Langfuse - Muster

How it works

If your team is already using Langfuse for LLM tracing, the muster Langfuse connector gives you immediate coverage with zero additional code. The connector:

Discovers agent names from existing Langfuse traces
Streams completed traces and maps them to muster quality signals
Extracts token usage from Langfuse generation observations automatically

Elitery deploys and manages the connector. Your developers do nothing.

What Elitery needs from you

LANGFUSE_HOST=https://cloud.langfuse.com   # or your self-hosted URL
LANGFUSE_PUBLIC_KEY=pk-lf-...
LANGFUSE_SECRET_KEY=sk-lf-...
LANGFUSE_PROJECT_ID=<optional, defaults to all projects>

What your developers do

Nothing — if you’re already tracing to Langfuse, you’re done. If you’re not yet using Langfuse but want to add tracing (which also feeds muster), add the Langfuse SDK to your agents:

from langfuse.decorators import observe, langfuse_context

@observe()
def run_agent(input: str) -> str:
    # Your agent logic here
    result = llm.invoke(input)
    return result

This gives Langfuse — and therefore muster — full execution visibility including prompts, completions, and token usage.