Repository navigation
Conversation
…aper-inference#3481) Codex's model-catalog refresh (codex_models_manager) does GET /v1/models?client_version=<v> and decodes a JSON object with a TOP-LEVEL `models` array. OmniRoute answers in the OpenAI-standard `{object,data}` shape, so codex fails with "missing field `models`" and logs "failed to refresh available models" on every startup. Detect codex clients via the `originator` / `user-agent` = `codex_*` headers they send and add an EMPTY top-level `models: []` so the decode succeeds. Non-codex OpenAI clients keep the byte-identical `{object,data}` response. The array is intentionally empty: codex replaces its built-in per-model agent prompt (`base_instructions`, ~21k chars) with whatever a populated entry carries for the selected model, so emitting our catalog would drop the agent prompt to nothing and break codex's agent behaviour (verified empirically against codex 0.137). An empty list keeps codex on its built-in model info — same inference as before, minus the error. Validated end-to-end with the real handler against codex 0.137: "failed to refresh available models" → 0 occurrences, instructions preserved (built-in Codex agent prompt, not empty). Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add generated quality gate reports, metrics files, and local setup prompt artifacts to .gitignore to prevent committing environment-specific or temporary files.
cheaper-inference#3490) Integrated into release/v3.8.18
…e#3451) Integrated into release/v3.8.18
…heaper-inference#3492) Integrated into release/v3.8.18
…cheaper-inference#3491) Integrated into release/v3.8.18
…inference#3473) Integrated into release/v3.8.18
…ity combos (cheaper-inference#3169) Integrated into release/v3.8.18
…-inference#1422 (cheaper-inference#3401) Integrated into release/v3.8.18
…-inference#3510) Integrated into release/v3.8.18
…g retry text (cheaper-inference#3513) PR cheaper-inference#3422 ('allow OpenAI usage-only empty choices chunks') reintroduced the assistant-content injection '[OmniRoute] Upstream returned an empty response. Please retry.' for empty `choices: []` chunks that carry no valid usage. Clients (Goose/opencode) feed that text back as a turn and spin in a retry loop -- the exact regression cheaper-inference#3400 had fixed by dropping the chunk. Restore the drop behavior for the no-usage case while preserving cheaper-inference#3422's standards-compliant forwarding of usage-only `include_usage` final chunks. Realign the mislabeled stream-utils test (it asserted the injection) and add a dedicated regression guard. Reported-by: @mochizzan Refs: cheaper-inference#3502, cheaper-inference#3388, cheaper-inference#3400, cheaper-inference#3422
…Bearer (cheaper-inference#3504) Integrated into release/v3.8.18
…aper-inference#3503) Integrated into release/v3.8.18
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a signature bypass mechanism ("skip_thought_signature_validator") for Antigravity/Gemini models to prevent 400 errors on signature-less tool calls, refactoring the request translator to emit native function calls and responses instead of legacy XML/text-based formatting. It also adds error parsing support for array responses and updates the unit tests. A critical issue was identified in openai-to-gemini.ts where a redundant check causes duplicate IDs to be pushed to toolCallIds, resulting in duplicated tool responses in the generated payload.
Important
The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.
| if (!signatureForToolCall && contextualizeSignaturelessToolResponses) { | ||
| if (!toolCallIds.includes(id)) toolCallIds.push(id); | ||
| } |
There was a problem hiding this comment.
Since the legacy signature-less text/context mode continue statements and custom formatting have been removed, all tool calls that pass the initial checks will unconditionally reach line 433 and have their id pushed to toolCallIds.
Keeping this block causes duplicate IDs to be pushed to toolCallIds when !signatureForToolCall && contextualizeSignaturelessToolResponses is true, which in turn duplicates the tool responses in the generated payload. We should remove this redundant block entirely.
cc0dc4a to
3be18d1
Compare
…per-inference#3513, cheaper-inference#3435-repair, gitignore chore (full commit↔changelog coverage)
…polynomial-ReDoS)
CodeQL js/polynomial-redos: unbounded \s* before an anchored \s*$ allowed
O(n²) backtracking on attacker-influenced display names. Bounded to {0,8}/{1,8}
(ample for any real label spacing). Plugin builds + 254 tests green.
- getPendingRequests() typed to real shape (was widened to object) → fixes unknown 'count' in the unified-requests view (cheaper-inference#3401) - streamChunks log payload cast to its declared type (callLogs.ts) - preScreenTargets aligned to canonical IsModelAvailable signature (cheaper-inference#3169), Promise.resolve-normalized so .catch never hits a bare boolean All 5 gates green: lint(0 err) + typecheck:core + cycles + docs-all + unit + vitest(146).
…mpatibility and JSON parsing
…ty preservation per rule #18
Standard Gemini API (FORMATS.GEMINI) does not support the thoughtSignature field and rejects requests containing it with 400. The bypass sentinel 'skip_thought_signature_validator' was being unconditionally injected by openaiToGeminiBase, which is shared between standard Gemini, Gemini CLI, and Antigravity. Fix: add supportsSignatureBypass option to GeminiToolNameOptions, set it true only in openaiToGeminiCLIRequest (which serves both CLI and Antigravity). Standard Gemini callers (openaiToGeminiRequest) leave it undefined/false. Addresses review comment from @gemini-code-assist on PR cheaper-inference#3414.
The conditional toolCallIds.push(id) was redundant since the unconditional push at the end of the loop always executes. This caused duplicate tool response entries when contextualizeSignaturelessToolResponses was true. Addresses HIGH priority review comment from @gemini-code-assist on PR cheaper-inference#3514.
1aca0cd to
d3978be
Compare
|
Thanks @oyi77. The intended fix here (thoughtSignature bypass) is a real problem worth solving, but I can't merge this PR as-is: it touches 136 files for what should be a ~6-file change. ~42 of them are This PR is also Could you land the fix on a clean branch off current |
|
Superseded by #3560 (focused 6-file re-cut off release/v3.8.20 per @diegosouzapw review). Closing this one in favor of the cleaner PR. |
Problem
PR #3414 introduced a
skip_thought_signature_validatorbypass sentinel injected viaopenaiToGeminiBase, which is shared between standard Gemini, Gemini CLI, and Antigravity. As @gemini-code-assist correctly identified:FORMATS.GEMINI) does NOT support thethoughtSignaturefield — it rejects requests with 400Fix
Added
supportsSignatureBypassoption toGeminiToolNameOptions:trueonly inopenaiToGeminiCLIRequest(which serves Gemini CLI + Antigravity)openaiToGeminiRequest) leave it undefined → no bypass injectedChanges
open-sse/translator/request/openai-to-gemini.ts: AddedsupportsSignatureBypassoption, conditional bypass injectiontests/unit/t43-gemini-tool-call-no-thought-signature.test.ts: Reverted test expectations for standard Gemini (should NOT get bypass)Tests
Supersedes #3414 (includes all original fixes + addresses review comments)