Skip to main content

Railway Deployment

Railway works best by building the existing server/Dockerfile and attaching managed Postgres, Redis, and an Ollama-compatible embedding endpoint.

  • server service from server/Dockerfile
  • worker service from the same image
  • worker-beat service from the same image
  • Railway Postgres
  • Railway Redis
  • External Ollama-compatible endpoint, or swap providers through env vars

Working railway.json​

{
  "$schema": "https://railway.app/railway.schema.json",
  "build": {
    "builder": "DOCKERFILE",
    "dockerfilePath": "server/Dockerfile"
  },
  "deploy": {
    "numReplicas": 1,
    "restartPolicyType": "ON_FAILURE",
    "restartPolicyMaxRetries": 10
  }
}

Service commands​

server: uvicorn app.main:app --host 0.0.0.0 --port $PORT
worker: celery -A app.celery_app worker --loglevel=info --concurrency=2
worker-beat: celery -A app.celery_app beat --loglevel=info

Required environment​

DATABASE_URL=${{Postgres.DATABASE_URL}}
REDIS_URL=${{Redis.REDIS_URL}}
SECRET_KEY=replace-me
EMBEDDING_PROVIDER=ollama
OLLAMA_BASE_URL=https://your-ollama-endpoint
OLLAMA_EMBEDDING_MODEL=nomic-embed-text

Run alembic upgrade head once after the first deploy.