Langfuse Prompt and Trace Debugger

About

Provides a mandatory set of tools for debugging AI interactions by connecting directly to Langfuse. Instead of guessing at prompt schemas, this skill allows for fetching the exact prompt text and configuration from both staging and production environments, which is critical for resolving `KeyError` and validation issues. It includes scripts to view specific traces, find recent errors, analyze performance, and understand the precise inputs and outputs of AI models, enabling data-driven investigation of application behavior and production incidents.

Key Features

• Mandatory usage for `KeyError` and schema validation failures to prevent guesswork.
• Fetches and caches live prompt schemas and configurations from separate staging and production servers.
• Provides command-line tools to fetch, view, and search Langfuse traces for debugging AI workflows.
• Includes proactive debugging workflows for production issues, performance analysis, and response validation.
• Offers scripts to reconstruct full, untruncated prompts for deep-dive analysis.
• 7 GitHub stars

Use Cases

• Resolving a `KeyError` in a test by fetching the actual prompt schema from Langfuse instead of guessing field names.
• Investigating a production incident where a user received an incorrect AI response by fetching and analyzing the specific error trace.
• Analyzing high API costs by fetching traces within a specific time window to identify slow or expensive prompts.