fix(catalog): sidecar-covered combo members keep image advertising - #4044
fix(catalog): sidecar-covered combo members keep image advertising#4044x3M3x wants to merge 1 commit into
Conversation
Complete discovery rows reach resolveComboCatalogMember un-hinted, so a text-only vision-sidecar consumer collapsed the whole combo to input_modalities text at derivation. The Codex app gates attachments client-side on input_modalities, so combos containing a no-vision member (e.g. planners with DeepSeek targets) blocked pasted images before the sidecar could run, even though the same member advertised image on its own provider row. Mirror the direct-row sidecar advertisement inside combo member resolution so derivation intersects the modalities the runtime actually serves.
|
✅ Deterministic PR hygiene checks passed. |
⏳ DRAFT
What to do
Review readiness checklist
0/4 boxes ticked. This PR stays in draft until every box above is ticked. |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .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; 9 remain after this review. 📝 WalkthroughWalkthrough
ChangesVision-sidecar modality resolution
Priority: ➖ Normal — Schedule the catalog modality fix because affected combo models can block image pasting before the proxy translates the request. Estimated code review effort: 2 (Simple) | ~10 minutes Merge Risk: ⚪ Minimal · up to Combo models backed by a vision sidecar now advertise image input correctly, while explicitly disabled image input and uncovered providers retain their existing behavior. The change is covered by focused regression tests and has no remaining merge-blocking risk. 🚥 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 |
리뷰 · 우선순위 71 / 80이 PR은 Codex 카탈로그에서 콤보(combo) 멤버가 비전 사이드카(vision sidecar)로 이미지를 처리할 수 있는데도, 카탈로그에는 흐름을 초등학생에게 설명하듯 풀면 이렇다. Codex 앱은 이미지를 붙이기 전에 카탈로그의 코드 위치는 회귀 테스트 라인 쪽은 전체적으로 의도가 맞고 범위도 작다. 한 가지 읽을 때만 주의할 점은 spread 순서다. 메인테이너의 판단이 필요한 지점
너의 추천 이 댓글은 grok-bot이 작성했습니다 |
Summary
["text"]in the Codex catalog, so Codex blocked pasting images client-side with "This model does not support image inputs" before the proxy could translate them.resolveComboCatalogMemberthrough thewithFallbackMetadataidentity fast path un-hinted, so the sidecarimageadvertising that fix(catalog): advertise image input for modalities-declared sidecar rows #3105 added for direct provider rows never reached derived combo members.withFallbackMetadatanow appends"image"to a member's modalities when the provider's vision sidecar covers that model — the same rule direct rows already get — so derived combo rows keep image support.imageInput: "disabled"still strips it, and providers without sidecar coverage keep byte-identical rows.Verification
tests/codex-integration/catalog-vision-sidecar-modalities.test.ts: a sidecar-covered discovery member gains["text","image"], the derived combo keepsimage,imageInput: "disabled"strips it, and an uncovered provider's row is identity-equal.bun run typecheckgreen.devwithout this patch. The three-OS CI run is the gate for those lanes.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.
Summary by CodeRabbit