fix: preserve Daybreak Blue Codex capabilities - #1604
Conversation
|
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
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 |
|
✅ Deterministic PR hygiene checks passed. |
⏳ DRAFT
What to do
Review readiness checklist
1/4 boxes ticked. This PR stays in draft until every box above is ticked. |
62327fb to
2ff758c
Compare
…eferral PR #1604 branched before fcbef38 ("restore deferred tool discovery for non-Cursor routed rows") landed on dev, so its negative control still expected supports_search_tool=false for a routed openai row. That field is now the ordinary routed default and carries no Daybreak-inheritance signal; the native-only assertions around it still prove the rejection.
|
Landed on One follow-up commit was needed on Verification on the merged GitHub does not auto-close PRs merged to a non-default branch, so closing manually. |
Summary
gpt-daybreak-blue-latestas an account-observed, selector-qualified native id without widening the bare native allowlist.openai/gpt-daybreak-blue-latestcustom model on the canonical Codex-forward provider inherit the pinned GPT-5.6 Sol capability snapshot while keeping the Daybreak wire id.openai-apikey/daybreak-blue-latestAPI route and its 1.05M / 922k limits isolated from the 372k Codex-login contract.authModeto its registry default (forward) for catalog discovery, matching routing behavior and avoiding a stray/modelsrequest.This builds on #1515: that PR preserves account-observed unknown native ids as selector-qualified rows. This follow-up covers the explicit canonical Codex-forward custom-model path and does not create a bare Daybreak row or grant account entitlement.
Runtime verification
The patch was deployed to a local OpenCodex 2.14.0 installation with an explicit canonical Codex-forward custom-model entry. After selecting a ChatGPT Codex account known to have Daybreak entitlement, one Codex-shaped streaming request to
openai/gpt-daybreak-blue-latestreturned HTTP 200, emittedresponse.completed, and producedOK. The usage ledger recorded the exact requested selector, exact upstream wire idgpt-daybreak-blue-latest, themainaccount label, andgpt-5.6-solas the capability resolution.This smoke test also confirmed that entitlement is account-specific: two pool accounts without Daybreak access returned the upstream unsupported-model response before the entitled account was selected. The PR preserves routing and metadata; it does not claim or manufacture entitlement for every ChatGPT account.
Verification
devcommit1849b947(HEAD2ff758c7, one commit ahead).bun run typecheckbun run privacy:scannode_modules/bun/bin/bun.exe test --isolate --timeout 30000 tests/codex-catalog.test.ts tests/codex-catalog-sync-hardening.test.ts tests/codex-convergence-account-selectors.test.ts tests/codex-routing.test.ts— 341 pass, 0 failcd docs-site && node_modules/bun/bin/bun.exe run build— 265 pages builtgit diff --check origin/dev...HEADI also attempted
bun run teston Windows. The default local runner still applies Bun's 5-second per-test ceiling and produced unrelated management/storage/Claude endpoint timeouts, followed by shared-fixture contamination. Once the suite was irreversibly non-green, I stopped only that owned run and verified that no repo-owned Bun/Node descendants survived. The changed-area suites above pass completely. CurrentdevCI separately runs isolated/sharded batches with a 60-second per-test timeout.Checklist
Review readiness checklist
This PR stays in draft until every box below is ticked. Tick all four boxes once the requirements are met:
All CI tests are green on my local testing.
I pushed my PR to the latest dev commit.
I resolved all correct Codex and CodeRabbit findings.
My PR is ready for review.