langfuse-optimization

What problem does it solve?

Debugging and optimizing AI workflow configurations (style.yaml, template.yaml, tools.yaml) based on Langfuse traces is a manual, iterative, and often frustrating process. This skill automates the identification of problems and generates precise, actionable YAML fixes, saving you countless hours of manual troubleshooting.

Core Features & Use Cases

• Trace-based Problem Extraction: Automatically identifies style check failures, template violations, and tool selection issues directly from Langfuse traces.
• Surgical YAML Fixes: Generates exact before and after YAML code blocks with clear rationales for each recommended change, ensuring easy implementation.
• Optimized Data Retrieval: Uses --filter-essential and --filter-all flags to drastically reduce trace data size (up to 99%), making analysis faster and more efficient.
• Use Case: If your AI-generated content consistently fails a tone_consistency check, this skill will analyze the traces, pinpoint the vague rubric in style.yaml, and provide the exact YAML update needed to clarify the grading criteria, improving content quality with minimal effort.

Quick Start

Retrieve the last 5 traces for case_id=0001 from Langfuse, applying essential filtering to reduce data size. Analyze these traces to identify any issues in style.yaml, template.yaml, or tools.yaml and provide specific YAML fixes.