Telemetry

Langfuse integration for AI call tracing

Endpoints

MethodPathDescription
GET/organizations/langfuse-credentialsGet current Langfuse configuration
POST/organizations/langfuse-credentialsSave Langfuse credentials
DELETE/organizations/langfuse-credentialsRemove Langfuse credentials

What's Captured Per Call

  • Full conversation transcript as Langfuse spans
  • LLM provider, model, prompt, completion, and token counts per turn
  • Tool call inputs and outputs
  • Node path traversed
  • Call duration and termination reason
  • Extracted data fields

Setup

Step 1: Create a Langfuse project

Sign up at https://cloud.langfuse.com or deploy self-hosted. Go to Settings → API Keys.

Step 2: Save credentials to Sysevo

bash
curl -X POST https://api.sysevo.io/api/v1/organizations/langfuse-credentials \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "host": "https://cloud.langfuse.com",
    "public_key": "pk-lf-xxxxxxxxxxxxxxxx",
    "secret_key": "sk-lf-xxxxxxxxxxxxxxxx"
  }'

Or in the portal: Developers → Telemetry.

Step 3: Verify

Place a test call. Within 30 seconds the trace should appear in your Langfuse dashboard under the project.

Remove Credentials

DELETE
/organizations/langfuse-credentials

Disables tracing immediately — subsequent calls are not exported.

Was this page helpful?
Open Dashboard →