fix: to_dict in OpenAIResponsesChatGenerator and json_schema for structured outputs

[Amnah199]

Related Issues

• The to_dict method currently converts Pydantic models to JSON Schema, which can cause errors when the deserialized component is executed.
• the generator needs to support JSON Schema with text instead of text_format.

Proposed Changes:

• Fix the to_dict method.
• Update the docstrings to explain how users can pass a JSON Schema instead of a Pydantic model.
• Update the docstrings to clarify that streaming and Pydantic models are not supported together.
• Note: OpenAI’s implementation with structured outputs is very unclear/messy. we still do not support streaming with structured outputs. Updated the docstring accordingly.

How did you test it?

Notes for the reviewer

Checklist

• I have read the contributors guidelines and the code of conduct
• I have updated the related issue with new insights and changes
• I added unit tests and updated the docstrings
• I've used one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test: and added ! in case the PR includes breaking changes.
• I documented my code
• I ran pre-commit hooks and fixed any issue

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
haystack-docs	Ready	Preview	Comment	Nov 11, 2025 10:42am

[sjrl]

@Amnah199 please make sure to update and fix the to_dict method of AzureOpenAIResponsesChatGenerator as well

[sjrl]

@Amnah199 should we also add an integration test where we pass in a pydantic model, serialize the component, deserialize it and run the deserialized component to ensure it has the expected output? I'm asking since it feels a bit odd that at serialization time we convert a text_format passed option by the user into text.

[Amnah199]

I tested this aspect locally but sure will add it to the tests.

[sjrl]

Looks good!