Repository navigation
fix(openai): trace computer use and access programs - #876
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bd8744bff5
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "Codex (@codex) review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "Codex (@codex) address that feedback".
| "browser_authentication_request": ("request_id",), | ||
| "computer_use_approval_request": ("request_id",), | ||
| "computer_use_approval_request_result": ("request_id",), |
There was a problem hiding this comment.
Preserve the approval payloads in tool spans
When computer use requests browser authentication, the actual computer_use_approval_request item carries a request object and the corresponding result carries a credential-free response describing submit/cancel and the selected option. These allowlists retain only request_id, while _agent_tool_span_output has no output keys for either type, so the emitted spans discard both the form being requested and the user's admitted response. Include request/response in the appropriate span data and cover the real item shapes with a cassette-backed test.
AGENTS.md reference: AGENTS.md:L176-L182
Useful? React with 👍 / 👎.
Summary
access_programsin request and response metadata, with defensive reads for older OpenAI pins.Fixes #869
Live traces
access_programstrace