Repository navigation
fix: support opt-in JSON validation for Codex tool parameters - #6067
Conversation
Codex Review Summary: ✅ CompletedThis 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. |
jbeckwith-oai
left a comment
There was a problem hiding this comment.
Reviewed commit ec2646b6f0567eac183c7156c29b184b1a41bf38. Approved; no actionable findings.
The explicit JSON-validation option uses Pydantic's existing JSON entry point while preserving Python-mode validation by default. Reviewed mapping/dataclass/keyword precedence, construction-time rejection of invalid modes, strict tuple/enum inputs, normalization and thread routing, validation-before-side-effects, single validator invocation, and redacted versus diagnostic error causes/contexts. The new tests exercise successful invocation and rejected input without starting or resuming a thread. Independent static review reached the same conclusion.
Hosted checks at review: {'SUCCESS': 21, 'IN_PROGRESS': 3}. Static review only; no candidate code or local tests executed.
This pull request adds
parameter_validation_mode="json"to the experimental Codex tool. Strict Pydantic models can accept JSON arrays for tuple fields and strings for enum fields. Existing tools retain Python-mode validation by default.Fixes #6057