What problem does it solve? Debugging LLM applications requires inspecting real traces, generations, and token usage, but Langfuse's API surface changed significantly in v4, breaking naive integrations and making manual querying error-prone. ## Core Features & Use Cases - Trace and Generation Inspection: List traces, view full trace timelines, and inspect LLM generations with model, prompt version, token usage, and tool calls. - Automatic API Generation Detection: Detects whether the server runs the legacy v1 API or the v4 read APIs and routes queries accordingly, including cursor-based pagination. - Documentation Access: Fetch Langfuse documentation via llms.txt, markdown pages, and search to guide SDK integrations. - Use Case: When an agent misbehaves in production, run langfuse_query.sh generations <trace-id> to see exactly what the LLM received and returned instead of adding print statements. ## Quick Start Ask the assistant to list the five most recent Langfuse traces and show the generations for the one that failed.
