Langfuse

Langfuse is an open-source LLM engineering platform for tracing, evaluation, prompt management, and metrics. Connect it to EUrouter to automatically capture every request as a Langfuse trace.

Setup

Get your Langfuse credentials

Sign in to Langfuse (or your self-hosted instance)
Go to Settings in your project
Copy your Secret Key and Public Key

For more details, see Where are my Langfuse API keys?

Add the destination in EUrouter

Go to Broadcast in your dashboard
Click Add Destination next to Langfuse
Enter your Secret Key and Public Key
If you're self-hosting, update the Base URL (default: https://cloud.langfuse.com)
Click Test Connection to verify
Click Create

Configuration

Field	Required	Description
Secret Key	Yes	Your Langfuse secret key (`sk-lf-...`)
Public Key	Yes	Your Langfuse public key (`pk-lf-...`)
Base URL	No	Custom endpoint for self-hosted Langfuse. Default: `https://cloud.langfuse.com`

What you'll see in Langfuse

Once connected, each EUrouter request appears as a trace in Langfuse with:

Model name and provider
Input/output messages (unless Privacy Mode is on)
Token usage and cost
Latency and timing breakdown
Any custom metadata passed via the trace field

Custom metadata

You can enrich Langfuse traces by including a trace field in your EUrouter requests:

{
  "model": "gpt-5.1",
  "messages": [{ "role": "user", "content": "Hello!" }],
  "trace": {
    "trace_id": "workflow_123",
    "trace_name": "Customer Support",
    "generation_name": "Initial Response",
    "user_id": "user_456",
    "session_id": "session_789"
  }
}

Langfuse recognizes these keys:

Key	Description
`trace_id`	Group multiple requests into a single Langfuse trace
`trace_name`	Display name for the trace (defaults to model name)
`generation_name`	Name for the specific LLM generation
`user_id`	Associate the trace with a specific user
`session_id`	Group traces into a Langfuse session