Langfuse setup and the OpenAI drop-in

Langfuse is a tracing backend for LLM applications. It stores traces, generations, and scores, and gives you a UI to search and read them. The Python SDK, version three, sits on OpenTelemetry, which is why nesting and context propagation work the way they do in later lessons.

The fastest way in is the OpenAI drop-in. You import openai from langfuse.openai instead of from openai, and every call the client makes is recorded: the model, the full input, the output, token counts, cost, and latency. Nothing else in your code changes.