feat: relax OpenAI and HF chat message conversions

[sjrl]

Related Issues

• fixes Agent crashes on the next LLM call after a Chat Generator discards a malformed tool call #12541

Proposed Changes:

• The OpenAI Chat Completions and Responses converters no longer raise on an assistant message with no content parts, which a Chat Generator returns when it discards a malformed tool call. It is sent with empty content, which the APIs accept, so the next LLM call goes through. ChatMessage.from_openai_dict_format accepts the same empty content, so such a message round-trips.
• The Hugging Face message converter no longer raises on an assistant message with no content parts, and sends it with empty content instead.

How did you test it?

New tests and expanded existing ones.

Also tested some of the HuggingFace integration components locally to check that their chat templates and models worked with empty assistant messages. E.g.

• TransformersChatGenerator with HuggingFaceTB/SmolLM2-135M-Instruct worked
• HuggingFaceAPIChatGenerator with Qwen/Qwen2.5-72B-Instruct and meta-llama/Llama-3.1-8B-Instruct worked

Notes for the reviewer

Integrations that support empty content:

• watsonx with ibm/granite-4-h-small works. no changes needed since it uses the openai conversion util method
• litellm with openai/gpt-4.1-nano works. no changes needed since it uses the openai conversion util method
• cometapi. no changes needed inherits from OpenAIChatGenerator
• stackit. no changes needed inherits from OpenAIChatGenerator
• aimlapi. no changes needed inherits from OpenAIChatGenerator

Integrations that support empty content but will need an update:

• anthropic with claude-opus-5 if we relax the converter
• ollama with qwen3:0.6b if we relax the converter
• llama_cpp with Qwen2.5-0.5B-Instruct-GGUF if we relax the converter
• google gen-ai works if we relax the converter

Integrations that don't support empty content

• mistral raises a 400 error if the history looks like [user, assistant("")], [user, assistant("Paris is")] or [user, assistant(""), user]
• perplexity raises a 400 error if the history looks like [user, assistant("")], or [user, assistant(""), user]

Checklist

• I have read the contributors guidelines and the code of conduct.
• I have updated the related issue with new insights and changes.
• I have 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 have documented my code.
• I have added a release note file, following the contributors guidelines.
• I have run pre-commit hooks and fixed any issue.

[vercel]

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

1 Skipped Deployment

Project	Deployment	Actions	Updated
haystack-docs	Ignored	Preview	Sep 2, 2026 7:37am UTC

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/components/generators/chat
openai_responses.py
haystack/dataclasses
chat_message.py
haystack/utils
hf.py
Project Total

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

[sjrl]

@anakin87 could you review this PR since you already have some context? Also just to let you know I will be opening follow up issues for the integrations that need to be updated.

Update:
Made the issues. You can see them at the bottom of the issue #12541

[anakin87]

From the code point of view, it looks good

I'd improve/remove some comments and a test