Query Langfuse data via CLI and retrieve Langfuse documentation for LLM observability workflows.
npx skills add https://github.com/celeroncoder/skills --skill langfuse-celeroncoder
Please help me install this Agent Skill. Skill: langfuse Source: https://github.com/celeroncoder/skills/tree/main/langfuse Command: npx skills add https://github.com/celeroncoder/skills --skill langfuse-celeroncoder
npx langfuse-cli with schema discovery and credential setup via environment variables. - Documentation Retrieval: Access Langfuse docs through the llms.txt index, markdown page fetching, and a search API that also indexes GitHub issues and discussions. - Use-Case Reference Guides: Follow dedicated guides for instrumenting applications, migrating prompts to Langfuse, capturing user feedback as scores, upgrading SDKs, and performing systematic error analysis on traces. - Use Case: You want to debug why your LLM app produces bad outputs. The Skill pulls traces via the CLI, guides you through the 5-step error analysis process, creates annotation queues, and links you directly to the Langfuse UI. ## Quick Start Ask the agent to list your recent Langfuse traces or fetch the Langfuse documentation page for the feature you are working on.High-intent search queries and answers about installing and using this skill.
Use `npx langfuse-cli api traces list` after setting LANGFUSE_PUBLIC_KEY, LANGFUSE_SECRET_KEY, and LANGFUSE_HOST environment variables. Run `npx langfuse-cli api __schema` to discover all available resources and `--help` on any resource or action for arguments.
Scan your codebase for prompts, convert variables to {{var}} double-brace syntax, create prompts via `langfuse.create_prompt()` with a production label, then refactor code to use `langfuse.get_prompt()` with `.compile()`. Complex Jinja logic like conditionals must move into application code.
Yes, feedback is stored as scores on traces. Use LangfuseWeb with only the public key in frontend code for explicit feedback, or `langfuse.create_score()` server-side for implicit signals like retries or copies. Always set dataType explicitly for boolean scores.
Fetch the official migration guide at langfuse.com first, then replace start_span/start_generation with start_observation, replace update_current_trace with propagate_attributes, and move release and environment settings to environment variables. Enable debug logging to catch dropped spans.
Common causes include missing `langfuse.flush()` before script exit, importing Langfuse before environment variables load, or importing the OpenAI client before Langfuse setup. Verify credentials with a test call to /api/public/projects and check for a 200 status.
Langfuse only supports simple {{variable}} substitution with no conditionals, loops, or filters. Pre-compute conditional content and formatted lists in your code and pass them as variables, or compile templates client-side with Jinja2 at the cost of losing Playground preview.
bilacchiInstrument LLM applications with tracing, prompt management, and evaluation using Langfuse.
adamatdevopsAnalyze LLM pipeline traces to prioritize failure modes by root cause.
Arize-aiRun, read, and compare dataset-backed experiments to evaluate prompt and pipeline changes.
tankpkgDiagnose LLM observability and evaluation gaps across production tracing and prompt workflows.