fix: map streamed reasoning item content

[Amazinghorseli]

Related Issues

• fixes OpenAIResponsesChatGenerator: streamed reasoning text (chunk.item.content[].text) is dropped #13061

Proposed Changes

Map reasoning_text content from a completed Responses API reasoning item into the Haystack reasoning text field when streaming. If a text delta has already been emitted for that output item, keep the streamed text and avoid appending it a second time. Preserve the full provider payload, including encrypted content, in ReasoningContent.extra.

Add conversion regression coverage for multiple text entries, existing streamed text, another output index, and empty or missing content. Update the component docs and release note.

How did you test it?

• pytest test/components/generators/chat/test_openai_responses_conversion.py test/components/generators/chat/test_openai_responses.py test/components/generators/chat/test_azure_responses.py -m 'not integration' -q — 93 passed, 23 deselected.
• Ruff lint and format checks passed for the changed Python files.

Notes for the reviewer

The change uses the completed item's reasoning_text entries only when the same output item has not already emitted reasoning text, preventing duplicate text while keeping the existing summary-delta path intact.

AI assistance was used to prepare the implementation and tests. Validation results are listed above.

[vercel]

@Amazinghorseli is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

All committers have signed the CLA.