Conversation
(cherry picked from commit 2095ba1)
(cherry picked from commit 6443712)
(cherry picked from commit 9f5df4e)
…deadline (cherry picked from commit 2149c4b)
A deferred settle can land after server.stop has released the spend-ledger owner: the journal is already closed, the outstanding sends have nobody to book against, and the thrown SpendLedgerOwnerError escapes as an unhandled rejection that kills the process. That crashed 'an unsafe configured key id is omitted with one id-free warning' in the refusal-recovery windows run when a client-cancel drain fired mid-test. Send records die with the discarded ledger; non-ownership failures still propagate. Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> (cherry picked from commit 3dd0dfd)
…matches is-gpt-nerfed-style auditability: the log already carried the routed model and, on some paths, the upstream-served one, but the GUI rendered resolvedModel alone and a virtual-model rewrite could overwrite the served value before it was ever compared. Record the upstream-served model on a dedicated servedModel field (openai-model header and response body), keep the exact wire id on wireModel when a route/virtual rewrite diverges it from the client-facing model, and render 'wire → served' in the Logs model column when they differ. (cherry picked from commit 10e3441)
(cherry picked from commit 37de7b0)
The reroute span's own title masked the cell's modelTitle diagnostics on exactly the rows where wire/served details matter most. Share one isModelRerouted predicate, fold the reroute notice into the tooltip, reuse the wire -> served arrow in the detail dialog, and cover the no-reroute case. (cherry picked from commit 3d989e3)
|
Important Draft PR not reviewedDraft PRs are not automatically reviewed by default.
To automatically review draft PRs, update your CodeRabbit configuration: reviews:
auto_review:
drafts: trueThanks 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 |
|
리뷰 · 우선순위 60 / 80이 PR은 Codex 쪽 문제를 세 덩어리로 묶은 스택입니다. 첫째, Responses Lite V2가 맨 위 라인 - 메인테이너의 판단이 필요한 지점 너의 추천 이 댓글은 grok-bot이 작성했습니다 |
⏳ DRAFT
What to do
This pull request was already a draft. Its draft status will be preserved after every issue above is resolved. |
|
Superseded by #5609, which is open and not yet merged. Cherry-picked (authored by you): refusal recovery with the pool harness (1b55d81 + c7e06af), served/wire model logging (632cf0d), the V2 catalog test isolation (44539f0) and the model-identity docs (271fde4). The logged served model is now bounded, and it is dropped when the secret redactor would change it. The settlement change (bba58c3, originally by Devin AI) is reimplemented crediting Devin AI and you, so that only an expired ledger lease is dropped. The reroute translations (24d8a0c) and the tooltip (4ea2059) are reimplemented with you as co-author; Turkish and Vietnamese are now translated. Two parts are already on |
Summary
Combine #5492, #5486, and #5485 above the Codex client compatibility layer. The client and routing layers carry seven original PRs; #5529 adds five related catalog/capability sources to this same Stack.
additional_toolscatalog for Responses Lite V2 plaintext forwarding when top-level tools are absent, preserving existing opt-in and collision boundaries.wire → servedmismatches with complete translated tooltips.Screenshot
Actual combined Logs component with synthetic API responses, including mismatched and matching model identities. No live user account or request data is included.
Related maintainer carry
Shared contribution: #5492 is also carried by maintainer-owned #5508. The relevant production file is identical between the two draft replacements; neither draft proves dev inclusion. Keep #5492 open until the landed implementation is verified, and reconcile this layer against dev when that carry lands.
Verification
c37f4f5f43351feee3ca46736770429ccbd068df; current head8ca80420b4da6cedac59738db850e85b48b10829.44539f0makes the existing catalog-state test override hermetic, avoiding real Windows PowerShell process enumeration and cross-test fixture interference. The final uninstrumented V2 file passed 27 tests / 152 assertions. Product behavior, ACL rules and timeout limits were unchanged.Checklist
Review readiness