langfuse-integration git:20260507.50823d8A

Immutable. This exact content is served forever at /api/v1/blob/3618fa23911ab03b.

---
name: langfuse-integration
description: LangFuse LLM observability integration for tracing, analytics, and cost tracking
allowed-tools:
  - Read
  - Write
  - Edit
  - Bash
  - Glob
  - Grep
graph:
  domains: [domain:software-engineering]
  specializations: [specialization:ai-agents-conversational]
  skillAreas: [skill-area:natural-language-processing]
  roles: [role:ml-engineer, role:backend-engineer]
  workflows: [workflow:ml-model-lifecycle, workflow:feature-development]

---

# LangFuse Integration Skill

## Capabilities

- Set up LangFuse tracing for LLM calls
- Configure cost tracking and analytics
- Implement prompt management
- Set up evaluation datasets
- Design custom trace metadata
- Create dashboards and alerts

## Target Processes

- llm-observability-monitoring
- cost-optimization-llm

## Implementation Details

### Core Features

1. **Tracing**: Track LLM calls, chains, and agents
2. **Prompts**: Version and manage prompts
3. **Analytics**: Usage, latency, cost metrics
4. **Datasets**: Evaluation and testing data
5. **Scores**: Track output quality

### Integration Methods

- LangChain callback handler
- Direct SDK integration
- OpenAI drop-in replacement
- Decorator-based tracing

### Configuration Options

- Public/secret keys
- Host URL (cloud or self-hosted)
- Sampling rate
- Metadata configuration
- User tracking

### Best Practices

- Consistent trace naming
- Meaningful metadata
- Regular prompt versioning
- Set up alerting

### Dependencies

- langfuse
- langchain (for callback integration)