request.<path>}}; substitution only) for any JSON business endpoint, reads the reply from a response path and returns it in the client's format; business requests are never replayed (no retry, no account or combo fallback). API-only for now, seedocs/guides/CUSTOM_JSON_API.md(#18353) — thanks @lyccccy - feat(providers): add ParalonCloud API-key provider (chat + embeddings) (#18356) — thanks @robertjitaru
- feat(dashboard): the OpenAI-compatible Add and Edit provider modals offer the Custom JSON API type — HTTP method, auth (bearer, API key header or none), a request-template editor with simple fields or full JSON, the response path and a "Send test request" button — validated in the browser with the same schema the server uses (#18359) — thanks @lyccccy
- feat(api-keys): rolling 5-hour USD budget per API key, enforced next to the daily/weekly limits and reported on the API key endpoints (#18362) — thanks @sajjadsaharkhan
- feat(services): the in-memory session TTL is now configurable via
OMNIROUTE_SESSION_TTL_MIN(minutes; default stays 15 and invalid values keep the default) (#18394) — thanks @jjxyxsjr - feat(dashboard): Codex connections on the provider page show whether each one is a personal plan or a Business seat, with the ChatGPT workspace name ("· Business · ", role in the tooltip), and the plan badge reads "Business Pro Lite" instead of
self_serve_business_prolite. The usage sync reads the workspace from ChatGPT's accounts endpoint at most once a week per connection, after a successful usage fetch and with a 15 s timeout; until then the kind falls back to the plan type (#18412 — thanks @feci). - feat(auto-combo): the reset-aware no-quota floor score is now a per-combo config field (
resetAwareNoQuotaFloor, default stays 0.5) instead of a hardcoded midpoint (#18435) — thanks @jjxyxsjr - feat(sse): a
STREAM_READINESS_TIMEOUTnow says which failure shape it was: the result carriesanyByteAt(null = the upstream went fully silent, not one raw byte; an epoch-ms value = transport alive with keepalive bytes but no content event) and the 504 response repeats it in anX-OmniRoute-Any-Byte-Atheader for external tooling. The timeout log line names the shape too. Pure telemetry — no routing or retry decision reads it. (#18586) — thanks @jjxyxsjr - feat(providers): add deAPI (
deapi), an API-key media provider on its OpenAI-compatible gateway (oai.deapi.ai/v1) — image generation, text-to-speech, transcription, embeddings, and async video through a newdeapi-video-jobpreset that downloads the key-gated/v1/videos/{id}/contentresult; the OpenAI-compatible speech path now forwardsinstructionswhen given (#18593) (#19319) — thanks @Pietrus-914 - feat(dashboard): Costs > Pricing gets a Providers "Configured only" filter (providers with a connection, plus no-auth providers that are turned on), and a model search lists only the matching models under each provider (#18651) (#18775) — thanks @yourspraveen
- feat(dashboard): "Needs verification (N)" status pill on the provider connections list that shows only the Antigravity accounts holding a stored Google verify-account link, so operators can work through
403 VALIDATION_REQUIREDaccounts one by one; hidden while no account needs verification (#18662) — thanks @bufftop25 - feat(combo): opt-in
OMNIROUTE_COMBO_MAX_DEADLINE_FAILURESsub-budget terminates a combo with a clean 504 after N stream-readiness deadline failures instead of rotating every remaining target against a stalled upstream (#18721) — thanks @jjxyxsjr - feat(routing): add endpoint-specific combo routes with independent targets, strategies and config for Chat, search/fetch, embeddings, rerank, images, video, music, OCR and audio surfaces.
/v1/search,/v1/web/fetch,/v1/music/generations,/v1/music,/v1/sound-generationand/v1/ocrresolve saved combo names while preserving native response contracts, account pins and caller key scopes. Mixed combinations filter incompatible targets before credential selection. Related to #10274; fixes #17858. (#18731) — thanks @Federicomang - feat(translator): per-node
thinkingReplaycapability declaration (redacted/text/none) on themodelCompatOverridesentry, settable throughPUT /api/provider-models, overrides the provider-prefix assumption about replayed thinking shape —textforces plainthinkingblocks for Claude-shape upstreams that rejectredacted_thinking,noneleaves replayed thinking untouched and keeps thinking enabled for upstreams that want no precursor (#18739) — thanks @jjxyxsjr - feat(vision): opt-in
OMNIROUTE_IMAGE_NORMALIZATIONnormalizes inline base64 images before dispatch (RGB, ≤4096px long edge, ~4MP, baseline JPEG/PNG) so Qwen2-VL-style vision pipelines stop 400ing or hanging on oversized/exotic images (#18740) — thanks @jjxyxsjr - feat(dashboard): Costs > Pricing can compare one model across providers: a View: Compare models toggle lists each model with every provider offering it, input / output / cached price, cheapest first, unpriced offers marked as such; the Configured-only and Auth filters apply and variants stay on their own row (#18775, closes #18642) — thanks @yourspraveen
- feat(providers): Gemini AI Studio Lyria 3 models (
lyria-3-pro-preview,lyria-3-clip-preview,lyria-3.5) are callable throughPOST /v1/music/generationsinstead of being advertised as music and rejected with400 Invalid music model(#18797) — thanks @LuisAlejandro - feat(gemini): Pass an explicit per-image
media_resolution(lowthroughultra_high) through the OpenAI to Gemini conversion asPart.mediaResolutionfor inline and remote images; opt-in, anddetailkeeps its meaning (#18844) — thanks @Bl0ck154 - feat(dashboard): the combo builder shows a Nadir settings section (API key, timeout, self-hosted base URL) when the router strategy is
nadir, so a dashboard-built combo no longer needs hand-edited JSON or env vars to setconfig.nadir(#18846) — thanks @doramirdor - feat(providers): add an opt-in schema-driven Fal gateway behind
FAL_NATIVE_ENABLED(off by default): the/v1/falroutes, the/api/providers/{id}/falschema routes, a Fal panel on the provider page and live catalog discovery, plus schema validation of Fal requests on the image, video and music routes. With the flag off, Fal requests keep their established shape, endpoint mapping and response format. (#18849) — thanks @KiaroSama - feat(executors): Gzip request bodies of 4096+ characters sent to the official
api.anthropic.com, as Claude Code 2.1.3xx does. Long agent sessions re-upload 0.5-10 MB of JSON each turn; it compresses 3-4x (gzip level 3: 3 MB in 27 ms). Skipped behind a proxy, TLS impersonation orNODE_EXTRA_CA_CERTS. A refused encoding (415, a non-Anthropic 400/403, or an origin JSON-parse 400) is resent uncompressed once and that origin stays uncompressed for one hour; ordinary API errors are never resent. Opt out withOMNIROUTE_GZIP_REQUEST_BODIES=0. (#18889) — thanks @serenity-X - feat(observability): opt-in V8 heap pressure log: with
OMNIROUTE_HEAP_DIAGNOSTICS_ENABLED=1the event-loop stall recorder logs heap, RSS and inflight numbers at most once a minute whenheapUsedpassesOMNIROUTE_HEAP_DIAGNOSTICS_THRESHOLD_MB(default 900) (#19054) — thanks @Bl0ck154 - feat(usage): Zed Hosted Models connections show the remaining USD credit on the Provider Limits page when the operator pastes the optional Orb billing portal link into the connection settings, stored as a credential (encrypted at rest, stripped from API responses) and displayed without affecting routing or connection parking (#19065) — thanks @wray-lee
- feat(providers): Upgrade the existing Microsoft 365 Copilot provider with work/school PKCE sign-in, account-safe reauthentication, the documented model catalog, single-generation emulated tools, bounded native turns, distinct model-licence and account-throttle handling, and Copilot safety refusals (
Disengaged) surfaced as terminal provider safety stops rather than retryable upstream errors. Follow-up requests continue the same Copilot conversation: one that extends the previous reply exactly (same connection, model profile and tool set, unchanged history, and that reply's tool-call ids or text) sends only the new messages instead of opening a new conversation with the whole history, which is what Copilot's per-user thread throttle counts; anything doubtful starts fresh, and a content-lessInternalErroris re-sent once in a fresh conversation and otherwise surfaces asm365_internal_error. Continuations live in process memory, so a restart starts fresh conversations. Image input works like the web client: each image (data URL or public HTTPS URL) is uploaded into the turn's conversation and attached to the message, and a failed upload fails the request./v1/images/generationsservescopilot-m365-web/m365-copilot-imagethrough Copilot's built-in image tool (one turn per image; the artifacts are downloaded with a Designer token minted from the sign-in's refresh token, and a used-up image budget is a 429). Answers keep their web sources: Copilot's inline citation markers become numbered links withurl_citationannotations instead of leaking as private-use characters, and its reasoning summary is returned asreasoning_content. An expired or revoked sign-in is reported as401 m365_unauthorizedinstead of a generic connection failure, and Test Connection fails on an expired sign-in that nothing can refresh, while a signed-in connection's test renews an expired access token instead of answering "Provider test not supported". A refusal carryingReasoningTurnExhausted(the per-account cap on Think deeper turns) is reported as429 reasoning_turns_exhausted, not as a model's priority access. The same refusal takes the othergpt-5.5/gpt-5.6Think deeper models out of rotation on that connection, so a combo does not spend a turn on each. Legacy HAR connections remain supported. (#19122) — thanks @aldoeliacim - feat(executors): opt-in Bun relay (
OPENCODE_FREE_TIER_RELAY_URL) and canonical OpenCode CLI prompt/tools for gated free-tier OpenCode requests, which the upstream otherwise answers with 429 FreeUsageLimitError to non-Bun clients; calls to injected CLI tools are filtered from the stream; plaintexthttp://relay URLs are accepted for loopback hosts only and a non-loopback relay bind requiresOPENCODE_FREE_RELAY_TOKENand a proxy allowlist (#19128) — thanks @DavyMassoneto - feat(skills): opt-in
agent-sandboxruntime for the skill sandbox —execute_command/eval_codecan run in a remote Kubernetes sandbox (kubernetes-sigs/agent-sandbox warm pools, e.g. gVisor) instead of a local container; selected withSKILLS_SANDBOX_RUNTIME=agent-sandbox, never auto-detected (#19130) — thanks @xiaoyaner0201 - feat(antigravity): add opt-in, per-account Antigravity physical-send phase timing diagnostics without logging request content or credentials (#19133) — thanks @Bl0ck154
- feat(antigravity): Expose physical-send timing diagnostics as a saved Settings → Resilience toggle. New sends use updated values through the settings cache without server restart. (#19133) — thanks @Bl0ck154
- feat(providers): add Replicate chat, Cerebrium, Hetzner Inference and Grokified providers, the MarkItDown MCP launcher, and desktop launchers with an Arabic setup guide (#19206) — thanks @jowimila
- feat(api): Add reusable key policy templates with editable drafts, secret-free capture and atomic key creation (#19311) — thanks @Szqub.
- feat(pii): opt-in reversible PII vault (
PII_REVERSIBLE_VAULT_ENABLED, default off, needsPII_REDACTION_ENABLED): request PII becomes per-request surrogate tokens instead of irreversible placeholders, and the chat handler restores the original values in 2xx responses (SSE split across events and JSON, tool arguments and reasoning included). Providers, fusion judges, logs and caches only see tokens; error bodies keep the tokens; memory, idempotency and semantic-cache writes are skipped for masked turns (#19326, builds on #18017) — thanks @Ardem2025 - feat(webhooks): a new
provider.exhaustedwebhook event alerts you when every account of a provider is cooling down, at most once per provider every 10 minutes. (#19566) — thanks @serenity-X - feat(db): New migrations are named
YYYYMMDDHHMMSS_description.sql(UTC,npm run db:new-migration -- <description>) instead of the next number, so parallel PRs no longer collide. The numbered migrations (001–241) stay frozen. The runner orders by the numeric value of the prefix. Against the base, the migration gate refuses deleted merged migrations and unsafe renames, and runs a commutation check on out-of-order and edited migrations. Release-Green applies every migration to an empty database and compares with the latest release tag. The order ofRENAMED_MIGRATION_COMPATIBILITYentries no longer matters: a database whose old PR rows block each other's slots boots instead of stopping withCannot reconcile(#20169) - feat(docker): releases are also published to the Cheaper Inference Docker Hub organization (
cheaperinference/omniroute) next todiegosouzapw/omniroute, which keeps its pull history and every image; without the org credentials the copy is skipped with a warning, anddocker-org-mirror.ymlrepublishes a released version later. The desktop app now auto-updates fromcheaper-inference/OmniRoute(#20344) - Add bounded response headers for account fallback count/reasons, same-account retries and routing wait, preserving response routing policies on immutable-header streaming responses. (#19163) — thanks @Bl0ck154
- feat(analytics): split the summary into total tokens, total input, input excluding cache reads, and cached input, with explicit incomplete-data handling and localized explanations while preserving provider accounting. (#19538) — thanks @KiaroSama
- feat(codex): opt-in auto-spend of Codex reset credits that are about to expire, so a banked credit is not silently lost when nobody opens the dashboard before its expiry (#18630) — thanks @Junior-HJ
- feat(dashboard): Cmd+K results are ranked by how well the page name matches instead of sidebar order: an exact name first, then a name or word that starts with the query, then a name that merely contains it, then matches in the subtitle or section. Typing
lognow shows Logs before Log export and the "Catalog" substring hit in Radar Catalog;log export,proxy logsandradarstill open their exact pages first. Arrow keys follow the displayed order (#17915) (#19175) — thanks @samimozcan / @anki1kr - feat(settings): Configure preferred-account response-start failover across all chat providers with a 20 s default, 5 s Antigravity / 12 s Gemini 3.8 heuristics and per-provider/per-model overrides in Settings → Resilience. (#19259) — thanks @Bl0ck154
- feat(routing): Opt-in preferred-account response-start failover across providers; one eligible sibling hop, no shared-account cooldown. Default 20 s globally, 5 s for Antigravity and 12 s for Gemini 3.8 Flash, with provider/model override support. (#19259) — thanks @Bl0ck154
- Add a bounded, trusted-internal-only request header to shorten the upstream response-start deadline without changing configured connection timeouts or allowing public clients to override it. (#19162) — thanks @Bl0ck154
- feat(video): expose conservative advanced-policy readiness (direct commit
56e39b80b9) - feat(video): wire independent confirmation into promotion evaluation CLI (direct commit
f699ee3e11) - feat(video): accept approved drilldown page limit with strict legacy alias (direct commit
df19b1407a) - feat(video): mint versioned drilldown tokens and retain only their digests (direct commit
4dd9717ff0) - feat(video): expose private drilldown usage and effective policy in dashboard (direct commit
99488921fe) - feat(ui): expose independent video drilldown consent controls (direct commit
4293493682) - feat(video): expose bounded drilldown handles on chat responses (direct commit
c29199cf45) - feat(video): publish consented tenant-bound drill-down handles (direct commit
a88092c233) - feat(quality): capture and compose common-source coverage evidence (direct commit
749acdd9ce) - feat(quality): freeze coverage source scope by exact commit (direct commit
2fa884af10) - feat(dashboard): "New models" inbox across connected providers (direct commit
73856d7aeb) - feat(search): add Keenable web search and fetch provider (direct commit
7f392f2a67) - feat(dashboard): custom JSON API form in the compatible-provider modals (direct commit
8c1e6ad41c) - feat(audio): Cloudflare Workers AI Nova-3 transcription with diarization (direct commit
8b9c88e954) - feat(dashboard): "Quota check" toggle for OpenAI/Anthropic-compatible nodes (direct commit
8f37d0f858) - feat(providers): custom JSON API provider nodes (request template + response path) (direct commit
9e9a3749fb) - feat(routing): support capability-aware search and fetch combos (#19771 — thanks @Federicomang)
- feat(routing): add capability-specific combo routes (#19771 — thanks @Federicomang)
- feat(pii): wire the reversible PII token vault into the chat pipeline (direct commit
7ab5d308df) - feat(providers): add Hetzner, Grokified, Replicate and Cerebrium (direct commit
cfdfef984a) - feat(quality): test-masking reports gate-mode conditional skips instead of a bare OK (direct commit
fdafea4710) (direct commitc20f9d67cb) - feat(ci): agent-instruction files carry no derived counts (invariant) (direct commit
a780923703) - feat(db): compare migrations with the base and check that out-of-order ones commute (direct commit
3310d89b94) - feat(ci): check:generated-artifacts judges an edited generated file with full history (direct commit
e6b75f7471) - feat(i18n): syncUiKeys rejects unknown locales; a dry run lists the compact catalogs (direct commit
5ffce6416f) - feat(i18n): pr mode judges the locale catalogs a PR changes against the merge base (direct commit
5c54e717ab) - feat(release): npm run release:refresh-derived [--check] (direct commit
9642d234c8) - feat(ci): docs-counts gate in pr and release modes (direct commit
692ea0c018) - feat(release): npm run facts and the committed derived-facts snapshot (direct commit
8804e34acc) - feat(ci): declare the translation family for check:pr-derived-paths (direct commit
c17498a608) - feat(ci): omni-queue workflow and the candidate profile in ci/quality (direct commit
b32d9830c0) - feat(ci): declare the generated and baselines families of derived files (direct commit
144fb9eb67) - feat(ci): resolve the base of an omni-queue candidate push (direct commit
37e1888cc4) - feat(ci): omni-queue history map from queue merge trailers to PR numbers (direct commit
e68488e5ae) - feat(quality): --require-tighten never fails a feature PR for an improvement (direct commit
a712d17340) - feat(i18n): translation-completeness gates run in release mode only (direct commit
e81745421d) - feat(ci): e2e-timings.json is synced at release, not by each PR that adds a spec (direct commit
39a72a8f94) - feat(i18n): syncUiKeys() API and --rewritten-since for the release refresh (direct commit
f68158cd8a) - feat(skills): check:agent-skills-sync judges only the skills a feature PR edits (direct commit
caaa3e1cd5) - feat(ci): omni-queue CAS state store and idempotent tick (direct commit
953ff8d6fc) - feat(ci): omni-queue dashboard, local CLI and label approval handler (direct commit
f054e80fa3) - feat(ci): scope generated-artifact checks to what a feature PR edited (direct commit
00f6716670) - feat(ci): omni-queue bisection of red batches (direct commit
3652598d24) - feat(ci): omni-queue REST client and in-process fake GitHub (direct commit
aaf5eca2c0) - feat(ci): omni-queue release ruleset and apply-ruleset script (direct commit
199856ba92) - feat(ci): check-docs-sync judges the mirrors a PR writes, even in PR mode (direct commit
a579509ad3) - feat(db): timestamped migration names, gate rules and generator (direct commit
5411622aed) - feat(ci): omni-queue candidate verdict and red classification (direct commit
0eee20f98f) - feat(ci): omni-queue candidate composition and landing (direct commit
c9588e7207) - feat(ci): omni-queue eligibility rules and batch selection (direct commit
c72acfae00) - feat(db): order migrations by the numeric value of their prefix (direct commit
ca1389b2e3) - feat(i18n): export syncLlmMirrors() for the release refresh, add --dry-run (direct commit
ced490f4f9) - feat(ci): check-docs-sync skips the translation-mirror checks in PR mode (direct commit
e2957ceb9c) - feat(ci): omni-queue shared contracts, config and module interfaces (direct commit
d7318d9039) - feat(ci): shared pr/release gate mode and a derived-files guard for feature PRs (direct commit
01258da100) - feat(release): refresh prunes and re-queues UI keys, e2e timings step, i18n gates in pr mode on the train (direct commit
7834d4f095) - feat(release): one-time pass that strips derived values from open PRs (direct commit
e7d603a641) - feat(quality): test-masking lists every gate-mode skip helper, not just releaseOnlyUnlessTouched (direct commit
c7f9eafdc5) - feat(ci): bless only with the queue refs protected and the candidate tree recomposed (direct commit
1374b95fed) - feat(ci): omni-queue environment, land token and manual bless (direct commit
e9cddb6f81) - feat(ci): omni-queue status door ruleset and queue refs ruleset (direct commit
c26aeaf8fb) - feat(ci): omni-queue blesses the candidate and lands as the owner (direct commit
7cf4ca4add) - feat(ci): omni-queue landing actor, identity and land status context (direct commit
9e605a3127) - feat(routing): add preferred and reserve accounts to Auto combos (#18430) — thanks @JxnLexn
- feat(i18n): complete European Portuguese (PT-PT) catalog (#18493) — thanks @DarkEsteves
- feat(antigravity): import a manual refresh token, validated before saving (#18497) — thanks @Junior-HJ
- feat(providers): route Mistral to its native /v1/responses endpoint (#18513) — thanks @hartmark
- feat(usage): opt-in auto-apply of banked resets (Codex credits, Claude grants, Claude session reset) (#19267) — thanks @fouadSalkini
- feat(dashboard): ratchet the kit type scale and cut 57 sub-12px hints (#19066) (#19398)
🐛 Bug Fixes and 📝 Maintenance
627 fixes and 235 maintenance entries are listed in full, each with its PR and author credit, in the attached CHANGELOG-v3.8.54.md and in CHANGELOG.md → [3.8.54] (the body of a GitHub release is capped at 125 KB).
🙌 Contributors
Thanks to everyone whose work landed in v3.8.54: