Repository navigation
Conversation
|
Navigate logical layers of code changes, visualize relationships, and explore their blast radius. No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository: lidge-jun/opencodex/.coderabbit.yaml Review profile: ASSERTIVE Plan: Advanced Run ID: 📒 Files selected for processing (2)
Included review availability: Your plan provides up to 10 included reviews per hour; 5 remain after this review. 📝 WalkthroughWalkthroughThe admission estimator now follows the routed provider’s handling of replayed assistant thinking. It excludes thinking for ChangesInput admission
Priority: ➖ Normal Estimated code review effort: 2 (Simple) | ~10 minutes Change: Bug fix · Severity of issue fixed: Medium Suggested reviewers: Merge Risk: ⚪ Minimal · up to The admission change is mergeable after normal checks; no actionable issue remains from this review. 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
@coderabbitai review |
|
✅ Deterministic PR hygiene checks passed. |
✅ Action performedReview finished.
|
✅ READY
Review readiness checklist
✅ 4/4 boxes ticked. This pull request has been marked Ready for Review. |
리뷰 · 우선순위 64 / 80이 PR은 글을 모델에게 보내기 전에 "너무 길지 않은가"를 재는 문을 고칩니다. 긴 대화에는 모델이 예전에 속으로 한 생각(thinking)이 많이 쌓입니다. 이제는 그 목록에 없는 docs-site/src/content/docs/ko/reference/architecture.md - 끝 문장이 일을 잘못 잇습니다. 잘못 거절을 막는 쪽은, 보내지 않는 생각을 계산에서 빼는 일입니다. 생각을 보존하는 모델을 계속 세는 일은 그 이유와 다른 문장으로 써야 합니다. tests/server/input-admission.test.ts - 직접 경로 테스트는 전송 글에 메인테이너의 판단이 필요한 지점 이 PR은 아직 초안입니다. 준비 체크는 4개 중 0개입니다. 설명에는 관련 테스트 51개 통과와 타입 검사가 적혀 있습니다. 전체 테스트는 이 작업 컴퓨터에서 끝까지 돌리지 못했다고 본인이 적었습니다. 코드 방향과 별개로, 저장소 게이트는 체크를 채워야 초안을 풉니다. 계산은 어댑터와 같은 너의 추천 고치는 방향은 맞습니다. 한국어 설명의 끝 문장을 나누고, 테스트가 긴 생각 문자열이 전송 본문에 없는지도 보면 됩니다. 그다음 준비 체크를 채우고 머지하면 됩니다. 이 댓글은 grok-bot이 작성했습니다 |
3f55bc6 to
ac697a7
Compare
|
@coderabbitai review |
✅ Action performedReview finished.
|
…, #5714, #5704, #5707) (#5738) * fix(claude): bound Responses user to 64 chars for long metadata.user_id Claude Code sends metadata.user_id as a ~186-char JSON string. It was copied verbatim into the Responses 'user' field, which Azure OpenAI and other OpenAI-compatible backends cap at 64 chars, so every Claude Code request routed to Azure failed with 400 "Invalid 'user': string too long". Keep short ids as-is and send the SHA-256 hex digest (already computed for prompt_cache_key) when the id exceeds 64 chars. Fixes #5705 Co-authored-by: Giulio Leone <giulioleone097@gmail.com> * Fix plaintext V2 SSE responses with missing content type Co-authored-by: Jerry WANG <jerrywang@Jerrys-MacBook-Pro-2.local> * fix(responses): exclude dropped chat reasoning from input admission * test(admission): verify reasoning payload matches gate decisions Co-authored-by: 정우철 <oocheol@naver.com> * Allow HTTP upstream for canonical ChatGPT provider An explicit upstreamWebsocket: false now routes streaming canonical ChatGPT turns over HTTP/SSE before sending. This gives operators a supported escape from intermittent post-send WebSocket closes without replaying ambiguous turns. The default remains WebSocket and native WS controls are unavailable when HTTP is selected. Verified: focused Responses/provider suites, typecheck, structure check, privacy scan, docs build. Changed-area suite rerun pending. Co-authored-by: kosta <kosta963@gmail.com> * test(claude): pin exact user hash and the 64/65-char boundary Refs #5705 Co-authored-by: Giulio Leone <giulioleone097@gmail.com> * docs(architecture): attribute the admission fix to excluding unsent thinking Split the preserved-reasoning statement from the refusal rationale in the English and Korean paragraphs, as review of #5714 asked. Refs #5696 Co-authored-by: 정우철 <oocheol@naver.com> * fix(anthropic): count ping events as upstream liveness Anthropic streams may include any number of ping events. The adapter only turned SSE comments into heartbeats, so a long silent thinking phase that pinged was cut off at stallTimeoutSec with upstream_stall_timeout. Named and data-only ping records now yield the same heartbeat. Refs #5707 * fix(management): report upstreamWebsocket as configured in GET /api/providers The row coerced an unset value to false. After the canonical ChatGPT opt-out, unset means upstream WebSocket and false means HTTP/SSE, so a save built from the row could turn WebSocket off. The row now omits the key when it is unset. Co-authored-by: kosta <kosta963@gmail.com> * docs(providers): describe the canonical ChatGPT upstreamWebsocket opt-out The reference row, its seven translations, and the provider type and schema comments still said the canonical ChatGPT transport ignores upstreamWebsocket. It now selects HTTP/SSE when set to false. The locale rows were also behind the English row on the first-party-only restriction; they are retranslated from it. Co-authored-by: kosta <kosta963@gmail.com> * fix(responses): bound the plaintext V2 SSE prefix probe by stallTimeoutSec The carried probe read the first chunk of an unlabeled body with no deadline, before the passthrough stall guard is attached. An upstream that sent headers but no body held the request and its host lease until the client gave up, and a failed read escaped the classifier. Each probe read now races a per-read inactivity window and one total budget, both stallTimeoutSec, plus the client abort signal. Timeout, abort, and read errors cancel the reader and return the existing unsupported-content-type 502. Docs now scope the recovery to missing or unrecognized non-JSON content types. Follow-up to #5683 review (maintainer, Codex, CodeRabbit). Co-authored-by: Jerry WANG <jerrywang@Jerrys-MacBook-Pro-2.local> --------- Co-authored-by: Giulio Leone <giulioleone097@gmail.com> Co-authored-by: Jerry WANG <jerrywang@Jerrys-MacBook-Pro-2.local> Co-authored-by: 정우철 <oocheol@naver.com> Co-authored-by: kosta <kosta963@gmail.com>
Summary
The pre-dispatch input gate counted replayed assistant thinking for every route, but
openai-chatdrops that history unless the model is inpreserveReasoningContentModels. A long Codex thread could therefore get a localcontext_length_exceededrefusal despite sending a much smaller upstream request. The gate now uses the routed provider's same preservation rule for both direct and combo-target admission. Preservedopenai-chatreasoning and other adapters still count it. Closes #5696.Both direct and combo regressions verify that the long thinking string is absent from the dropped wire body and present when preserved, alongside the admission decisions. English/Korean architecture guidance and the Responses failover contract are updated.
Verification
Windows, Bun 1.4.0, isolated test homes:
Results: 51 pass, 0 fail; typecheck, structure, privacy, and diff checks passed.
docs-sitebuild passed (505 pages).An initial parallel run of the same three files failed in
responses-context-overflow.test.tsbecause multiple test servers in one process conflicted over the process-global spend-ledger home (SPEND_LEDGER_OWNER_HOME_CONFLICT); the complete serial rerun passed. Full-suite exception: the earlier--changed=devrun on this shared Windows host selected a large import-connected set and was stopped after more than four minutes without a result. It is not passing evidence. This bounded gate change was validated with the estimator, adapter, and request-level overflow suites; full-suite and cross-platform validation remain for CI.Checklist
Review readiness checklist
This PR stays in draft until every box below is ticked. Tick all four boxes once the requirements are met: