Langfuse¶

Open-source LLM observability platform for tracing, evaluating, and monitoring AI applications. Provides detailed visibility into LLM calls, token usage, latency, and costs.

Overview¶

Langfuse is the default observability backend for the starter kit: - LLM Tracing: Trace every LLM call with input/output, tokens, latency - Cost Tracking: Monitor spend per model, user, or feature - Evaluations: Score traces with manual or automated evaluations - Prompt Management: Version and manage prompts - Datasets: Create test datasets for evaluation - Dashboards: Built-in analytics and monitoring - S3 Storage: Uses Terraform-provisioned S3 bucket for blob storage

Installation¶

Prerequisites¶

Configure in .env:

LANGFUSE_USERNAME=admin
LANGFUSE_PASSWORD=your-password
LANGFUSE_PUBLIC_KEY=pk-lf-xxxxxxxx
LANGFUSE_SECRET_KEY=sk-lf-xxxxxxxx

Install¶

./cli o11y langfuse install

The installer: 1. Provisions an S3 bucket via Terraform for blob storage 2. Adds the Langfuse Helm chart repository 3. Renders Helm values with credentials and bucket configuration 4. Deploys Langfuse to the langfuse namespace

Verification¶

# Check pods
kubectl get pods -n langfuse

# Check service
kubectl get svc -n langfuse

# Port-forward for UI access
kubectl port-forward svc/langfuse-web 3000:3000 -n langfuse --address 0.0.0.0 &

# Access Langfuse UI
open http://localhost:3000

Log in with LANGFUSE_USERNAME and LANGFUSE_PASSWORD.

Integration¶

LiteLLM¶

LiteLLM automatically sends traces to Langfuse when installed. Traces include: - Request/response pairs - Token usage per request - Latency breakdown - Model routing decisions

OpenClaw¶

OpenClaw sends agent execution traces to Langfuse when LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY are configured.

Custom Applications¶

from langfuse import Langfuse

langfuse = Langfuse(
    public_key="pk-lf-xxx",
    secret_key="sk-lf-xxx",
    host="http://langfuse-web.langfuse:3000"
)

trace = langfuse.trace(name="my-app")
generation = trace.generation(
    name="chat",
    model="qwen3-30b-instruct-fp8",
    input=[{"role": "user", "content": "Hello"}],
    output="Hi there!"
)

Configuration¶

Environment Variables¶

Infrastructure¶

Langfuse uses Terraform to provision: - S3 bucket for blob storage (traces, datasets) - IAM roles via IRSA for S3 access

Troubleshooting¶

UI not loading¶

# Check web pod status
kubectl get pods -n langfuse -l app=web

# Check logs
kubectl logs -n langfuse -l app=web

Traces not appearing¶

# Check LiteLLM integration
kubectl logs -n litellm -l app=litellm | grep langfuse

# Verify Langfuse credentials
echo $LANGFUSE_PUBLIC_KEY
echo $LANGFUSE_SECRET_KEY

Database issues¶

# Check database pod
kubectl get pods -n langfuse -l app=postgresql

# Check database logs
kubectl logs -n langfuse -l app=postgresql

Learn More¶

• Langfuse Documentation
• Langfuse GitHub
• Langfuse Python SDK
• LiteLLM Integration