Langfuse LLM Observability

Instrument LLM apps with Langfuse for tracing, prompt management, evaluation, and datasets. Captures nested traces across LangChain, LlamaIndex, and OpenAI calls to debug latency, cost, and quality in production.

This skill adds observability to LLM applications with Langfuse. It instruments nested traces and spans across chains and agents, manages versioned prompts, builds evaluation datasets and scores, tracks token cost and latency, and integrates with LangChain, LlamaIndex, and the OpenAI SDK for production debugging.

When to use

Use when you need to trace, debug, evaluate, or monitor an LLM application in production — tracking cost, latency, prompt versions, and output quality.

Examples

Trace an agent

Capture nested LLM calls

Instrument my LangChain agent with Langfuse so I can see per-step traces, token cost, and latency for each tool call

Evaluate outputs

Score responses on a dataset

Set up a Langfuse dataset and run an evaluation scoring my RAG answers for relevance and faithfulness