Langfuse Skill

When to Use

Use this skill when you need to:

• Setup Langfuse — Cloud or self-hosted observability platform
• Add observability — Instrument an LLM application with tracing and monitoring
• Trace LLM — Record LLM calls, inputs, outputs, latency, and costs
• LLM monitoring — Track production performance, errors, and usage patterns
• Evaluate LLM outputs with scores and datasets
• Manage prompts with version control

What It Covers

1. Install and configure Langfuse (cloud or self-hosted)
2. Trace LLM calls manually with the Python SDK
3. Integrate with LangChain using the Langfuse callback handler
4. Integrate with LangGraph for agent and workflow tracing
5. Create scores and evaluations — manual, automated, and LLM-as-a-Judge
6. Manage prompts with version control and programmatic retrieval

Templates

This skill provides ready-to-use templates:

Examples

• examples/full-observability.py — Complete example with traces, spans, generations, and scores

Resources

• Langfuse Documentation
• Langfuse GitHub
• references/tracing-guide.md — Core concepts: traces, spans, generations, attributes
• references/evaluation.md — Scores, LLM-as-a-Judge, datasets
• references/self-hosting.md — Docker Compose setup with Postgres and Clickhouse

Use When

The user asks anything about: setup langfuse, add observability, trace LLM, LLM monitoring, instrument LLM app, evaluate LLM output, manage prompts, self-host langfuse, langchain observability, langgraph tracing, or debug LLM behavior.