pytracer-langgraph

pyTracer adapter for LangGraph and LangChain. It auto-instruments an agent through LangChain's callback system, so you do not add @op by hand: model calls, tool calls, and graph nodes are captured as pyTracer spans automatically.

import pytracer_sdk as pytracer
from pytracer_langgraph import PyTracerCallbackHandler

pytracer.configure(
    pytracer.HTTPTransport("https://gateway.internal", api_key="YOUR_KEY"),
    service_name="my-agent", branch="main", commit="abc123",
)

handler = PyTracerCallbackHandler()
app.invoke(inputs, config={"callbacks": [handler]})   # spans captured for this run

on_chat_model_start / on_llm_start become LLM spans, on_tool_start become tool spans, and LangGraph nodes become agent spans, all linked into one trace by the run tree.