Inconsistent Message ID Formats with LangGraph and LangChain Integration

·Sep 17, 2025 09:29 AM·

I'm using openinference-instrumentation-langchain==0.1.51 with LangGraph and running into message ID format issues that seem inconsistent. When my LangGraph creates messages with different ID formats, I get different errors from the instrumentation: 1. When I set message id as an empty list []: I get "IndexError: list index out of range" 2. When I set it as a list with a string ["user_msg"]: I get "AttributeError: 'list' object has no attribute 'startswith'" 3. When I set it as just a string "user_msg": I get "AssertionError: expected list, found <class 'str'>" It seems like the instrumentation code expects the ID to be a list in some places but then tries to call string methods on it in other places. Is there a correct message ID format for LangGraph integration? Or is this a known compatibility issue between the LangChain instrumentation and LangGraph?

4 comments

· Sorted by Oldest

RunLLM
APP
·
·
This is a known compatibility issue: the OpenInference LangChain instrumentation has had inconsistent handling of message `id` formats, especially with LangGraph, leading to errors when the `id` is a string, list, or empty list. The correct format expected by the instrumentation has changed over time, and recent versions have fixed message parsing for LangGraph, but version 0.1.51 still has these issues. Upgrading to at least openinference-instrumentation-langchain==0.1.52 or later is recommended, as recent changelogs note bug fixes for message parsing in LangGraph (<https://github.com/Arize-ai/openinference/blob/main/python/instrumentation/openinference-instrumentation-langchain/CHANGELOG.md|see changelog>). If you must use 0.1.51, there is no single "correct" id format that will work in all cases due to these bugs. The errors you see are a result of the instrumentation inconsistently expecting either a list or a string and then applying the wrong logic. Upgrading is the only reliable solution (<https://github.com/Arize-ai/phoenix/issues/7441|see related issue>). Would you like more detail on the specific code paths or a workaround for your current version? Sources: - <https://github.com/Arize-ai/openinference/blob/main/python/instrumentation/openinference-instrumentation-langchain/CHANGELOG.md|openinference-instrumentation-langchain CHANGELOG.md> - <https://github.com/Arize-ai/phoenix/issues/7441|phoenix/issues/7441>
💡 Hint: Mention RunLLM in the thread for followups.
GGhiasi
·
·
It still has the bug even with the update
Roger Y.
·
·
Can you provider a minimal reproducible example?
RunLLM
APP
·
·
If you tell me the right answer to this question, I'll make sure I don't make the same mistake twice!
Feedback
Type the right answer here...
↵ Please 'enter' to submit

Inconsistent Message ID Formats with LangGraph and LangChain Integration | Arize AI Community