[mosafariuk]

What does this PR do?

The LangChain CallbackHandler drops the name of every message from the captured generation input.

_convert_message_to_dict only reads message.additional_kwargs["name"]. Current LangChain stores the name on the message itself (BaseMessage.name), so it never reaches Langfuse. That includes HumanMessage(name="alice"), ToolMessage(name=...) and FunctionMessage, whose name is required, even though LangChain does send those names to the provider. For example, with a real ChatOpenAI and only the HTTP layer mocked:

Sent to OpenAI (names): [('user', 'alice'), ('assistant', None), ('tool', None), ('function', 'get_weather')]
Langfuse input (names): [('user', None), ('assistant', None), ('tool', None), ('function', None)]

This PR reads message.name and falls back to additional_kwargs["name"], so integrations that still put it there keep working. With the change:

Langfuse input (names): [('user', 'alice'), ('assistant', None), ('tool', 'get_weather'), ('function', 'get_weather')]

The langfuse-js counterpart (#950) describes this as matching what the Python SDK already does, so this brings the Python behaviour in line with that.

Fixes # (no existing issue)

Type of change

• Bug fix
• New feature
• Breaking change
• Refactor
• Documentation update
• Tooling, CI, or repo maintenance

Verification

uv run pytest tests/unit/test_langchain.py tests/unit/test_langchain_utils.py   # 30 passed
uv run pytest tests/unit                                                        # 946 passed; 18 errors in test_prompt.py, identical on unmodified main (fixture issue addressed by #1979)
uv run ruff check . && uv run ruff format --check .
uv run mypy langfuse

The new test_chat_model_input_keeps_message_names in tests/unit/test_langchain.py follows the existing exporter-local pattern (patched ChatOpenAI._generate, in-memory exporter). It covers message.name on human, tool and function messages, plus the legacy additional_kwargs["name"] fallback. It fails on main (('user', None) != ('user', 'alice')) and passes with this change.

Checklist

• I self-reviewed the diff using code_review.md.
• I added or updated tests for behavior changes.
• I updated docs, examples, or .env.template if needed.
• I did not hand-edit generated files; if generated files changed, I used the upstream regeneration path.
• I did not commit secrets or credentials.

Confidence Score: 5/5

The PR appears safe to merge.

Summary

The PR keeps message names in LangChain generation input.

• Langfuse records names attached directly to LangChain messages.

No actionable issues were found. Tests were not run during this review.

_{Reviews (1) · Last reviewed commit: "fix(langchain): keep message names in ge..." · Reviewed by Greptile}

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[CLAassistant]

All committers have signed the CLA.

[chrikrah]

@mosafariuk I'd merge this at d9cec13. On langchain-core 1.2.22 the name lives on BaseMessage.name. The merge base reads only additional_kwargs, so a name set with name= never reached the input.

$ .venv/bin/python -m pytest tests/unit/test_langchain.py tests/unit/test_langchain_utils.py -q -p no:cacheprovider   # d9cec13, langfuse from the tree
30 passed in 5.10s

# langfuse/langchain/CallbackHandler.py from 21834c5 (merge base), your test kept
FAILED tests/unit/test_langchain.py::test_chat_model_input_keeps_message_names
E         At index 0 diff: ('user', None) != ('user', 'alice')
1 failed, 29 passed in 3.94s
# not run: the rest of tests/unit, mypy, and a live provider

non-blocking: _convert_message_to_dict also builds the generation output at CallbackHandler.py:1309. A named AIMessage from the model now keeps its name there too. That looks right to me, but no test pins it. One assertion on the output span would cover it.

@hassiebp when both are set, message.name now wins over additional_kwargs["name"]. I'd keep that order. The old value still shows, because additional_kwargs is copied in whole. Does any integration you know read name from the old place?

[mosafariuk]

@hassiebp could you take a look when you have a moment? It's approved by @chrikrah, CI is green and the CLA is signed. It's a small fix: the LangChain handler now reads message.name, falling back to additional_kwargs["name"], so message names reach the generation input again.

[hassiebp]

thanks for your contribution!