Tracing in LangChain

As organizations mature with XLMs(x=Large, Medium, Small), the deployment architecture will be more agentic in nature. That agentic architecture will be similar to a distributed architecture. Tracing is a very important feature in such architectures. This video demonstrates "Tracing", an upcoming feature in LangChain.

I forgot to mention in the video how to instrument the code to record the trace in the tracing server. For that, we just need the below three lines of code at the top of the agent program

import os
os.environ["LANGCHAIN_TRACING"] = "true"
os.environ["LANGCHAIN_SESSION"] = "agent_chain" # This is session