[bug] Python LangChain 1.0 instrumentation error in _extract_message_role(): expected list, found <class 'NoneType'>

[scribu]

Describe the bug

ERROR:openinference.instrumentation.langchain._tracer:Failed to get attribute.
Traceback (most recent call last):
  File "/Users/scribu/my-project/.venv/lib/python3.13/site-packages/openinference/instrumentation/langchain/_tracer.py", line 324, in wrapper
    yield from wrapped(*args, **kwargs)
  File "/Users/scribu/my-project/.venv/lib/python3.13/site-packages/openinference/instrumentation/langchain/_tracer.py", line 574, in _extract_message_role
    assert isinstance(id_, List), f"expected list, found {type(id_)}"
           ~~~~~~~~~~^^^^^^^^^^^
AssertionError: expected list, found <class 'NoneType'>

Steps To Reproduce

Pass a raw dict as a message to a Langchain agent:

from langchain.agents import create_agent
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage

model = ChatOpenAI()

agent = create_agent(model=model)

# This works fine
# message = HumanMessage(content="What's the weather in London?")

# This throws the assertion error
message = {"role": "user", "content": "What's the weather in London?"}

result = agent.invoke({"messages": [message]})

output = result["messages"][-1].content

print(output)

Note that using HumanMessage() instead of a raw dict works

Versions

openai==2.7.2
langchain==1.0.5
langchain-openai==1.0.2
openinference-instrumentation==0.1.42
openinference-instrumentation-langchain==0.1.54

[dosubot]

Hi @scribu! I'm Dosu and I’m helping the openinference team.

This is a known compatibility issue between OpenInference instrumentation and LangChain 1.0. The error happens because the instrumentation expects the message id field to always be a list, but in LangChain 1.0, this field can be missing or None, which triggers the assertion error you’re seeing. This has been reported by other users as well and is tracked in the OpenInference repo (issue #1356, issue #2314).

A workaround is to patch the _parse_message_data (or _extract_message_role) function to check if id_ is a list before using it, and handle the case where it’s None or missing. For example, you could replace the assertion with:

id_ = message_data.get("id", [])
if not isinstance(id_, list):
    # handle gracefully, e.g., skip or set a default role
    return
if not id_:
    # handle empty list case
    return

Recent PRs have updated some compatibility for LangChain 1.0, but the strict type check for id_ still exists in the released code (PR #2336). Until an official fix is released, patching locally or pinning to a pre-1.0 version of LangChain are your best options.

Let me know if you want a more detailed patch or have questions about where to apply it!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[caroger]

potentially the same root cause for an issue reported on AX side and would be resolved by #2412. I will investigate and confirm