← Back to Documentation

Deployment

Production deployment options.

You will learn: - Direct vs sidecar deployment - Environment configuration - Platform-specific setup

Deployment Options

Option Best For Latency
Direct Most apps ~50ms
Sidecar High-throughput <1ms

Direct (Default)

Traces go straight to Kalibr cloud. No extra infrastructure.

import kalibr
import os

os.environ["KALIBR_API_KEY"] = "sk_live_..."
os.environ["KALIBR_COLLECTOR_URL"] = "https://api.kalibr.systems/api/ingest"

Good for: Most applications, <10k traces/day.

Sidecar

For high-throughput apps. Buffers and batches traces locally.

# docker-compose.yml
services:
  app:
    environment:
      - KALIBR_COLLECTOR_URL=http://sidecar:8080
    depends_on:
      - sidecar

  sidecar:
    image: ghcr.io/kalibr-ai/kalibr-sidecar:latest
    environment:
      - KALIBR_API_KEY=sk_live_...

Environment Variables

Variable Required Description
KALIBR_API_KEY Yes Your API key
KALIBR_COLLECTOR_URL Yes* Where to send traces
KALIBR_WORKFLOW_ID No Default workflow
KALIBR_TENANT_ID No Default customer
KALIBR_ENVIRONMENT No prod/staging/dev

For cloud: https://api.kalibr.systems/api/ingest For sidecar: http://localhost:8080 or http://sidecar:8080

Kubernetes

containers:
  - name: app
    env:
      - name: KALIBR_API_KEY
        valueFrom:
          secretKeyRef:
            name: kalibr-secrets
            key: api-key

Troubleshooting

Problem Solution
Traces not appearing Check KALIBR_API_KEY
High latency Use sidecar deployment
Import order import kalibr before LLM SDKs

Next Steps

← Back to Documentation