how to customize trace id when using callback handler in langchain integration

[shawnzhu]

When following the Customize IDs section, it supports changing trace id for a function decorated with @observe().

How to do the same thing without the decorator?

[dosubot[bot]]

Hey there, @shawnzhu! I'm here to help you with any bugs, questions, or contributions you have while waiting for a human maintainer. Let's dive in and solve this together!

To customize the trace ID when using a callback handler in Langchain integration without using the @observe() decorator, you can pass the userId as a parameter to the CallbackHandler constructor:

import { CallbackHandler } from "langfuse-langchain/src/callback";

// Create a callback handler with a custom user ID
const langfuseHandler = new CallbackHandler({ userId: "your-user-id", sdkIntegration: "your-sdk-integration" });

You can then use this handler in your Langchain invocations:

chain.invoke({ person: "obama", language: "spanish" }, { callbacks: [langfuseHandler] });

This approach allows you to set a custom trace ID without using the @observe() decorator ^{[1][2][3][4][5]}.

To continue talking to Dosu, mention @dosu.

[shawnzhu]

it could not customize trace id like this when using the low level API.

[shawnzhu]

I figured it out by customizing run_id of a Langchain runnable, which is later used as trace id.

from langfuse.callback import CallbackHandler

run_id = uuid4()
langfuse_handler = CallbackHandler();
agent.invoke({"input": input}, config={"callbacks":[langfuse_handler], "run_id": run_id})

Notice that it will also changed the run_id of this langchain runnable as well.

[marcklingen]

this is great! alternatively, you could use the low-level sdk to create a trace and get a langchain callback handler for it:

# Intialize Langfuse Client
from langfuse import Langfuse
langfuse = Langfuse()
 
# Get Langchain handler for a trace
trace = langfuse.trace()
langfuse_handler_trace = trace.get_langchain_handler(
   update_parent=True # add i/o to trace itself as well
)
 
# Add Langfuse handler as callback (classic and LCEL)
chain.invoke({"input": "<user_input>"}, config={"callbacks": [langfuse_handler_trace]})

[marctorsoc]

FWIW, this works too

langfuse_handler = CallbackHandler();
agent.invoke({"input": input}, config={"callbacks":[langfuse_handler], "run_name": specific_name_for_this_input})