fix: stop Agent retries after output limit

[Nikhi00718]

Related Issues

• fixes Agent loops to max_agent_steps when a reasoning model returns no text because it hits max_output_tokens #12300

Proposed Changes:

• Map OpenAI Responses API completions stopped by max_output_tokens to Haystack's existing length finish reason in both streaming and non-streaming paths.
• Treat a tool-call-free assistant reply with that finish reason as a text Agent exit, while retaining the retry behavior for other empty replies such as discarded malformed tool calls.
• Add sync, async, streaming, and non-streaming regression coverage.

How did you test it?

• C:\Program Files\Hatch\hatch.exe run test:unit test/components/agents/test_agent.py test/components/generators/chat/test_openai_responses_conversion.py (121 passed, 4 integration tests deselected)
• C:\Program Files\Hatch\hatch.exe run test:unit test/components/generators/chat (251 passed, 43 integration tests deselected)
• C:\Program Files\Hatch\hatch.exe run test:types
• C:\Program Files\Hatch\hatch.exe run ruff check --fix
• C:\Program Files\Hatch\hatch.exe run ruff format
• C:\Program Files\Hatch\hatch.exe run pre-commit run --files ... (all scoped hooks passed)

Notes for the reviewer

The new Agent regression tests fail on main because the Agent makes a second model call after a length-truncated empty response. The Responses conversion regressions also fail on main because the same condition is exposed as stop in streaming mode and has no normalized finish reason in non-streaming mode.

hatch run fmt is not directly executable on Windows because the repository script's Unix-style semicolon is passed literally to Ruff; the two configured Ruff commands above were run separately through Hatch.

AI assistance disclosure

I identified and reproduced the issue, then used OpenAI Codex to assist with the implementation and regression-test work. I personally reviewed the resulting diff and ran the validation commands listed above to verify the fix. Any full-suite or local-environment limitations are documented in the validation section.

Checklist

• I have read the contributors guidelines.
• I have updated the related issue with helpful information about this PR.
• I have added unit tests and documentation, if applicable.
• I have followed the commit conventions.
• I have documented any new or changed behavior.
• I have added a release note.
• I have run the pre-commit hooks.

[vercel]

@Nikhi00718 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.

[julian-risch]

@sjrl Could you please take over this review? Thank you!

[sjrl]

@Nikhi00718 please also handle the merge conflicts with main

[Nikhi00718]

Addressed all review feedback:

• Handle both response.completed and response.incomplete terminal streaming events.
• Normalize max_output_tokens to length and content_filter to content_filter for streaming chunks and ChatMessage metadata.
• Updated the release note to describe the generic Agent behavior, the reasoning-token exhaustion scenario, and the FinishReason metadata.
• Updated coverage to use the real ResponseIncompleteEvent and test both incomplete reasons.
• Confirmed the branch is up to date with main and mergeable.

Verification:

• 126 passed, 2 skipped across the two affected test modules.
• Ruff lint and format checks pass for all affected files.
• Targeted mypy check for the updated conversion test passes.

All four review threads have been replied to and resolved.

[HaystackBot]

Hi @Nikhi00718, thanks a lot for your contribution! 🙏

We noticed that the Contributor License Agreement (CLA) check (license/cla) hasn't passed yet, so we've temporarily moved this PR to draft and paused the review assignment.

To get your PR reviewed, please sign the CLA via the link in the license/cla check below (or in the CLA bot comment). As soon as the check turns green, this PR will automatically be marked ready for review again and a reviewer will be re-assigned.

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/components/agents
agent.py					159
haystack/components/generators/chat
openai_responses.py					716
haystack/core/pipeline
pipeline.py
Project Total

This report was generated by python-coverage-comment-action

[Nikhi00718]

@sjrl The CLA Assistant says all committers have signed, but the license/cla status is missing on the latest commit and the cla-pending label remains. Could you please retrigger the CLA check and review assignment?

[sjrl]

@sjrl The CLA Assistant says all committers have signed, but the license/cla status is missing on the latest commit and the cla-pending label remains. Could you please retrigger the CLA check and review assignment?

Yeah sorry about that. I'll be able to finish reviewing this early next week and look into the bot issue.

[HaystackBot]

Thanks for signing the CLA, @Nikhi00718! 🎉 This PR is now ready for review again and the reviewer has been re-assigned.