Skip to content

fix(codex): expose vision settings for Responses providers - #2019

Draft
wangziheng437 wants to merge 1 commit into
jlcodes99:mainfrom
wangziheng437:codex/fix-responses-vision-settings
Draft

fix(codex): expose vision settings for Responses providers#2019
wangziheng437 wants to merge 1 commit into
jlcodes99:mainfrom
wangziheng437:codex/fix-responses-vision-settings

Conversation

@wangziheng437

Copy link
Copy Markdown

Summary

  • expose the existing image-input, per-model vision, and vision-routing settings for third-party Responses providers
  • keep official OpenAI and native DeepSeek Responses providers read-only where these snapshots do not apply
  • synchronize linked API-key provider snapshots and show a restart hint when active Codex/API Service processes may still use the old manifest
  • align all localized help text with the existing v1.3.8 behavior: omit unsupported images and continue with text unless a vision route is available
  • add regression coverage for UI capability policy, provider snapshot persistence, and Provider Gateway manifest serialization

Validation

  • npm run typecheck
  • node --test src/utils/codexModelProviderAccountSync.test.ts src/utils/codexProviderPresets.test.ts
  • node scripts/check_locales.cjs
  • npm run build
  • npm exec -- vite build
  • git diff --check

Environment limitation

  • Rust test compilation was attempted, but this Windows machine does not have the MSVC link.exe prerequisite. The targeted tests did not compile locally and are left for CI.
  • cargo fmt --all --check could not run because Windows application control blocked rustfmt (os error 4551).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant