Repository navigation
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Merge Risk: ⚪ Minimal · up to
|
| Check name | Status | Explanation |
|---|---|---|
| Docstring Coverage |  |
Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 5 functions across 3 files. (5 skipped: 5 … |
| Linked Issues check | |
Check skipped because no linked issues were found for this pull request. |
| Out of Scope Changes check | |
Check skipped because no linked issues were found for this pull request. |
| Description Check | |
Check skipped - CodeRabbit’s high-level summary is enabled. |
| Title check | |
The title clearly and concisely describes the primary change: adding Ultra Fast to live Codex model discovery. |
✨ Finishing Touches
-
🧪 Generate unit tests (beta)
-
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
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 help to get the list of available commands.
|
✅ Deterministic PR hygiene checks passed. |
✅ READY
Review readiness checklist
✅ 4/4 boxes ticked. This pull request is already Ready for Review. |
|
Being carried into #7220 with your authorship preserved (own commit + Co-authored-by); this PR will be closed when the bundle lands on dev. |
… ids and auto-review override (#7115, #7126, #7173, #6867) (#7220) Carried from #7115, #7126, #7173, #6867 (one commit per source PR, original authors preserved). Remote exact-head validation (hosted GitHub Actions stalled; local tests not run by maintainer rule): union candidate codex/prr-integ-6 (368006f, tree 27fdc37) on mini (Windows) and lidge (Linux), typecheck + full suite, compared with base-dev 4ae4936 on each host. Co-authored-by: 정우철 <oocheol@naver.com> Co-authored-by: Rafael Moreira <rrmlima@gmail.com> Co-authored-by: sechmachine <97589681+sechmachine727@users.noreply.github.com>
2 participants
Summary
GET /v1/models?client_version=...received only Fast even when a routed provider explicitly declaredultrafast -> ultrafast. The public catalog builder now applies the existing configured-tier projection, using the same captured declarations and native/account/combo exclusions as the persisted merge. Defaults and the persisted writers' paths stay unchanged.ultraFastTierretains independently supplied exact routed-row tiers, and separately generates missing tiers on eligible native Astra/Sol rows. Explain generated-tier withdrawal and preservation of operator edits. Update the catalog contract.Follow-up to the post-merge review on #6927: live catalog projection and native-setting documentation.
Verification
Based on upstream
dev7640a942b35044c974e534b84d5a8049f4a8ae9a, fetched and checked on 2026-10-15. Final tree committed as27d095b4f9731e366fc191e4655bb17bb2d0a45d; one fix commit, eight changed files. Local execution used Windows and repository-pinned Bun 1.4.3 (c6da4a4d3) through the isolated test wrapper.The direct-builder regression first failed on the original code: an explicit eligible declaration produced only
priority, withoutultrafast. The final regression file has eight passing cases, including a real local HTTP server queried twice with an active declaration. The HTTP fixture uses the server's startup configuration; it does not mistake a disk-only config save for an update to that server's captured live configuration. Returned routed rows contain both tiers, retaindefault_service_tier: null, and do not grant the choice to the ordinary unsupported row.Focused run:
802 passed, 0 failed, 3,892 assertions across 13 files, 26.02 seconds. This includes both persisted writers, native projection, operator retention, wire behavior and the optional-Lab boundary.
Additional bounded changed-mode run over all remaining test files mentioning the builder family:
969 passed, 2 skipped, 0 failed, 9,362 assertions across 41 files, 75.62 seconds. The wrapper resolved
origin/devto the base above and selected all 41 supplied files. The two existing Windows skips areEACCES on a configured regular catalog file is indeterminateandan unresolvable surface symlink is indeterminateincodex-native-residue.test.ts.Combined: 1,771 passed, 2 skipped, 0 failed across 54 distinct files.
bun run typecheck: passed.bun run structure:check: passed.bun scripts/file-size-ratchet.ts: passed; no baseline increases.bun run privacy:scan: passed.git diff --checkand staged diff check: passed.docs-site/,bun install --frozen-lockfilereported no package changes;bun run buildpassed with 577 pages / 81,200 internal links, 22.27 seconds.Full-suite scope exception: the repository has 2,326 backend test files. An exhaustive local run is disproportionate to this bounded catalog projection, so the focused regressions and remaining named builder consumers above were run instead. The full repository suite was not run. Whole-repository and cross-platform execution remain separate CI coverage; this PR does not change the Ubuntu-only PR CI policy.
An additional in-session static review found no concrete blocker in declaration conflicts, row authority, stale tiers, defaults or writer interactions. That review did not execute tests and is not maintainer approval. The local HTTP test uses disposable configuration and synthetic provider metadata; it does not establish rendered Desktop selection, provider entitlement or a successful paid Ultra Fast inference. Hosted results are separate from the local receipts above.
Current-head hosted status (
27d095b4f9731e366fc191e4655bb17bb2d0a45d), refreshed by lane 7115 on 2026-10-15: maintainer authorization was granted to the existing Ubuntu PR CI, Catalog async contracts, and React Doctor runs. Runner selection, change detection, Catalog async contracts (Ubuntu), and React Doctor passed; main Ubuntu test shards, gates and smoke jobs remain queued. This is not passing executable CI evidence. The pre-edit label, hygiene and enforce-target runs succeeded; earlier cancelled label/enforce-target runs were superseded. This verification update schedules fresh description checks. No cross-platform workflow was dispatched. GOAT rejected CI watching from this delegated subagent because the parent thread owns the PR; the parent must monitor the remaining jobs. Maintainer approval and passing current-head executable CI remain outstanding.Review follow-up on 2026-10-15: CodeRabbit completed its current-head review with no actionable comments and 100% docstring coverage. There are no inline review comments or unresolved review threads. The unchanged source diff was checked against that feedback; no additional code fix or validation rerun was warranted. This is not maintainer approval. The later lane verification below records workflow authorization separately.
Lane 7115 verification (2026-10-15)
Reviewed the contributor's unchanged head against current
origin/dev(7640a942b35044c974e534b84d5a8049f4a8ae9a). The live builder previously skipped the persisted merge's configured-tier overlay. Its new overlay uses the same captured capability declarations and native/account/combo exclusions, strips stale template tiers, and preserves the ordinary default. No fix-up commit was necessary; the original author and commit remain intact. There are no inline review threads.Regression proof: temporarily restoring only
build-entries.tsfromorigin/dev, then runningbun scripts/test.ts --timeout 45000 tests/codex-integration/ultrafast-live-catalog.test.ts, produced 5 pass / 3 fail, specifically missing Ultra Fast in the direct builder, routed alias and real HTTP response. The original contributor file was restored in afinallyblock; the worktree is clean.Exact focused command:
811 pass / 0 fail across 14 files; 3,977 assertions. Includes persisted writers, native tiers, wire policy, optional-Lab boundary, test-layout guards and file-size guard.
Exact bounded changed-mode command (the 41 remaining files found with
rg -l 'buildCatalogEntries' tests --glob '*.test.ts', excluding the already verified focused files):971 pass / 0 fail across 41 files; 9,366 assertions. Comparison resolved to
origin/devat the base above. Combined lane execution: 1,782 pass / 0 fail across 55 distinct files.Additional commands:
bun run typecheck,bun run structure:check,bun run privacy:scan,bun scripts/file-size-ratchet.ts, andgit diff --checkall passed. No source changes followed these checks. The contributor's unchanged documentation build receipt above remains applicable; lane 7115 did not rerun that build.Resource exception: concurrent lanes share one 16-core machine. The focused invocation was initially queued behind the user-wide test lock for over five minutes and was stopped before executing; the runner's supported
OCX_TEST_NO_QUEUE=1option then enabled intentional bounded overlap with one worker.OCX_TEST_NO_QUEUE=1 bun run test:changed --parallel=2 --timeout 45000was also attempted without filters, but stopped after over three minutes of graph evaluation with no test result, to bound contention. That unfiltered attempt is not a passing result. The bounded changed-mode run above passed. No fullbun run testwas run, as requested for the concurrent triage bundle. Exhaustive import-graph/full-suite coverage remains with Ubuntu CI; macOS/Windows coverage remains with the orchestrator. No rendered Desktop selection, provider entitlement or paid inference was exercised.Checklist
Review readiness checklist
This PR stays in draft until every box below is ticked. Tick all four boxes once the requirements are met:
Summary by CodeRabbit
New Features
Bug Fixes