Capability tables are hand-maintained, so a model gains vision or a wider
context only when someone notices and edits the file. This adds a daily
sync that fills the gap for models already in the registry.
How it decides:
- Modalities (vision/pdf/audio/video) belong to the MODEL — every gateway
serving glm-5.3-flash serves the same weights — so they are keyed by
model id and shared. A majority of sources must declare one, which keeps
out lone mis-declarations: minimax-m2.5 (1 of 45), glm-4.7 (1 of 44) and
gpt-oss-120b (2 of 76) are text-only despite a reseller claiming vision.
- Context/output limits belong to the GATEWAY — each truncates differently
(glm-5 ships as 202752/16384 on one host and 204800/131072 on another) —
so they are keyed by provider + model and only the matching provider's
own numbers are trusted.
Both layers are strictly additive and sit BELOW the hand-written tables,
which short-circuit first. A capability already true stays true.
Mechanics: worker thread (the 4MB parse would block the loop ~20ms),
ETag so an unchanged catalog costs one empty request, 60s startup delay,
30min backoff on failure, MODEL_CATALOG_SYNC=off to disable. Only the
~57KB delta is kept; lookups cost ~0.1us via an mtime-guarded cache.
capabilities.js is bundled into the browser through useModelCaps, so it
cannot import node:fs — the server injects the reader via
setCatalogSource() from instrumentation.
visionPatterns.js is the last resort: a model nobody has catalogued yet
still accepts images when its id says so (qwen3-vl-plus, glm-4.6v, llava),
with image-generation and embedding ids excluded.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Vendors shipped four multimodal models the registry did not carry:
- glm-5.3-flash — z.ai's first natively multimodal GLM-5, 1M context,
image + video + pdf input (glm, glm-cn, opencode-go)
- deepseek-v4-flash-vision-exp — image input at V4-Flash text parity,
1M context / 384k output (deepseek, opencode-go)
- grok-4.6, grok-4.5 — 500k context; 4.6 has no text output limit (xai)
Capabilities needed hand entries because the existing globs mis-matched:
*glm-5* and *deepseek-v4* carry no vision, and *grok-4* would have capped
grok-4.6 at 256k instead of 500k. The grok-4.6 pattern sits above the
generic *grok-4* so it wins the first-match lookup.
Also corrects glm-4.6v / glm-4.5v, which were missing video input and
declared no maxOutput, and backfills glm-4.6v on glm-cn — zhipuai serves
it and the sibling provider already listed it.
tests/unit/opencode-go-models.test.js pins the opencode-go model list, so
its expected array moves with the registry.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
On a 409/429 from Antigravity, fetch live quota to learn the exact
per-model resetAt instead of guessing a backoff, then skip only the
exhausted account/model pair until that time.
- antigravityQuota.js: in-memory quota cache, coalesced concurrent
refreshes, 30s throttle per connection (applied to failures too),
keeps known cache when upstream returns 401/403 error payloads
- auth.js: pre-filter exhausted account/model pairs; report the
earliest quota reset when every account is blocked; skip the
30-minute cooldown cap so the upstream resetAt is not truncated
- chat.js: antigravity 409/429 falls back on the RAM cache only, no
persistent modelLock_* for this path
- Logs identify accounts by id prefix, never email or name
Closes#3561
Xquik needs a GET request with x-api-key auth and a tweets envelope
normalizer, neither of which the generic search fallback provides. Adds a
dedicated request builder and normalizer, cursor pagination passthrough,
result-based credit usage reporting, and a validateUrl probe so key
validation hits the no-charge credits endpoint.
The openai-responses branch of compressWithHeadroom returned null without
recording a reason, leaving the diagnostics panel blank and making Codex
translation failures indistinguishable from a successful compression.
Add a "Bulk Add" flow for the grok-cli provider, mirroring the existing
codex one: paste a JSON array/object or drag-drop multiple .json files,
then create all OAuth connections in a single call.
- BulkImportGrokCliModal: flexible JSON parsing (array, single object,
{accounts:[...]}, concatenated objects) + multi-file upload
- POST /api/oauth/grok-cli/bulk-import: serial createProviderConnection,
snake_case/camelCase token fields, email backfilled from id_token or
access_token, authMethod "device_code" to match the login flow
## Features
- **Auth**: native SAML 2.0 SSO alongside OIDC — AuthnRequest generation, ACS
assertion handling, SP metadata export, admin config test, replay-protected
via a `saml_state` cookie matched against `InResponseTo`
- **Providers**: add Alibaba Token Plan (`token-plan.ap-southeast-1`) — the
fourth Alibaba key type, Singapore-only and OpenAI-compatible transport only
- **Providers**: add `glm-5.3` to GLM Coding and GLM (China)
- **Providers**: Kimchi accepts API keys as well as OAuth (dual auth), with a
working Test Connection for both modes
- **Antigravity**: add Gemini 3.7 Flash and its tiered high/medium/low variants
(also in the Gemini registry) with pricing and quota tracking
- **TTS**: add Fish Audio — model id travels in an HTTP `model` header, voice
is a `reference_id` (preset or cloned voice model)
- **OpenCode-Go**: route by request format via declared transports instead of
forcing every client into `/messages` — Codex/OpenAI clients no longer pay a
lossy Responses→OpenAI→Claude double translation. Per-model `supportedFormats`
guard; the bespoke executor is gone (its shared `_lastModel` cache could cross
auth headers between concurrent requests)
- **Usage**: dedup + cache Claude quota calls (120s TTL keyed by access token,
in-flight promise dedup, last-good read on soft failure) to stop multiple
tabs tripping 429; manual refresh (↻) sends `force=1` to bypass the cache
## Fixes
- **Docker**: ship `sql.js` in the image so the pure-JS DB fallback can start —
file tracing carried the package's JS without `dist/sql-wasm.wasm`, so a
container with no native driver aborted with ENOENT and never got a database
(#3248)
- **Usage**: read Gemini `usageMetadata` out of the antigravity `{ response }`
envelope — every non-streaming antigravity request logged `IN 0 | OUT 0`
(#3260)
- **Claude**: re-anchor passthrough cache breakpoints — the client's own
`cache_control` markers point at pre-normalization offsets, so the tail was
re-cached every request. Last system block and last tool pinned at 1h TTL,
last assistant turn at 5m, mid-conversation system messages folded into the
neighbouring user turn instead of hoisted into `body.system`
- **Combos**: detect images from Hermes and attachment payloads (`images[]`,
`experimental_attachments`, message-level `image_url`/`audio_url`, inline
`data:` URIs) so the Vision Adapter auto-switch fires for Hermes/Ollama/
Vercel AI SDK shapes
- **Kiro**: intercept chat via `x-amz-target` — Kiro IDE 1.0.228+ moved
`GenerateAssistantResponse` to `POST /` + header, bypassing MITM. Also emit
the now-mandatory initial-response frame and map the `auto` model slot
- **Kiro**: report real output tokens and stop discarding usable turns
- **Qoder**: detect billing blocks at stream start and return a synthetic 403
so combo/account fallback triggers instead of leaking the error into chat
- **Antigravity**: strip competitive system prompts (Zed IDE's Claude-agent
prompt) that Antigravity flags with a 429 Quota Exhausted
- **OpenCode**: send the official client fingerprint on free-tier requests so
the Console stops classifying traffic as unidentified and rate-limiting it;
session id resolves conversation-stable to preserve prompt caching
- **Responses**: don't close the message on an empty `tool_calls` array — some
providers attach one to every chunk, and the truthy check ended the message
on the first content token (#3234)
- **Translator**: preserve `prompt_cache_key` when converting chat to responses
- **Models**: expose snake_case token limits on `/v1/models`
- **Combos**: strip `stream_options` from the Fusion panel fan-out to avoid a
DeepSeek 400 (#3024); raise the dashboard model-test probe budget to 1024 and
soft-pass reasoning-only responses (#3010)
- **Headroom**: the toggle reflects the `headroomEnabled` setting even when the
proxy is down — it previously showed OFF while the engine kept calling
`/v1/compress`; proxy status stays visible via the status chip
- **Hermes**: add the `api_key` parameter to the model block in YAML config
- **Providers**: add llm7 to provider test support
## Docs
- **i18n**: add Spanish, French, and Brazilian Portuguese README translations
## Security
- **Real IP**: `x-9r-real-ip` and the Host fallback were trusted from
client-controlled headers whenever `custom-server.js` was not in the request
path (`npm run start`, `start:bun`), letting a remote caller pose as local to
skip API key auth and reach `LOCAL_ONLY_PATHS` (`/api/mcp/*`,
`/api/tunnel/enable`, `/api/auth/reset-password`). The server now stamps a
per-process `x-9r-peer-token` on every request it sanitizes and only trusts
`x-9r-real-ip` behind it — falling back to Host in development and failing
closed in production (GHSA-pjm4-8fpg-f9p6). Also fixes IPv6 loopback
detection (`::1`, `::ffff:127.0.0.1`) and routes `npm run start` /
`start:bun` through `custom-server.js`
- **Search**: `resolveBaseUrl()` rejects client-supplied non-public baseUrls
(SSRF guard on `/v1/search`)
- **Login**: fresh-install remote login with the default password returns 403
without issuing a JWT
- **Usage**: `/api/usage/request-details` redacts request/response payloads
opencode-go hard-coded targetFormat: claude per model, so every client
format was force-routed to /messages (Codex/OpenAI clients paid a lossy
Responses->OpenAI->Claude double translation). Declare the existing
upstream multi-endpoint transports [openai, claude, openai-responses]
and guard per model via registry supportedFormats: kimi/glm/mimo only
support /chat/completions, minimax/qwen add /messages, deepseek adds
/responses. Undeclared models keep the upstream default.
Drop the bespoke OpenCodeGoExecutor (its shared _lastModel cache could
cross auth headers between concurrent requests); DefaultExecutor already
consumes runtimeTransport and injects reasoning content.
Next file tracing follows JS imports, and sql.js loads dist/sql-wasm.wasm by
path at runtime, so the standalone output carries the package's JS without its
wasm binary. When both native drivers fail the last-resort adapter then aborts
with ENOENT on the missing binary and the container never gets a database.
The CLI bundle already guards this explicitly (build-cli.js step 3b,
ensureModuleInBundle("sql.js")); the image just never got the same treatment.
Copy the package the same way node-forge and next already are.
Fixes#3248
Antigravity and gemini-cli wrap their payload in { response: {...} }.
extractUsageFromResponse only tested top-level usageMetadata, so every
non-streaming antigravity request logged zero usage (IN 0 | OUT 0) and
zeroed rows in the usage dashboard. Read the envelope the same way
usageTracking.js and nonStreamingHandler.js already do; top-level
metadata keeps priority and the OpenAI/Claude branches are untouched.
Fixes#3260
x-9r-real-ip and the Host fallback were trusted from client-controlled
headers whenever custom-server.js was not in the request path (npm run
start, start:bun), letting a remote caller pose as local to skip API key
auth and reach LOCAL_ONLY_PATHS (/api/mcp/*, /api/tunnel/enable,
/api/auth/reset-password).
custom-server.js now generates a per-process secret at boot and stamps it
as x-9r-peer-token on every request it sanitizes. hasTrustedPeerHeaders()
(src/lib/auth/trustedPeer.js) gates trust in x-9r-real-ip on that secret;
otherwise the guard falls back to Host only in development, and fails
closed in production. Same gate on loginLimiter.getClientIp() so a spoofed
header cannot rotate the login lockout bucket.
Also: fix isLoopbackHostname for IPv6 (::1, ::ffff:127.0.0.1) which the
old split(":")[0] reduced to empty string; route npm run start /
start:bun through custom-server.js (postbuild copies it into
.next/standalone, build-cli.js fails without it) so documented deployments
keep passwordless local access.
Fourth Alibaba key type — Coding Plan (alicode/alicode-intl) and Model Studio
(alims-intl) both reject Token Plan keys. Registry entry only; PROVIDER_MODELS
builds from providers/registry so no executor or translator work is needed.
Singapore-only (eu-central-1 answers IllegalEndpoint) and OpenAI-compatible
transport only (the Anthropic surface is not authorized for this plan).
Closes#2754Closes#2806
Add gemini-3.7-flash and its tiered high/medium/low variants to the
Antigravity and Gemini registries, with matching capabilities, pricing
and Antigravity quota tracking.
extractModel now recognises gemini-3.7-flash-tiered alongside 3.6 and
derives the version from the request, so thinkingLevel still maps to the
right tiered alias.
Closes#3286Closes#3281
Registry entry plus one config-driven FORMAT_HANDLERS handler. The model id
travels in an HTTP `model` header rather than the JSON body, and the voice is
a reference_id (preset or cloned voice model).
Closes#2411
Zhipu released GLM-5.3 on both api.z.ai and open.bigmodel.cn coding
endpoints. Verified live against both, returning model:"glm-5.3" with
native reasoning_content.
No other changes needed: the '*glm-5*' family pattern in capabilities.js
and 'glm-5*' in pricing.js already cover it.
Passthrough kept the client's own cache_control markers, which point at
pre-normalization offsets. Once normalize/dedupe reshaped system and tools,
the breakpoints landed mid-array and the tail was re-cached every request.
- Pin the last system block and last tool at ttl 1h (was the client's 5m)
- Anchor the last assistant turn at 5m, falling back to the final message
so a first turn still gets a breakpoint
- Fold mid-conversation system messages into the neighbouring user turn
instead of hoisting them into body.system, where the volatile token
counters invalidated the prefix on every request
- Run the anchoring after every token saver, at the final body
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Inspect images[], experimental_attachments/attachments, message-level
image/image_url/audio_url, and inline data:image|audio|pdf URIs on trailing
user turns so Vision Adapter auto-switch fires for Hermes/Ollama/Vercel AI
SDK shapes. stripOpenAI now also drops msg.images and image attachments when
the active model lacks vision support.
Kimchi's transport is OpenAI-compatible (Authorization: Bearer) but the
registry declared it OAuth-only, so the dashboard, /api/providers, and
the connection test all rejected API keys. Enable dual auth
(authModes: ["oauth", "apikey"]) and add a kimchi case to
testApiKeyConnection so the Test Connection button works for both modes.
Regenerate the golden snapshot with the Kimchi entries (+ other
previously-missing providers).
Mirror the official opencode CLI fingerprint (User-Agent, x-opencode-session, x-opencode-request, x-opencode-project) on free-tier requests so the Console no longer classifies traffic as an unidentified client and rate-limits it with FreeUsageLimitError / HTTP 429.
Session id resolves conversation-stable via resolveSessionId (client session to assistant-text hash to connection) to preserve prompt caching, normalized into opencode ses_ format with a generated fallback. When the downstream client is already opencode, its headers are forwarded as-is.
Zed IDE injects a Claude-agent system prompt that Antigravity flags as
competitive, blocking the request with a 429 Quota Exhausted response.
Scan systemInstruction.parts and remove the prompt before dispatch.
Kiro IDE 1.0.228+ moved GenerateAssistantResponse from path
/generateAssistantResponse to POST / + x-amz-target header, so chat turns
bypassed MITM. The SmithyMessageDecoderStream also now requires an
initial-response frame at stream start, and agent/vibe mode sends
modelId "auto" which had no mappable slot.
- Add isChatRequest() header-based match for kiro in mitm/config.js
- Add buildInitialResponseFrame/withInitialFrame to emit the mandatory
initial-response once per stream (kiro.js)
- Add "auto" model slot and update mitmDomain to runtime.us-east-1.kiro.dev
Peek the first SSE frame in wrapQoderSSE; if statusCodeValue != 200 and the
body carries a billing signature (code 112/10605 or pricingUrl), return a
synthetic 403 so chatCore marks the connection unavailable and triggers
combo/account fallback instead of leaking the error text into chat.
wrapQoderSSE becomes async; consumed peek bytes are re-processed in the
stream start() seed loop so nothing is dropped.
Some providers (e.g. codebuddy/cbcn) attach an empty tool_calls array to every streaming chunk. An empty array is truthy in JS, so the guard 'if (delta.tool_calls)' closed the message on the first content token and emitted response.output_text.done early, dropping the remaining deltas. Guard on a non-empty array; finish_reason still closes the message and real tool calls still close it before emitting function_call items.
fixes#3234
Manual refresh (↻) sends ?force=1 so it bypasses the Claude quota cache (dedup + TTL) added in cd4003bc. Auto-refresh and multi-tab stays cached, so Anthropic's usage endpoint is no longer hammered.
Co-Authored-By: Claude <noreply@anthropic.com>
Multiple tabs/accounts/auto-refresh funneled straight to Anthropic and tripped 429. Add a 120s TTL cache keyed by access token with in-flight promise dedup, serve the last good read on soft failure, and thread a force flag through getUsageForProvider for manual refresh. Also lower the dashboard poll cadence (180s to 600s) and stable group-by-provider so connection order stops jumping.
Co-Authored-By: Claude <noreply@anthropic.com>
Toggle was checked={headroomEnabled && headroomRunning} and disabled when the proxy was down, so a downed proxy showed OFF while headroomEnabled stayed true in the DB. The engine only checks headroomEnabled, so it kept calling /v1/compress. Toggle now reflects the user setting; proxy up/down stays visible via the status chip.
Co-Authored-By: Claude <noreply@anthropic.com>
## Features
- **Providers**: add TokenRouter (300+ models via OpenAI-compatible gateway) with
exact per-model pricing for 110 models and `reasoning_effort` thinking config
- **Providers**: add Self-hosted STT / TTS / Embedding — point 9Router at your own
OpenAI-compatible speech and embedding servers (whisper.cpp, faster-whisper,
Kokoro-FastAPI, llama-server, vLLM, Infinity). Unlike the named cloud providers
these read `baseUrl` per connection, so one provider can front several machines
- **Combos**: default-enable vision/audio capacity adapter (auto-routes to a
vision/audio-capable model when the target lacks that capability, falling back
to `oc/mimo-v2.5-free`), wired into chat handler routing
- **Endpoint**: auto-provision a "Default Key" for first-time users so `/v1`
works without a manual dashboard step
- **Codex**: support GPT-5.6 Max/Ultra reasoning-level overrides (cx/ routes only)
- **Qoder**: support PAT (Personal Access Token) connections end-to-end, alongside
OAuth device flow
- **CLI tools**: add OpenDesign (manalkaff/opendesign) support
- **Headroom**: report effective payload savings (tool schema/history bytes broken
out, byte-savings % reflects actual outbound reduction)
- **Ollama**: Cloud quota tracker (session + weekly) + proactive background OAuth
token refresh scheduler for all providers
## Fixes
- **Providers**: remove Qwen (OAuth flow stopped working reliably)
- **Passthrough**: detect codex-tui/Codex Desktop as native Codex client — they
were falling through to the translator and losing fields like `reasoning.summary`
- **OAuth**: scope antigravity header fixes to loadCodeAssist/onboardUser only
- **OAuth**: keep `open` external in the build so xAI/Grok token refresh works on
Windows
- **OAuth**: declare missing `searchParams` in register-session handler (was a
500 instead of JSON on error)
- **DB**: `ENABLE_REQUEST_LOGS` env var now overrides the UI setting correctly;
observability defaults to off (opt-in)
- **Translator**: preserve Codex Responses Lite tool use across chat-native
OpenAI-compatible providers
- **Translator**: don't drop image-only user messages in `prepareClaudeRequest`
- **Translator**: drop JSON Schema keywords Gemini rejects (`uniqueItems`,
`contains`, `multipleOf`, `unevaluatedProperties`, `unevaluatedItems`,
`contentSchema`)
- **Claude**: remove global header cache that leaked one client's identity
headers onto another client/account sharing the server; gate `anthropic-beta`
by model instead
- **Antigravity**: drop retired Gemini 3.0 quota tiers, show Gemini 3.6 Flash
usage bars
- **Cloudflare AI**: declare API key authentication (dashboard showed "No
connections" despite an active key)
- **GitHub Copilot**: hold monthly-exhausted accounts until UTC month reset
instead of only cooling down 120s
- **CodeBuddy**: dodge Tencent CN content filter, add usage tracking, normalize
codebuddy-intl messages
- **Usage**: stop losing cached prompt tokens in the forced-SSE→JSON path
- **Grok CLI**: display the public subscription tier from the OAuth token claim
- **Providers**: count apikey connections for Ollama free-tier card; free-tier/
apikey providers without `authModes` now default to apikey (were treated
oauth-only)
- **Build**: include static/public assets in standalone output (login page hung
on 404s when run via PM2)
- **Server**: support IntelliJ IDEA OpenAI-compatible clients over HTTP (h2c
upgrade handling)
- **Auth**: redirect already-logged-in sessions away from `/login`
- **CLI tools**: enable Apply button for dynamic OpenAI/Anthropic-compatible
provider connections
- **CLI**: include complete API artifacts in the CLI package
- **TTS**: a bare self-hosted model name is the MODEL, not the voice — `kokoro`
was parsed as a voice against a default model, 404ing or synthesising with the
wrong one
endpoint that drops packets never returns headers, so the request previously
hung indefinitely
Google fingerprints User-Agent/Client-Metadata on loadCodeAssist and
onboardUser, silently refusing to provision a cloudaicompanionProject
when they don't match the real IDE. Split antigravity's headers out of
the shared gemini-cli constants instead of overwriting them, so the fix
doesn't touch gemini-cli or any other provider.
Inspired by #3000 (thanks @stoXmod for flagging the resource-exhausted
issue), rewritten to keep gemini-cli untouched.
Adds exact per-model rates for 110 TokenRouter models (pulled from
TokenRouter's own pricing API) plus a dedicated thinkingFormat case
(reasoning_effort enum low/medium/high/xhigh/max) and the provider
logo. Provider registration itself already landed in a prior commit;
this fills in what PR #3043 added on top.
Sync alias-baseline.json and providers-baseline.json with the current
registry (poolside, tokenrouter, selfhosted-* providers already added;
stale claudeOverlay hook and Kiro X-Amz-Target header already removed).
- Endpoint page auto-creates a "Default Key" when no keys exist yet,
so /v1 works out of the box without a manual dashboard step
- Show/copy key buttons stay visible instead of opacity-0 by default
Qwen OAuth flow (portal.qwen.ai) stopped working reliably; drop the
executor, registry entry, OAuth provider/service, token refresh
profile, usage handler, and related test coverage and baselines.
- detectRequiredCapabilities: infer audioInput/videoInput from block
type and embedded mime, not just vision/pdf
- handleChat / handleSingleModelChat: augment combo and single-model
routing with capacity-adapter models when the target lacks a
required capability, wrapped with history stripping for the
adapter model's context window
- Enable vision + audioInput capacity-adapter pools by default for new
and existing users (mergeWithDefaults backward-compat)
- Fall back to oc/mimo-v2.5-free when an enabled pool has no models
configured, both in the backend resolver and the combos UI (auto
refill on removing the last model from a pool)
- Hide PDF/Video from the Vision Adapter UI (PDF never implemented,
Video lacks translator support) while keeping the settings shape
- Exclude combos from the model picker when opened from the Vision
Adapter section
- mimo-v2.5 registry entry now declares audioInput/videoInput
- Simplify combo strategy and Vision Adapter descriptions
detectClientTool only matched the legacy "codex-cli" User-Agent, so the
current codex-tui CLI and Codex Desktop (UA "Codex Desktop", originator
"codex_work_desktop") fell through to null and lost native passthrough —
their requests got re-translated, stripping/overwriting fields like
reasoning.summary instead of forwarding the client body as-is.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Add Self-hosted STT/TTS/Embedding providers that read baseUrl per connection
instead of a fixed registry endpoint, so 9Router can point at whisper.cpp,
faster-whisper, Kokoro-FastAPI, llama-server, vLLM, Infinity, and similar
OpenAI-compatible local servers.
Self-hosted Embedding refuses to run without a baseUrl rather than falling
back to api.openai.com like openaiCompatNode does, since that fallback would
silently send input text and the API key to OpenAI under a provider named
"Self-hosted". Also fixes embeddingsCore to catch adapter build errors as a
400 instead of letting them escape uncaught, and bounds the upstream fetch
with FETCH_CONNECT_TIMEOUT_MS to avoid hanging forever on a dead endpoint.
Self-hosted TTS treats a bare model value as the model rather than the voice,
since the generic OpenAI TTS convention (bare = voice) is backwards for a
provider where the model is the variable part.
