LangChain & LangSmith

Two integrations let existing LangChain and LangSmith code report to langprobe with minimal change.

LangChain / LangGraph

Install the langchain extra and drop in a callback handler:

pip install "langprobe[langchain]"

from langprobe.integrations.langchain import LangprobeCallbackHandler

handler = LangprobeCallbackHandler()

# pass it wherever LangChain takes callbacks:
chain.invoke(inputs, config={"callbacks": [handler]})

The handler bridges LangChain / LangGraph runs onto langprobe's run and span model — you keep your chains and graphs as-is.

LangSmith-compat shim

If you already write LangSmith-shaped code (traceable, the Client), the langsmith extra re-exports the same names so migration is a one-line import change:

pip install "langprobe[langsmith]"

from langprobe.integrations.langsmith import Client, traceable

@traceable
def my_agent(prompt: str) -> str:
    ...

This is the code-side counterpart to the env-var migration in Migrate from LangSmith — either re-point your client with LANGSMITH_ENDPOINT, or switch the import here.