Whisper-compatible upstreams pick the decoder from the multipart filename
against an allow-list (flac, m4a, mp3, mp4, mpeg, mpga, oga, ogg, wav,
webm) that has no `opus`, and OmniRoute forwarded the client's filename
verbatim. The same bytes transcribed as `note.ogg` and 400'd as
`note.opus`. Since /v1/audio/speech emits audio/opus for
`response_format=opus`, clients re-uploading their own voice notes hit
this on every round trip.
A `.opus` file is Opus in an Ogg container (RFC 7845), so `.ogg` is a
truthful relabel and is already on the allow-list. Rewrite the extension
in getUploadedFileName, the single choke point feeding
buildMultipartBody.
The OpenRouter STT path had the same root cause with a quieter symptom:
`.opus` matched neither its extension list nor its MIME map, so it fell
through to the "wav" default and announced Opus bytes as WAV. Map both
the extension and audio/opus to its already-supported ogg container.
Fixes#10588
* fix(xai): cap chat history at xAI 800-message limit
xAI returns 413 when messages/input exceed 800 items. Token
compression never fires on a long tool loop that still fits the
context window, so trim at the executor edge after Responses
expansion and drop orphaned tool pairs from the cut.
* chore(changelog): attach PR number to xAI 800-message fragment
* fix(xai): resolve TS2339 generic assignment in capXaiRequestHistory
Drop the T extends Record<string, unknown> generic on
capXaiRequestHistory and type it directly as
Record<string, unknown> -> Record<string, unknown>. Assigning
next.messages / next.input onto a generic T was rejected by
TypeScript even though every call site already passes/consumes a
JsonRecord (= Record<string, unknown>), so no caller relied on the
generic preserving a narrower type.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: mikolaj92 <mikolaj92@users.noreply.github.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* fix(audio): fall back nested STT models when the prefix provider has no credentials
Bare ids such as deepgram/nova-3 prefix-match the native provider and 400
when that key is missing, even if OpenRouter lists the same model. Retry
the gateway and mention qualified catalog ids in the error.
Closes#10583
* test(audio): scope whisper-1 fallback test to a 2-provider registry
nanogpt was added to AUDIO_TRANSCRIPTION_PROVIDERS (already merged,
unrelated to this fix) with a bare "whisper-1" model id, which now
intercepts findAlternateAudioProvider's first candidate before the
qualified-alias branch this test exists to cover. Scope the test to a
local {openai, openrouter} registry subset so it deterministically
exercises the qualified `${provider}/${model}` fallback regardless of
future providers that also list a bare "whisper-1" id.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* feat(providers): complete Jina AI via OmniRoute including Omni multimodal
Dashboard and env keys share one Jina credential pool, native v5 Omni
{text}/{image}/{content} docs pass through /v1/embeddings intact, and
classify/segment/search are proxied without a third unused Jina card.
* chore(changelog): name Jina complete-provider fragment for #10581
* feat(providers): make Gemini Embedding 2 multimodal work via OmniRoute
Route gemini-embedding-2 through embedContent/batchEmbedContents so N
OpenAI input items become N vectors, pass through native multimodal
parts, and use dashboard Gemini keys (GEMINI_API_KEY only as fallback).
* fix(providers): resolve rebase fallout for Jina/Gemini embeddings
- narrow the two new no-explicit-any violations introduced by this PR
(validateJinaFoundationProvider's params + catch, search.ts's
normalizeJinaSearchResponse data param)
- cast credentials to Record<string, unknown> at the two quota-preflight
call sites in src/sse/services/auth.ts so the new JinaEnvCredentials /
GeminiEnvCredentials union members type-check without loosening the
allRateLimited narrowing used elsewhere in the same function
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Add githubSkillTools to getAllToolDefinitions() so the searchable MCP
catalog matches TOTAL_MCP_TOOL_COUNT, which already counts them. The
GitHub skill tools were registered and counted but missing from the
catalog, so omniroute_tool_search could not surface them.
Adds regression tests at both layers: catalog aggregation and
client-visible discovery via the MCP client.
Co-authored-by: Brandon Bennett <brandonbennett@macbookair.myfiosgateway.com>
* fix(opencode): session stability, free-tier routing, and CLI defaults
- Wire generateSessionId() into opencodeHeaders so x-opencode-session
is a deterministic fingerprint instead of randomUUID() per request,
enabling upstream prompt caching across a conversation
- Thread request body through buildHeaders() so session fingerprint
has access to model, system, messages, and tools
- Default CLI header synthesis to ON (opt-out via false), align
values with 9router proven defaults (opencode/desktop/global)
- Auto-echo listing-valid model names for noAuth providers so
response.model matches /v1/models listing
- Short-circuit free-tier model resolution to opencode provider first
to prevent prefix inference misrouting when catalog is unreachable
* fix(opencode): make free-tier default flip self-consistent + add coverage
PR #10571 flipped OPENCODE_SYNTHESIZE_CLI_HEADERS to on-by-default and
changed the synthesized UA/client/project default values, but shipped
with 2 broken assertions in the existing #5997 regression test and no
coverage for the new session-fingerprinting, free-tier routing, or
noAuth echoModel logic (Hard Rule #18).
- Update tests/unit/opencode-cli-headers-synthesis-5997.test.ts to match
the new on-by-default behavior and new default values; add an explicit
opt-out coverage test so the forward-only path is still guarded.
- Fix 20 further test failures in tests/unit/opencode-executor.test.ts
and tests/unit/refactor-buildHeaders-opencode.test.ts caused by the
same default flip (pin OPENCODE_SYNTHESIZE_CLI_HEADERS=false for the
characterization suites that predate #10571; use a genuinely
CLI-looking UA where the preserved-UA test requires one).
- Fix a real bug found via TDD while adding the mandated free-tier
routing regression test: the big-pickle/*-free short-circuit in
open-sse/services/model.ts checked activeProviders?.has("opencode")
literally, but getActiveProviderSet() canonicalizes every connection's
provider id through resolveProviderAlias(), which rewrites "opencode"
to "opencode-zen" via a manual override — so an active no-auth
opencode connection could never satisfy the check. Now checks both
opencode-family candidate ids. Proven with a test that fails on the
original code and passes with the fix (both connections active with a
stale synced catalog omitting big-pickle).
- Extract the noAuth-provider echoModel aliasing in chatCore.ts into a
pure, directly-testable helper (open-sse/handlers/chatCore/noAuthEchoModel.ts),
matching the existing chatCore god-file decomposition pattern.
- Add regression tests for generateSessionId()-based x-opencode-session
fingerprinting (stable within a conversation, changes on model/message
changes), the free-tier routing short-circuit, and the noAuth echoModel
aliasing.
- Add the changelog.d/ fragment and sync docs/reference/ENVIRONMENT.md's
OPENCODE_SYNTHESIZE_CLI_HEADERS/OPENCODE_USER_AGENT/OPENCODE_CLIENT/
OPENCODE_PROJECT rows to the new defaults.
Does NOT resolve whether flipping OPENCODE_SYNTHESIZE_CLI_HEADERS's
default was the right call, and does NOT touch the separate open PR
#10357 which flips the same flag with a different literal default value
- that decision is left to the maintainer at merge time.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
GET /v1/models listed google/gemini-embedding-001 but omitted
google/gemini-embedding-2 even though that id already returns 3072-d vectors.
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(vision): preserve high detail for inline images
* fix(vision): scope high-detail image default to OpenCode clients
defaultImageDetail() was applied at prepareUpstreamBody, the shared
upstream-body prep path for every provider and format, not just the
OpenCode path the fix targets. Gate it on isOpencodeClient (the
existing User-Agent/x-opencode-* header signal already used for
bypassDefaultToolLimit at this call site) so non-OpenCode callers keep
the provider's own image detail default. Adds a regression test
covering a non-OpenCode caller against the same opencode-zen provider.
* fix(vision): document and test the global vs OpenCode-only detail scope
The OpenCode-only high-detail default in chatCore/upstreamBody.ts
(defaultImageDetail, gated on isOpencodeClient) forwards the caller's
own image_url.detail and was already correctly scoped in a prior
commit on this branch.
The internal vision-bridge describe self-loop (visionBridgeHelpers.ts)
is architecturally global: VisionBridgeGuardrail runs for every
caller/provider whenever the target model lacks vision support, and
there is no client-identity signal at that layer to gate on. Its
describe prompt explicitly asks the vision model to transcribe visible
text, so requesting "high" detail unconditionally is justified on its
own merits (OCR accuracy), independent of the OpenCode motivation.
Adds a compatibility assertion proving the Anthropic wire-format
branch of the same describe self-loop carries no `detail` field (it
has no such concept) and is therefore unaffected by this default, and
documents the split (OpenCode-only forwarding vs. global describe
default) in docs/security/GUARDRAILS.md.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: rinseaid <rinseaid@rinseaid.net>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* feat(providers): optional AI Horde API key and live image catalog
Allow a registered Horde key on the no-auth connection and send it for
chat and image jobs. List only image models that currently have workers,
and generate through Horde's native async API.
# Conflicts:
# open-sse/config/imageRegistry.ts
# src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx
# src/shared/constants/providers.ts
# src/sse/services/auth.ts
* fix(providers): validate AI Horde keys against find_user
The OpenAI-compatible /v1/models probe returns 200 for any Bearer token
on oai.aihorde.net, so Check always succeeded. Use Horde's /v2/find_user
lookup instead; an empty key still counts as the optional anonymous path.
* chore(changelog): name the AI Horde fragment for #10542
* fix(images): harden AI Horde optional-key selection and outbound fetches
- Optional-key selection now honors connection health (rate-limit cooldown
and terminal/unavailable test status) before handing a stored key back,
rotating to the next healthy key or falling back to the anonymous no-auth
path instead of using an unhealthy stored key.
- Route the Horde submit/check/status/cancel and catalog calls through the
repository's bounded outbound-fetch helper (timeout, no more bare fetch())
and route R2 image downloads through the established bounded remote-image
fetch (SSRF host guard, DNS-rebinding pin, streaming byte cap, redirect
limit) instead of an unbounded fetch().
- Extend the generation deadline to cover the full request lifecycle
(catalog freshness check, submit, polling, and image download), and add a
regression test proving that exceeding the deadline issues a DELETE
cancel to Horde's API rather than only timing out locally.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: pqr <pqr@soraka.ititti.es>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* fix(settings,auth): default debugMode to false and skip account rotation on model-unsupported 400
* fix(auth): disambiguate model-unsupported from auth-credential 400
The model-unsupported guard used MODEL_ACCESS_DENIED_PATTERNS directly,
which also matches auth-credential errors like 'invalid api key for
model X'. Add the AUTH_CREDENTIAL_ERROR_PATTERNS exclusion (same as
checkFallbackError) and use provider_model_unsupported log reason.
Addresses maintainer feedback on PR #10525
Signed-off-by: Minxi Hou <houminxi@gmail.com>
* fix(auth): narrow model-unsupported guard to avoid misclassifying account-scoped entitlement 400s
The #10460 guard reused MODEL_ACCESS_DENIED_PATTERNS directly, which also
matches ambiguous "access"/"permission" phrasing (e.g. "does not have
permission to access this model") that commonly signals an ACCOUNT-scoped
entitlement gap (PRO vs free tier) rather than a genuinely provider-wide
unsupported model — a different account of the same provider may still
have access, so those must keep rotating normally instead of being
short-circuited.
Extract isProviderModelUnsupported400() in accountFallback.ts: reuses the
same AUTH_CREDENTIAL_ERROR_PATTERNS exclusion checkFallbackError's 400
branch already applies, narrowed to a strict subset of unambiguous
"provider does not serve this model at all" phrasings. auth.ts now calls
this shared helper instead of testing the broader patterns in isolation,
and exposes the sanitized reason ("provider_model_unsupported") on the
returned result, not just in the log line.
Also fix DATA_DIR test-isolation ordering in
account-fallback-service.test.ts: it was assigned after the first
dynamic import of accountFallback.ts, which transitively imports
src/lib/db/core.ts (DATA_DIR is captured once at module-load time), so
the intended isolated test directory was silently never used. Move the
assignment before any transitive DB import, and add regression tests for
the 3-account rotation contract: exactly one upstream call for an
unambiguous provider-wide 400 with the combo advancing to the next
target, continued rotation for account-scoped 401/403/429 and for the
permission/entitlement 400 case that motivated this narrowing.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Signed-off-by: Minxi Hou <houminxi@gmail.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* fix(combo): surface context-overflow before compression so oversized requests fail fast with a clear error (#10225)
* fix(combo): make context-overflow deferral target-aware for native Codex passthrough (#10225)
The deferral added by the prior commit checked only operator-named
compression exclusions when deciding whether at least one target "can
compress" — it never accounted for native Codex Responses passthrough
targets, which chatCore.ts unconditionally excludes from compression
(compressionExcluded = nativeCodexPassthrough || ...). Deferring on such
a target's account let an oversized request skip both the combo preflight
AND compression, reaching fetch() uncompressed.
Thread the same request-shape facts chatCore.ts uses
(shouldUseNativeCodexPassthrough: provider/sourceFormat/endpointPath/body/
headers) down into getKnownContextOverflow so the deferral decision can
never drift from chatCore's own — a native-codex-passthrough target now
never counts as "compressible", so a pool made only of such targets keeps
the fast local 400 instead of a wasted round trip.
Adds regression coverage: the pure getKnownContextOverflow target-aware
check, an end-to-end handleComboChat proof that a native-codex-only pool
fails fast with zero dispatches, and two real handleChatCore-path tests
proving compression actually reduces the dispatched body when eligible,
and that a still-too-large-after-compression request is rejected locally
without an upstream call.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(resilience): keep combo quality and auth reasons separate and redact connection labels in terminal errors (#10314)
* fix(resilience): sanitize identifiers in error text, add explicit terminal-status policy, fix classifier ordering (#10314)
Four gaps in the prior combo-error-aggregation fix:
- formatComboOutcomes() only redacted connection identifiers in the model
label, never in the raw upstream error TEXT — a proxy echoing a
connection/account id back in its error body leaked it into the
client-facing terminal message. Redact both.
- The terminal HTTP status was still `lastStatus` — whichever target
happened to fail last, independent of the other targets' reasons. Add
resolveComboTerminalStatus(): preserve a 4xx only when every eligible
target's failure is genuinely "the request is invalid" (model-class);
a heterogeneous mix (e.g. a quality failure + a sibling's 401) now
normalizes to a 5xx-class status reflecting an infra/provider problem,
never a misleading client error borrowed from an unrelated target.
- classifyComboOutcome()'s ordering had `status === 408 || status >= 499`
checked before `status >= 500`, making the provider branch permanently
unreachable — every real 5xx (500/502/503/504) was silently mislabeled
as "timeout". Fixed to an exact match (408/499) and gave 429 its own
explicit `rate_limit` kind instead of falling into the generic "model"
(request-invalid) bucket by accident.
- Added an integration-level regression driving the real handleComboChat
wiring end-to-end (quality failure + sibling 401, and a success-after-
quality-failure case), not just the pure aggregation helpers.
Updated three pre-existing tests whose assertions encoded the OLD
last-writer-wins contract this fix intentionally supersedes (#8486 Part B
antigravity retryAfter tests, two combo-routing-engine status/message
tests) to the new, more precise contract; verified the underlying #8486
concern (wrong target's retryAfter header) is still honored under the new
status policy.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(providers): fall back to public Code Suggestions endpoint on GitLab Duo direct_access 401 (#10365)
* fix(providers): extend GitLab Duo 401 fallback to the connection-test path (#10365)
The chat-completion path (open-sse/executors/gitlab.ts) already falls back to
the public Code Suggestions completions endpoint when the direct_access
exchange is rejected with 401, but testOAuthConnection() / the dashboard
Retest button still reported the connection unhealthy on the same 401 —
even though a real chat request through that connection would have
succeeded via the fallback. Apply the identical fallback contract to the
connection-test path (first attempt and the post-refresh retry), sharing the
predicate with the executor via shouldFallbackToPublicCodeSuggestions.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(compression): add i18n support for less-code and terse-prose
Translates less-code output style to pt-BR, vi, ja, and id. Adds missing vi translation to terse-prose caveman mode. Removes less-code from English-only allowlist and updates matrix tests.
Fixes#10426
* docs(compression): add output styles coverage table
Adds the requested Output Styles matrix to the compression guide covering styles, supported languages, and intensity levels.
Fixes#10426
* fix(sse): downgrade adaptive thinking and gate context-1m beta on model eligibility (#10119)
* fix(sse): thread resolved model into DefaultExecutor's anthropic-beta merge (#10119)
DefaultExecutor.buildHeaders() merged the client-negotiated anthropic-beta
header without ever passing the resolved target model into
mergeClientAnthropicBeta(), so the context-1m-2025-08-07 eligibility gate
added earlier in this PR could not see which model a combo/fallback had
actually routed to at this call site. buildHeaders() now accepts an
optional model parameter (mirroring BaseExecutor.buildHeaders' existing
signature and the pattern already used by grok-cli.ts/qoder.ts) and
forwards it through, so an ineligible model target (e.g. Haiku) has the
beta dropped instead of forwarded blind.
Restores a CHANGELOG bullet (PR #10366) that a prior merge auto-resolve
had dropped from this branch.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(admission): resolve adaptive latency-collapse self-lock with solo-progress and idle recovery (#10111)
* fix(admission): refresh recovery ceiling on updateConfig (#10111)
updateConfig() clamped currentLimit to the new min/maxLimit but left
recoveryCeiling pinned to the value computed at construction time, so
a raised initialLimit could never recover past the stale ceiling and
a lowered one could leave the ceiling above the new maxLimit.
Recompute recoveryCeiling from the new initialLimit on every
updateConfig call, clamped to the (possibly also new) min/maxLimit.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(usage): wire agentrouter balance quota into dashboard Quota UI (#10078)
* fix(usage): render AgentRouter wallet balance as USD in the Quota UI (#10078)
The prior fix wired AgentRouter's balance into getUsageForProvider() and
USAGE_SUPPORTED_PROVIDERS, but the actual dollar figure never reached the
Dashboard Quota UI: quotas.balance.remaining carried a synthetic two-state
percent (100/0) instead of the real dollarBalance, and the Provider Limits
renderer only formats a row as "$X.XX" when isCredits/currency/creditCount
are set, which the generic quota-parsing path never sets. A configured
balance rendered as a bare "100% left" percentage, not USD.
Shape quotas.balance.remaining as the real USD amount (clamped to 0) and add
an agentrouter branch to quotaParsing.ts that builds a credits-style row
(same buildCreditsQuota() pattern as DeepSeek/Claude extra-usage), so a
configured balance shows a currency-formatted dollar amount and an
exhausted balance always renders as exactly $0.00.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(providers): resolve combo names on /v1/audio/speech and /v1/videos/generations
`GET /v1/models` advertises combos with `owned_by: combo`, and chat, embeddings,
transcriptions (#9134) and images (#8986, #9239) all resolve those names. Speech
and video did not: both rejected a combo name at model validation, before any
resolution could happen.
POST /v1/audio/speech {"model":"my-combo","input":"hi"}
-> 400 Invalid speech model: my-combo. Use format: provider/model
POST /v1/videos/generations {"model":"my-combo","prompt":"a cube"}
-> 400 Invalid video model: my-combo. Use format: provider/model
A client picking a model out of /v1/models therefore could not tell which
entries the catalogue would actually accept, and callers ended up hardcoding
vendor ids for these two routes while using combo names everywhere else.
Both routes now mirror the images route: detect a combo name before the
provider lookup and divert to a strategy executor. The two new executors follow
imageCombo — expand targets with resolveComboTargets(), filter to targets the
route can actually serve, walk them in priority order, and return the first
success or the last failure, with 400/401/403 treated as terminal.
Two details differ from the image strategy:
Speech filters at model level rather than provider level. parseSpeechModel()
resolves a provider prefix without checking that the model behind it can speak,
so `openai/gpt-4o` would otherwise be accepted as a target and fail only once
dispatched. The filter now checks the provider's own model list, and keeps
targets from dynamic provider nodes that do not enumerate models.
Speech also returns the handler's Response untouched instead of building a JSON
body, because that route streams audio; only the ADD-only meta headers are
attached, exactly as the direct path does. The failure branch is the only place
the body is read.
successfulMediaGenerationResponse() gains optional `strategy` and
`fallbackAttempts` so the video strategy can report them the way imageCombo
does, rather than duplicating the cost calculation. Both are omitted on the
direct single-model path, where neither is meaningful.
Tests mirror tests/unit/combo/image-combo.test.ts for both routes: combo not
found, no capable targets, empty combo, and targets present with no provider
connection. 16/16 pass across the three combo test files.
* fix(providers): preserve local overrides, custom models and per-target prompt rules through video combo dispatch
executeVideoCombo() diverged from the direct /v1/videos/generations route in
three ways: it dropped the ComfyUI-style local-override credential lookup for
authType:"none" targets, its capability filter only matched the built-in
video registry (skipping custom OpenAI-compatible provider nodes tagged with
the "videos" endpoint), and the route validated the prompt against the
unresolved combo name before combo targets were expanded — rejecting
prompt-optional I2V targets that never got the chance to opt out.
Extracts the shared resolution rules (resolveVideoModelTarget,
isVideoPromptOptional, resolveLocalOverrideCredentials) into
src/app/api/v1/_shared/videoModelResolution.ts so the direct route and the
combo executor apply identical rules, moves the combo-name diversion ahead of
the prompt-required check so validation runs against the real resolved
target, and adds per-target prompt validation inside the combo loop so a
missing prompt only rules out that target instead of the whole combo.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* fix(open-sse): stop concurrent requests colliding on dedup hash for non-OpenAI formats
computeRequestHash() in requestDedup.ts projected the prompt content from
body.messages only. The dedup site in chatCore.ts hashes the *translated*
(target-format) request body, and non-OpenAI target formats don't carry a
messages field: Gemini-translated bodies use `contents`, Responses-API
bodies use `input`. So for those formats messages was always undefined,
every prompt hashed to the same null-backed value for a given model, and
concurrent requests with different prompts joined the same in-flight
promise -- the second caller silently received the first caller's
response verbatim (#10249).
Fix: project body.messages ?? body.contents ?? body.input ?? null instead
of only body.messages, keeping the rest of the canonical hash projection
unchanged. Genuinely identical concurrent requests still dedupe (the
intended perf behavior); different prompts under Gemini/Responses-API
target formats no longer collide.
Regression test: tests/unit/request-dedup-10249.test.ts reproduces the
two collision scenarios from the plan-file (Gemini `contents`,
Responses-API `input`), confirms the OpenAI `messages` case was already
correct, and asserts identical-request dedup keeps working. Verified
RED (byte-identical hashes 0b24fd88.../dc16d5b7... pre-fix) -> GREEN
(distinct hashes, dedup preserved) against this exact diff.
* fix(open-sse): cover nested translator shapes + system fields in dedup hash (#10438)
computeRequestHash() only read top-level body.messages ?? body.contents ??
body.input, but several translated request shapes nest their prompt
content: the Antigravity Cloud Code envelope under request.contents, and
Kiro under conversationState.currentMessage.userInputMessage.content (plus
conversationState.history). Two different concurrent prompts to those
targets could hash identically and share/leak a response between callers.
Adds extractPromptContent()/extractSystemContent() helpers covering every
prompt-bearing shape produced by open-sse/translator/request/*.ts
(OpenAI/Cursor messages, Claude messages+system, Gemini contents+
systemInstruction, Responses input+instructions, Antigravity and Kiro
nesting), and folds system/instructions/systemInstruction into the
canonical hash so two requests with the same user message but a different
system prompt no longer collide either.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(antigravity): strip trailing model turn for native Gemini requests too
Newer Gemini endpoints reject a request ending on a model turn with HTTP
400 'Requests ending with a model turn are not supported' — the same
rejection class Claude hits via Vertex. transformRequest() previously
wired stripTrailingAntigravityAssistantTurn() only into the isClaude
branch, so native Gemini models routed through Antigravity kept a
trailing role:model entry and hit the 400.
Extend the guarded strip (never empties contents) to native Gemini
models too, gated by upstreamModel including "gemini". The Claude
path is untouched (byte-identical), preserving PR #6114's live
validation against Vertex Claude.
Flips tests/unit/antigravity-claude-prefill-strip.test.ts test (b),
which previously asserted the buggy pass-through, and adds (b2) for
the gemini-3-flash-agent tier.
Closes#10104
* fix(antigravity): scope Gemini trailing-turn workaround
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
* fix(translator): Normalize tool call names from lowercase to PascalCase (#1)
* Fix: Map lowercase tool names from Antigravity (Gemini format) to Claude Code expected PascalCase
* Fix: toolNameMap in fun restoreClaudePassthroughToolUseName
* fix(translator): Normalize tool call names from lowercase to PascalCase when translating upstream responses (OpenAI, Gemini, Antigravity) to Claude Messages API format
This resolves `Error: No such tool available: read`/`bash`/`write` errors when using Claude Code CLI with third-party providers that emit lowercase tool names. The fix adds case-insensitive tool name lookups in `openai-to-claude.ts`, `gemini-to-claude.ts`, and related translators, ensuring tool names like `read`/`bash` are mapped to `Read`/`Bash` before being sent to Claude Code. Includes unit tests and comprehensive changelog notes ([#10250](https://github.com/diegosouzapw/OmniRoute/pull/10250))
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(translator): Parse <tool_call> JSON and TOOL_CALL text formats fr… (#2)
* fix(translator): Parse <tool_call> JSON and TOOL_CALL text formats from model output
Some models (DeepSeek, Qwen) emit tool calls as text instead of proper
tool_calls JSON: either <tool_call>{...}</tool_call> or TOOL_CALL Name: {...}.
Extend extractXmlInvokeBlocks to handle all 3 formats in a single scan pass,
picking whichever pattern appears first. Includes unit tests for all formats.
* fix(translator): Parse text-format tool calls in gemini-to-claude translator
Extend the Gemini->Claude translator to detect <invoke>, <tool_call> JSON,
and TOOL_CALL text formats emitted inline in text parts (Antigravity/Gemini
models), converting them to proper tool_use content blocks instead of leaking
raw text to Claude Code.
* docs(changelog): Add changelog entry for text tool call parsing fix
* fix(translator): consolidate tool name casing normalization and restore thought-signature persistence (#3)
* fix(translator): sanitize tool_use.id and tool_result.tool_use_id to match Anthropic schema (#4)
Ensure tool IDs from OpenAI-compatible upstreams (which may contain dots, colons, or special characters) are sanitized to ^[a-zA-Z0-9_-]+$ in response translators and passthrough requests before reaching Claude endpoints.
* fix(responses): preserve native tools for openai-compatible Responses targets (#5)
A Responses-shaped request to a custom openai-compatible connection whose
outbound protocol is Responses took a Responses -> Chat -> Responses round
trip, so Codex custom tools lost their grammar (`exec`), namespace groups were
flattened (`collaboration`), and tool invocations failed upstream.
Gate a native Responses passthrough on the connection's configured protocol
(`apiType: "responses"` / `_omnirouteForceResponsesUpstream`) so the original
tool definitions reach a Responses-capable upstream unchanged. Chat-only
connections keep the existing downgrade.
Closes#10374
* fix(translator): add support for 'applypatch' tool name in tool call checks
* test(translator): add unit test for apply_patch and applypatch tool name remapping
* fix(translator): remove no-explicit-any lint errors in tool-use-id-sanitization test
Type the openaiToClaudeResponse/translateNonStreamingResponse return
values with narrow local shapes instead of `any`, satisfying the
repo's no-explicit-any = error rule for tests/. No behavior change —
the same 3 assertions still pass.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* test: update 9568 casing regression to match #10392's consolidated fix
restoreClaudeToolName's static casing map now normalizes known
lowercase tool names to canonical PascalCase unconditionally on the
gemini-to-claude and openai-to-claude Claude Messages API paths (not
gated behind toolNameMap), superseding the earlier per-map-only fix
that the original #9568 regression test locked in as "expected" (it
was previously labeled a known bug case). The gemini-to-openai
passthrough path is unaffected by #10392 and keeps its original
pass-through assertion.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Family resolves like auto/zai with no connected models logged a warn
on every call (about once a minute per poll). Keep the empty-pool
behavior; emit the warn at most once per label per 60s.
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* feat(oauth): add gemini-3.7-flash models with reasoning tiers for antigravity
Support gemini-3.7-flash and its thinking tiers (low/medium/high) for antigravity and agy providers.
- Define public models, pricing, modelSpecs, and CLI tool definitions
- Map tiers to live upstream id gemini-3.7-flash-tiered
- Configure defaultThinkingBudget (low: 1024, medium: 8192, high: 32768)
- Allow executor fallback on upstream 404 and 5xx errors
- Add unit tests in antigravity-model-aliases.test.ts
* fix(oauth): expose gemini-3.7-flash as one callable antigravity/agy model
Upstream (fetchAvailableModels on daily-cloudcode-pa) only accepts the single
upstream id gemini-3.7-flash-tiered; the high/medium/low suffixed tier ids
404. Registering all four as distinct public model ids violates the base
#3696 uniqueness invariant (no two ANTIGRAVITY_PUBLIC_MODELS entries may
resolve to the same upstream id). Collapse to the single live gemini-3.7-flash
public model (aliased to gemini-3.7-flash-tiered) and drop the tiered specs,
pricing, free-catalog and CLI entries accordingly, keeping the leading public
model order (Gemini 3.6 tiers first) intact.
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
Co-authored-by: Chewji9875 <Chewji9875@users.noreply.github.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* remove: drop sunset MiMoCode provider from model catalog
* remove: drop sunset MiMoCode provider from model catalog (shared.ts)
Remove unused imports, types, and comments from shared.ts.
* remove: MiMoCode provider (Xiaomi sunset) — executor, registry, no-auth config, icon, tests
* refactor(providers): finish MiMoCode removal — sweep remaining no-auth references
Drop the leftover mimocode entries from the no-auth provider controls, the
translate-path snapshot, the eslint suppressions, and the #3061 auth-loop
test. Re-point the fingerprint-pin (#6696) and proxy-noauth (#6272) tests at
opencode, which exercises the same fingerprint path, so the removal does not
break runtime behavior.
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* docs(providers): reconcile provider/executor counts after MiMoCode sunset
The base's parallel doc-count sync (#10433) pinned 340 providers / 101
executors. With mimocode removed, live code has 339 providers and 100
executors; refresh the user-facing counts (package.json description,
llm.txt, README/AGENTS, i18n llm.txt, provider reference, diagrams) so the
check-docs-counts STRICT gate stays green.
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* test(providers): fix orphaned mimocode references after MiMoCode sunset
The sunset removed mimocode/mcode from the free-onboarding candidates and
from FINGERPRINT_PROVIDERS, but two tests still referenced them:
- free-provider-onboarding-setup: the mimocode->theoldllm substitution
introduced duplicate 'opencode' rows (impossible given the request-set
dedupe) and the wrong display name; align expectations with the actual
{opencode, theoldllm} dedupe behavior and 'The Old LLM (Free)' name.
- combo-system-prompt-templates-5501: resolveTargetFingerprint tested with
provider 'mcode', which is no longer a fingerprint provider; point it at
the remaining fingerprint provider 'opencode'.
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
---------
Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
Co-authored-by: Tushar49 <Tushar49@users.noreply.github.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* feat(quota): Phase 2 adapters, reset timers, analytics, and dashboard API
* feat(routing): add quota-aware provider scheduling (opt-in)
* fix(db): rename migration to 148_provider_quota_state.sql
* fix(quota): harden quota state route, isolate phase2 tests, slim env diff
- route: requireManagementAuth + Zod body validation + buildErrorBody
sanitization (Hard Rule #12); fix clearProviderQuotaState -> clearProviderQuota
- .env.example/ENVIRONMENT.md: drop ~20 foreign vars, keep only
OMNIROUTE_QUOTA_AWARE_ROUTING (migration 148)
- tests/unit/quota-phase2.test.ts: DATA_DIR mkdtemp + resetDbInstance teardown
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* chore(ci): fix docs-sync + eslint-suppression drift for quota branch
CI gates flagged on PR #10126 head 43335f07:
- migration counts in README/AGENTS/llm.txt were stale (145 -> 146)
- regenerate docs/reference/PROVIDER_REFERENCE.md (gen-provider-reference)
- sync root llm.txt body into all 42 i18n mirrors (headers preserved)
- prune eslint suppressions that no longer occur
--no-verify: pre-commit docs-sync was failing on a pre-existing
release-base artifact (changelog 3.8.49 vs package 3.8.50) — fixed by
the changelog entry in the prior commit; re-verify in CI.
* chore(skills): regenerate agent skills (add omni-settings)
Merge-integrity CI gate flagged a missing generated skill. Regenerated
with check:agent-skills-sync --apply: +omni-settings, 45 unchanged.
* fix(ci): resolve Fast Quality Gates regressions on quota branch
- check-migration-numbering: migration 148 (provider_quota_state) landed
on this branch, so the KNOWN_GAPS allowlist entry is stale — remove it
(stale-enforcement 6A.3: 'REMOVA a entrada')
- open-sse/utils/stream.ts: duplicate sseCommentsEnabled import from a
bad merge (lines 31 + 77) — TS2300 duplicate identifier; drop the
duplicate so the open-sse typecheck gate is back within baseline
* docs: sync migration count to 149 after release merge
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* test(migrations): align 148 gap assertion after 148_provider_quota_state.sql landed
The phase-2 branch added 148_provider_quota_state.sql, and 148 was already
removed from KNOWN_GAPS in scripts/check/check-migration-numbering.mjs. The
frozen-allowlists assertion still expected 148 to be a gap, so it failed.
Flip the assertion to match the allowlist (same pattern as 143/147).
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
---------
Co-authored-by: benzntech <benzntech@users.noreply.github.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* fix(adobe-firefly): open browser sign-in and resolve provider slug in /login
POST /api/providers/[id]/login passed the connection DB id to
inAppLoginService.startLogin, but that service looks up the provider by
slug in TOKEN_EXTRACTION_CONFIGS. The lookup always missed and returned
"No extraction config" without launching a browser — so the VibeProxy
"Sign in" button for Adobe Firefly (and every other web-cookie provider)
never opened a browser.
Adobe Firefly additionally had no extraction config because its IMS JWT
is never in cookies/localStorage — it only rides on the Authorization:
Bearer header of firefly-3p.ff.adobe.io XHRs.
- Resolve the provider slug from the connection row and pass the slug
(not the DB id) to inAppLoginService.startLogin.
- Add open-sse/services/adobeFireflyBrowserLogin.ts: a Playwright
service that launches a visible browser at firefly.adobe.com and
intercepts firefly-3p requests to capture the IMS JWT + sherlockToken
cookie. Wire it into the /login route for the adobe-firefly slug.
- Fix latent bug: updateProviderConnection reads camelCase keys
(apiKey, providerSpecificData), so the previous snake_case call never
persisted extracted credentials.
* fix(adobe-firefly): open browser sign-in and resolve provider slug in /login
POST /api/providers/[id]/login passed the connection DB id to
inAppLoginService.startLogin, but TOKEN_EXTRACTION_CONFIGS is keyed by
provider slug — so browser login never launched for web-cookie providers.
Adobe Firefly also cannot use cookie extraction: the IMS JWT only appears
on Authorization headers to firefly-3p.ff.adobe.io. Add a dedicated
Playwright interceptor and persist credentials with camelCase keys that
updateProviderConnection actually reads.
* fix(adobe-firefly): use system Chrome/Edge CDP for browser sign-in
Playwright is not available inside the pkg-packaged VibeProxyServices.exe,
so import('playwright') always failed with 'Playwright not installed' and
never opened a window. Launch Chrome/Edge with --remote-debugging-port and
capture the firefly-3p Authorization Bearer via pure CDP WebSocket instead.
* fix(adobe-firefly): live x-arp-session-id / Arkose wire (stop 408 under load)
Browser generate-async requires x-arp-session-id as base64({sid,ark,ftr}) with a
real Arkose blob (sherlockToken). JWT alone frequently returns colligo HTTP 408
system under load while credits still work.
- Match live ftr magic __UDF43-m4_31ck + Arkose pk in synthetic ARP fallback
- Ranked extract of sherlockToken / x-arp from Cookie, HAR, fetch() paste, and
space-joined JWT+ARP (PasswordBox newline collapse)
- Reuse one ARP for storage upload + generate-async
- Clearer 408 errors when browser ARP is missing vs stale
- Unit suite 42/42
* fix(adobe-firefly): durable session ARP rebuild and aux_sid false-positive
Rebuild x-arp-session-id from forterToken/arkose/ff_session_guid instead of
ranking long Cookie pairs (e.g. aux_sid=…) as opaque ARP, which caused colligo
HTTP 408. Cache IMS JWT + cookie sessions, rotate ARP on 408 retries, and keep
Playwright warm-up opt-in only (headless Forter is rejected).
Also expand synthetic ARP shape with bfp/fpjs to match live successful captures.
* fix(adobe-firefly): durable session, off-screen Chrome recovery, browser sign-in
Rebuild x-arp-session-id from Cookie pieces (sid/ark/forter) so aux_sid is never
sent as ARP. Sticky ARP + submit spacing reduce mid-batch colligo 408 thrash.
Add optional managed Chrome warm (off-screen headed by default; Forter rejects
headless) and POST /api/providers/{id}/login browser sign-in that returns JWT+Cookie
after a fresh SSO. Visible sign-in resets off-screen window placement and clears
prior Adobe session when adding another account.
* fix(adobe-firefly): renew sessions through durable CDP
* fix(adobe-firefly): isolate browser sessions per account
* fix(adobe-firefly): make account login fresh and deterministic
* chore(adobe-firefly): remove obsolete browser fallback
* docs(adobe-firefly): document renewal controls
---------
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
The model ships only as `qwen3.8-max-preview` across every provider that serves it (bailian-coding-plan, qoder, qwen-cloud-token-plan, qwen-web), so the bare `qwen3.8-max` missed MODEL_SPECS: the chatCore context preflight fell back to contextManager's `default: 128000` and rejected prompts with `context_length_exceeded` despite the model's real 1M window, and the unknown id would have reached the upstream verbatim.
Both symptoms share one cause, so the alias goes in BUILT_IN_ALIASES, which resolveLifecycle() applies before the preflight and before dispatch.
Merged with the inherited OmniGlyph base-red (#9985) documented: its two failing compression tests were reproduced on the pure base tip aa912c42a7, with no commit from this branch.
* fix(usage): read Gemini usageMetadata out of the antigravity response envelope
Port decolua/9router#59d858b: antigravity/gemini-cli wrap non-streaming
payloads in { response: {...} }, so extractUsageFromResponse only saw the
top-level usageMetadata and every non-streaming antigravity request logged
zero usage (IN 0 | OUT 0) and zeroed usage-dashboard rows. Top-level
metadata keeps priority; OpenAI/Claude branches untouched.
* chore(changelog): fragment for #10430 antigravity usage envelope
* fix(usage): surface Gemini cachedContentTokenCount as cached_tokens
Review follow-up on #10430: the Gemini branch of extractUsageFromResponse
ignored cachedContentTokenCount, so non-streaming cache-hit tokens never
reached the cached_tokens field the OpenAI/Claude/Responses branches
already populate (and the streaming path surfaces at usageTracking.ts:684).
Adds cached_tokens: usageMetadata.cachedContentTokenCount || 0, updates
the three Gemini assertions (envelope fixture already carried
cachedContentTokenCount: 7), and adds a dedicated regression test.
* chore(changelog): fragment for #10465 Gemini cached_tokens surfacing
* test(fix): refresh expired alibaba quota sample validity and onnxruntime pin for v3.8.50 base
- alibaba-free-tier-quota-fetcher.test.ts: sample quotaValidityPeriod
(2026-08-16 16:00 UTC) is in the past, making every quota entry classify
as expired/not_capable; bump to 2028-01-01 UTC so the text/merge
classification tests exercise the intended path again.
- optional-transformers-dependency.test.ts: onnxruntime-node pin assertion
updated from ~1.24.3 to ~1.27.0 to match package.json (bumped by #10403);
the regular-not-optional intent is unchanged.
---------
Co-authored-by: Rouzbeh <rqzbeh@users.noreply.github.com>
* fix(antigravity): heal empty-projectId accounts via retryable auto-onboarding
Accounts with an empty Cloud Code projectId get a permanent 422 "Missing
Google projectId" when loadCodeAssist returns no project. The 3.8.50
bootstrap attempts to CREATE the project via onboardUser, but a single failed
attempt (transient network/upstream error) was memoized forever in
onboardAttemptedCache: every later request in the process skipped onboarding
and 422'd, even though a retry would succeed.
Replace the permanent per-token Set with a failure-backoff map: failed onboard
attempts are retried after a 5-minute backoff (bounded, self-healing), the
in-flight lock still dedupes concurrent calls, and success clears the failure
marker and memoizes the project as before. Accounts that CAN be onboarded now
heal automatically on a later request or token refresh — no user action.
Tests: the existing "does not retry" case is now framed as the backoff window;
a new case proves the account heals (retries onboarding and recovers the
project) once the backoff expires.
* chore(changelog): fragment for #10424 antigravity project autocreate
* feat(antigravity): BYOP fast-fail + manual GCP project-id override
Port decolua/9router#2934 + VansRouter 802a859:
- tryOnboardUser now returns a three-way status; a 200 onboardUser response
WITHOUT cloudaicompanionProject means Google deprecated automatic project
creation for standard-tier (personal) accounts (BYOP). Such accounts are
cached permanently (no pointless ~18s re-onboard) and the executor fails
fast with 403 GCP_PROJECT_REQUIRED + actionable 'enter your project id'
message instead of the generic 422 or a delayed 429.
- Transient onboard failures keep the existing 5-min backoff heal.
- Manual project-id override: the EditConnectionModal now stamps
providerSpecificData.isProjectIdManual when the operator enters a project
id, and tokenRefresh skips auto-discovery for flagged accounts so the
manual value is never overwritten.
* chore(changelog): cover BYOP fast-fail + manual override in #10424 fragment
* test(antigravity): expect fast 403 GCP_PROJECT_REQUIRED when loadCodeAssist finds no project (#10424)
Google now marks accounts without an onboarded project as BYOP (automatic
project creation deprecated for standard-tier accounts, #2934). The PR's
BYOP fast-fail path returns 403 gcp_project_required instead of the old
generic 422 missing_project_id; align the #2334 executor test with that
contract so CI unit-test shard 2/4 passes.
* fix(antigravity): persist isProjectIdManual, fix BYOP citation, dodge refresh-retry
Review follow-up on #10424:
1. EditConnectionModal: isProjectIdManual was set on
updates.providerSpecificData right after the project-id field, then the
OAuth path (Antigravity is always OAuth) rebuilt providerSpecificData from
connection.providerSpecificData before the request went out, discarding the
flag — tokenRefresh.ts was guarding a field never actually persisted. The
flag now lands in the single surviving antigravity merge, with a jsdom
regression test (modeled on edit-connection-modal-openai-store-toggle).
2. The '#2934' citation for the Google BYOP claim pointed at an unrelated
closed issue. Swapped for the real tracking issue #8491 (empty Google
projectId -> 422 class) across bootstrap/executor/test comments.
3. BYOP fast-fail now returns 422 instead of 403: chatCore's generic
401/403 -> refresh-and-retry path was hitting Google's OAuth token
endpoint on every request from an affected account (pointless — refreshing
cannot create a GCP project), and 422 matches the sibling
missing_project_id error the client already maps to an action-needed
prompt.
Also: eslint-disable-next-line for the pre-existing
react-hooks/set-state-in-effect baseline noise in the modal (repo
convention, same pattern as 11 other dashboard files).
* chore(ci): drop unused eslint-disable in EditConnectionModal form hydration
The react-hooks/set-state-in-effect disable added in the previous commit is
unused under the repo's pinned eslint-plugin-react-hooks (7.0.1) — the rule
does not fire on this line at that version, so the unused directive tripped
the whole-repo 'No new ESLint warnings' gate (max-warnings 0). Verified with
the lockfile-pinned plugin: lint:json is clean (0 errors, 0 warnings).
* fix(build): bound and retry the opencode-plugin npm install in prepublish
The plugin's node_modules is gitignored, so every fresh CI checkout runs a
full npm install inside @omniroute/opencode-plugin during build:cli. npm's
unbounded fetch retries turn a stalled registry CDN connection (the recurring
onnxruntime-class ETIMEDOUT flake) into a 20-30 minute hang — the DAST
'Build CLI bundle' step has been cancelled at the 30m cap repeatedly.
- Bound npm fetch: --fetch-timeout 60s, 2 retries with capped backoff — a
stalled connection now fails fast instead of hanging the job.
- Retry the install up to 3 times with a 10s pause between attempts, so
transient CDN failures recover in-build.
Net effect: the step either completes (network OK) or fails quickly with a
clear error (network down) — it can no longer eat the whole job budget.
* ci(quality): use the npm-ci-retry action on every install step
Fast Quality Gates failed on the recurring onnxruntime-node postinstall
ETIMEDOUT (Microsoft CDN 150.171.x.x) - the same transient flake that has
hit Vitest and dast-smoke today. Only the Build job used the retry action;
the other five jobs (Docs, Fast Quality Gates, Vitest, Unit Tests,
changelog) still ran a bare install and die on any CDN hiccup. Use the
existing retry action (3 attempts, exponential backoff) on every install
step for consistency.
* Merge branch 'release/v3.8.50' into fix/antigravity-project-autocreate
* test(fix): refresh expired alibaba quota sample validity and onnxruntime pin for v3.8.50 base
- alibaba-free-tier-quota-fetcher.test.ts: sample quotaValidityPeriod
(2026-08-16 16:00 UTC) is in the past, making every quota entry classify
as expired/not_capable; bump to 2028-01-01 UTC so the text/merge
classification tests exercise the intended path again.
- optional-transformers-dependency.test.ts: onnxruntime-node pin assertion
updated from ~1.24.3 to ~1.27.0 to match package.json (bumped by #10403);
the regular-not-optional intent is unchanged.
* test(fix): widen modelsDevSync lastSync wait from 200ms default to 2000ms
The truthy-spellings loop asserted each enabled case completes its first
fetch within waitFor's 200ms default timeout, which trips under CI runner
load (observed on PR 10424 shard 2/4). Match the file's other lastSync
waits (2000ms) so the sync-completion assertion is load-tolerant.
---------
Co-authored-by: Rouzbeh <rqzbeh@users.noreply.github.com>
* fix(deps): bump nanoid, dompurify for 2 new Dependabot alerts (#189, #190)
Bumps: nanoid ^3.3.17 (was transitive, now overridden), dompurify ^3.4.13
(with monaco-editor scoped override). Closes Dependabot #189, #190.
Remaining #182-#188 (js-yaml + mermaid) already closed by #9651 merge —
awaiting Dependabot re-scan.
npm audit → 0 vulnerabilities.
* fix(repo): harden .gitignore to also ignore a _tasks symlink (/_tasks)
_tasks is a SEPARATE nested git repo (gitignored). The pattern _tasks/ (trailing
slash) ignores only a directory, not a SYMLINK named _tasks. A self-referential
_tasks symlink can slip in via git add -A and, once pulled, checkout materializes
it over the real _tasks repo (destroying plans/specs/hands-off). Anchored /_tasks
ignores the symlink too, preventing re-capture.
* Hide health-check excluded models from /v1/models catalog (#10026)
Mirror the request-time exclusion rule (provider_specific_data.excludedModels)
in the unified catalog builder: a model is hidden when its provider has
connections but none of them is eligible for it. Applied across the
PROVIDER_MODELS, synced, custom, alias-backed, and managed-fallback loops
so ghost models no longer appear as available.
Co-authored-by: ritheshcn25 <ritheshcn25@users.noreply.github.com>
* fix(models): memoize getModelsDevPricing (event loop / healthz) (#10055)
* fix(models): memoize getModelsDevPricing for /v1/models catalog
resolveCatalogPricing called getModelsDevPricing once per model while
building GET /v1/models. Each call re-scanned models_dev_pricing and
JSON.parsed every row (~10k SQL scans + multi-GB parse work), pegging
the event loop so even /healthz timed out (#9685, #10052).
Memoize the parsed map until saveModelsDevPricing / clearModelsDevPricing
and add a unit test for invalidation.
Signed-off-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
* fix(db): invalidate modelsDevPricing cache on DB reset (#10055)
Copilot review fixes:
1. Register invalidateModelsDevPricingCache() with DB state reset system
so resetDbInstance() clears the process-local memo, preventing stale
pricing data from surviving across DB reset/restore operations.
2. Add test assertion verifying DB reset bypasses the memo (Copilot #10055).
The process-local memo at modelsDevSync.ts:204 caches getModelsDevPricing()
results until saveModelsDevPricing()/clearModelsDevPricing() to avoid
re-scanning all pricing rows on every /v1/models request. Without this hook,
backup restore and test DB resets would serve stale cached data from the
previous connection.
Tests: npm run test:unit:serial -- tests/unit/modelsDevSync-extended.test.ts
---------
Signed-off-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* fix(sse): keep Codex quota headers under the forwarding budget
The 768-byte cap plus priority-3 for any name that does not contain
"ratelimit" dropped x-codex-*-used-percent / reset / credits on every
stream. x-codex-turn-state (314 bytes) ate the budget. Raise the cap,
treat Codex quota headers as rate-limit priority, and do not forward
turn-state.
---------
Signed-off-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouzapw@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>
Co-authored-by: ritheshcn25 <rithesh.chandran@snb.ca>
Co-authored-by: ritheshcn25 <ritheshcn25@users.noreply.github.com>
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* feat(responses): virtualize previous_response_id continuation regardless of upstream support
OmniRoute now exposes OpenAI-compatible previous_response_id/store
continuation to clients unconditionally, even when the selected upstream
provider has no native Responses-API state support. Reconstruction happens
server-side in handleChatImplementation, before any downstream validation
or provider translation: OmniRoute resolves the response id back to the
full input/output it previously produced, prepends it to the client's
delta, and forwards the full reconstructed history upstream exactly as it
does today. Client<->OmniRoute traffic shrinks to the new delta only;
OmniRoute<->provider traffic is unchanged.
Storage reuses the existing call-log pipeline artifact (already gated by
call_log_pipeline_enabled, already retained/cleaned up by the existing
call-log lifecycle) instead of duplicating conversation content into a
second store -- only a lightweight call_logs.response_id index is new.
Every lookup is scoped by api_key_id so one client can never resolve
another client's stored conversation, and any unresolvable/missing/
size-limit-omitted state fails closed with OpenAI's own
previous_response_not_found contract.
Stacked on feat/openai-responses-store-toggle (#10121).
* fix(db): re-export responsesContinuationStore from the localDb barrel
check-db-rules requires every db/ module to be re-exported (or explicitly
allowlisted as intentionally-internal) for discoverability. Missed this
when the module was first added.
* fix(db): renumber previous_response_id index migration to 154
The migration was numbered 153, but release/v3.8.50 already carries
153_radar_local_model_state.sql. The emngrating runner's collision guard
throws on two live .sql files sharing a numeric prefix, so the refreshed
merge would fail DB startup. Renumber to the next free slot (154).
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* docs(db): sync migration count to 149 across llm.txt mirrors
The responses-continuation store adds one migration, so the docs'
migration count is now 149 (was 148). Update README/AGENTS/llm.txt and
regenerate the i18n llm.txt mirrors to keep check:docs-all green.
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* fix(responses-continuation): respect preserve mode, drop dead export
- Un-export ResponsesContinuationState: it's never imported outside
responsesContinuationStore.ts, its own defining file. Fixes the
check:dead-code regression (410 > baseline 409).
- Scope the previous_response_id virtualization interception in chat.ts to
skip entirely when responsesPreviousResponseIdMode=preserve. The
interception ran unconditionally before target/connection selection,
ahead of applyResponsesPreviousResponseIdPolicy (chatCore.ts) -- the
existing per-target enforcement point for this setting -- so "preserve"
(the explicit, connection-independent contract for "let the upstream
resolve previous_response_id natively") was silently unreachable: the
field was already deleted and replaced with locally-reconstructed input
by the time that policy ran. This also broke Codex's own executor, which
relies on an untouched previous_response_id to delegate history
resolution upstream (see stripOrphanedCodexFunctionCallOutputs in
codex.ts). "auto" and "strip" modes are unaffected -- virtualization is
a strict improvement over their old "drop the field, hope the client
resent everything" behavior.
- Add a regression test exercising the actual chat.ts handler (not just
the policy helper in isolation): confirms mode=preserve now proceeds to
normal routing instead of the virtualization's previous_response_not_found
rejection, and that default/auto mode's existing virtualization behavior
is unchanged. Verified the test fails for the right reason against
pre-fix chat.ts.
Addresses PR review feedback.
---------
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
Co-authored-by: hartmark <hartmark@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
* fix(deps): bump nanoid, dompurify for 2 new Dependabot alerts (#189, #190)
Bumps: nanoid ^3.3.17 (was transitive, now overridden), dompurify ^3.4.13
(with monaco-editor scoped override). Closes Dependabot #189, #190.
Remaining #182-#188 (js-yaml + mermaid) already closed by #9651 merge —
awaiting Dependabot re-scan.
npm audit → 0 vulnerabilities.
* fix(repo): harden .gitignore to also ignore a _tasks symlink (/_tasks)
_tasks is a SEPARATE nested git repo (gitignored). The pattern _tasks/ (trailing
slash) ignores only a directory, not a SYMLINK named _tasks. A self-referential
_tasks symlink can slip in via git add -A and, once pulled, checkout materializes
it over the real _tasks repo (destroying plans/specs/hands-off). Anchored /_tasks
ignores the symlink too, preventing re-capture.
* Hide health-check excluded models from /v1/models catalog (#10026)
Mirror the request-time exclusion rule (provider_specific_data.excludedModels)
in the unified catalog builder: a model is hidden when its provider has
connections but none of them is eligible for it. Applied across the
PROVIDER_MODELS, synced, custom, alias-backed, and managed-fallback loops
so ghost models no longer appear as available.
Co-authored-by: ritheshcn25 <ritheshcn25@users.noreply.github.com>
* fix(models): memoize getModelsDevPricing (event loop / healthz) (#10055)
* fix(models): memoize getModelsDevPricing for /v1/models catalog
resolveCatalogPricing called getModelsDevPricing once per model while
building GET /v1/models. Each call re-scanned models_dev_pricing and
JSON.parsed every row (~10k SQL scans + multi-GB parse work), pegging
the event loop so even /healthz timed out (#9685, #10052).
Memoize the parsed map until saveModelsDevPricing / clearModelsDevPricing
and add a unit test for invalidation.
Signed-off-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
* fix(db): invalidate modelsDevPricing cache on DB reset (#10055)
Copilot review fixes:
1. Register invalidateModelsDevPricingCache() with DB state reset system
so resetDbInstance() clears the process-local memo, preventing stale
pricing data from surviving across DB reset/restore operations.
2. Add test assertion verifying DB reset bypasses the memo (Copilot #10055).
The process-local memo at modelsDevSync.ts:204 caches getModelsDevPricing()
results until saveModelsDevPricing()/clearModelsDevPricing() to avoid
re-scanning all pricing rows on every /v1/models request. Without this hook,
backup restore and test DB resets would serve stale cached data from the
previous connection.
Tests: npm run test:unit:serial -- tests/unit/modelsDevSync-extended.test.ts
---------
Signed-off-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* fix(providers): remove invalid CodeBuddy CN glm-4.7 and add hy3 (0.0x credit)
Swap the GLM-4.7 model for the Hunyuan hy3 model in the codebuddy-cn
registry, keep the catalog at 15 models, and update the matching provider
test expectations. Regenerated the auto-generated provider reference doc.
---------
Signed-off-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouzapw@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>
Co-authored-by: ritheshcn25 <rithesh.chandran@snb.ca>
Co-authored-by: ritheshcn25 <ritheshcn25@users.noreply.github.com>
Co-authored-by: Ravi Tharuma <25951435+RaviTharuma@users.noreply.github.com>
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Keep the #8350 Hermes system-prompt drops, but remove hermes from the
factory obfuscate_words list so hostnames and CLI mentions stay intact.
Co-authored-by: Ravi Tharuma <RaviTharuma@users.noreply.github.com>
* fix(mcp): dynamically generate web search provider enum from registry
* test(mcp): add contract test for dynamic web search provider enum
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
* fix(mcp): restore search.ts eslint suppression, type builder maps, fix firecrawl searchType arg
The enum-dynamic refactor dropped search.ts's no-explicit-any suppression
while a new Record<string,any> map re-introduced anys, and the response
normalizer map swapped the firecrawl searchType argument with query. Type
both maps explicitly, restore the base suppression (33 pre-existing anys),
and pass searchType (not query) to normalizeFirecrawlSearchResponse.
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
---------
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
Co-authored-by: adevwithpurpose <adevwithpurpose@users.noreply.github.com>
Co-authored-by: sadSanta-07 <sadSanta-07@users.noreply.github.com>
Co-authored-by: diegosouzapw <diegosouza.pw@gmail.com>
Issue #10443: when the upstream kills an SSE stream mid-generation
(antigravity/Gemini does this under its own rate enforcement), OmniRoute
closed the stream silently for OpenAI-format clients - HTTP 200, a few
content chunks, no finish_reason. The client sees a truncated turn.
resolveSilentCloseReason() only flagged that shape for Claude clients
(#7699). Extend it to OpenAI chat completions guarded on sawContent(),
and teach hasClientTerminalSseMarker() that a non-null finish_reason
chunk is a terminal marker (some providers omit data: [DONE]). Every
known OpenAI-producing path ends with one of the two, so content
forwarded without either is an upstream drop and now surfaces the
in-band 502 error chunk + [DONE] instead of a silent close.
TDD: tests/unit/silent-sse-close-openai-10443.test.ts - core case RED
before / GREEN after, plus guard cases for finish_reason-only close,
[DONE] close, empty-content (#8649 verdict preserved), and literal
finish_reason text inside model content (JSON escaping keeps the raw
bytes from matching the unescaped-field regex).
Signed-off-by: Minxi Hou <houminxi@gmail.com>
Stage 7 of issue #10321 moves the optional ML and browser automation dependency closures out of the desktop bundle into checksummed, versioned packs installed on demand through the omniroute packs command.
- scripts/build/optionalPackStaging.mjs stages pack members under .build/optional-packs, creates release tarballs, and emits optional-packs.index.json with per-member SHA-256 checksums.
- scripts/packs provides manifest, install, remove, and verification helpers plus the packs CLI commands.
- Runtime lookup includes installed pack node_modules directories, while LLMLingua and browser executors continue to degrade gracefully when packs are absent.
The measured darwin-arm64 staging closure was about 534 MB of the 929 MB standalone node_modules tree (57%).
