[markstuart-oai]

Summary

Change the output format for later Responses calls by adding a configuration update to input. Use parse_as to choose the Python type that reads the answer.

Changes

• Set configuration_update.text.format to a JSON schema or to {"type": "text"}. Retrieved input items expose the same format. A reasoning-only update keeps the active format.
• Pass parse_as to synchronous or asynchronous responses.parse(). It selects the local parser. If supplied, text_format still sets the top-level request schema.

response = client.responses.parse(
    model=model,
    previous_response_id=previous.id,
    input=[
        {"type": "configuration_update", "text": {"format": detailed_format}},
        {"role": "user", "content": "Include the evidence."},
    ],
    parse_as=AnswerWithEvidence,
)

On a later request that carries only a previous response ID and a new message, pass parse_as=AnswerWithEvidence to keep reading that type locally.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-30T18:53:32.835013Z	2df1423	New commits
🔒 Security Review	✅ Completed 2026-10-30T18:54:22.212930Z	2df1423	New commits

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 review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.