[codex] Unify OpenAI auth provider picker#82324
Conversation
|
Codex review: needs real behavior proof before merge. Summary Reproducibility: no. high-confidence interactive reproduction was run in this read-only review. Source inspection does show current Real behavior proof Next step before merge Security Review detailsBest possible solution: Keep this PR open for maintainer review, undraft/update it when ready, and land the unified picker behavior only with the focused auth-choice/model-picker/provider-wizard tests green. Do we have a high-confidence way to reproduce the issue? No high-confidence interactive reproduction was run in this read-only review. Source inspection does show current Is this the best way to solve the issue? Yes, the proposed direction appears maintainable: it keeps provider-specific choice IDs while changing wizard grouping metadata and scopes the model allowlist by selected provider. It still needs maintainer judgment because this is a visible OpenAI setup UX/policy change and the PR is draft. What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning high; reviewed against e22a7e45a40a. |
3bddda6 to
f92fd41
Compare
|
Landed via rebased PR branch onto current main.
|
Summary
Real behavior proof
Behavior addressed: OpenAI configure no longer shows separate first-tier OpenAI and OpenAI Codex provider entries; Codex model selections are carried into the saved model allowlist.
Real environment tested: local OpenClaw checkout on macOS, rebased onto current main.
Exact steps or command run after this patch: node --input-type=module manifest probe for extensions/openai/openclaw.plugin.json; node scripts/run-vitest.mjs src/plugins/provider-validation.test.ts src/commands/configure.gateway-auth.prompt-auth-config.test.ts src/commands/model-picker.test.ts src/commands/auth-choice-options.test.ts extensions/openai/openai-codex-provider.test.ts extensions/openai/openai-provider.test.ts extensions/openai/openclaw.plugin.test.ts extensions/openai/setup-api.test.ts src/wizard/setup.test.ts; git diff --check.
Evidence after fix: terminal output from manifest probe showed the visible OpenAI choices after the patch:
Observed result after fix: the visible OpenAI setup surface is one featured OpenAI provider group with ChatGPT/Codex login/device pairing and OpenAI API key choices; legacy OpenAI OAuth choices remain manual-only compatibility routes. The rebased focused test lane also passed: 9 test files passed, 154 tests passed; git diff --check passed; codex-review was clean before PR creation.
What was not tested: live OAuth/browser login against OpenAI.