Instruments Python LLM pipelines with Langfuse tracing, observability patterns, and performance scoring.
This skill streamlines the process of adding observability to AI applications by guiding users through Langfuse SDK configuration, environment validation, and the implementation of best-practice tracing patterns. It provides specialized templates for various architectures—including RAG, agentic workflows, and multi-model pipelines—ensuring that LLM calls, tool executions, and spans are correctly captured. By identifying anti-patterns and offering pre-built scoring modules, it helps developers gain deep insights into their AI pipeline performance and debug SDK configurations effectively within the Claude Code environment.
Key Features
01Anti-pattern detection for LLM instrumentation
02Automated environment and SDK validation scripts
03Implementation of custom scoring and evaluation modules
04Architecture-specific tracing templates for RAG and Agents
050 GitHub stars
06Verification testing with automated trace run validation
Use Cases
01Adding performance scoring to evaluate LLM output quality
02Debugging missing traces in complex agentic workflows
03Setting up production-grade observability for a RAG application
