docker-compose.yml

# Local development only: this is not a production deployment manifest.
services:

  ollama:
    image: ollama/ollama
    ports:
      - "127.0.0.1:11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
      interval: 10s
      timeout: 5s
      retries: 12
      start_period: 30s
    restart: unless-stopped

  ollama-init:
    image: ollama/ollama
    depends_on:
      ollama:
        condition: service_healthy
    volumes:
      - ollama_data:/root/.ollama
    entrypoint: ["ollama", "pull", "qwen2.5:7b"]
    environment:
      - OLLAMA_HOST=http://ollama:11434
    restart: "no"

  postgres:
    image: postgres:16-alpine
    restart: unless-stopped
    environment:
      POSTGRES_DB: rag_assistant
      POSTGRES_USER: rag
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-rag_dev_password}
    ports:
      - "127.0.0.1:5432:5432"
    volumes:
      - pgdata:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U rag -d rag_assistant"]
      interval: 10s
      timeout: 5s
      retries: 5

  redis:
    image: redis:7-alpine
    restart: unless-stopped
    ports:
      - "127.0.0.1:6379:6379"
    volumes:
      - redisdata:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 5

  jaeger:
    image: jaegertracing/all-in-one:1.59
    restart: unless-stopped
    ports:
      - "127.0.0.1:16686:16686"
      - "127.0.0.1:4317:4317"
      - "127.0.0.1:4318:4318"

  app:
    build: .
    ports:
      - "127.0.0.1:8000:8000"
    env_file:
      - .env
    environment:
      - RAG_ENV=development
      - OLLAMA_BASE_URL=http://ollama:11434
      - DATABASE_URL=postgresql://rag:${POSTGRES_PASSWORD:-rag_dev_password}@postgres:5432/rag_assistant
      - REDIS_URL=redis://redis:6379/0
      - OTEL_ENABLED=${OTEL_ENABLED:-false}
      - OTEL_EXPORTER_OTLP_ENDPOINT=http://jaeger:4317
      - OTEL_SERVICE_NAME=rag-support-assistant
    volumes:
      - ./data:/app/data
    depends_on:
      ollama-init:
        condition: service_completed_successfully
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
    restart: unless-stopped

  # Exactly one Celery ingestion worker (concurrency 1). Shares /app/data with
  # the app. Not a second Uvicorn web process — web stays --workers 1 / one
  # replica until session/confirm-action state is externalised.
  # Also executes plan §4.6 escalation outbox retry tasks when beat schedules them.
  worker:
    build: .
    command:
      - celery
      - -A
      - tasks.celery_app:celery_app
      - worker
      - --concurrency=1
      - --hostname=ingest@%h
      - --loglevel=INFO
    env_file:
      - .env
    environment:
      - RAG_ENV=development
      - OLLAMA_BASE_URL=http://ollama:11434
      - DATABASE_URL=postgresql://rag:${POSTGRES_PASSWORD:-rag_dev_password}@postgres:5432/rag_assistant
      - REDIS_URL=redis://redis:6379/0
      - OTEL_ENABLED=${OTEL_ENABLED:-false}
      - OTEL_EXPORTER_OTLP_ENDPOINT=http://jaeger:4317
      - OTEL_SERVICE_NAME=rag-support-assistant
      - RAG_OUTBOX_RETRY_BEAT=${RAG_OUTBOX_RETRY_BEAT:-true}
      - RAG_OUTBOX_RETRY_INTERVAL_SEC=${RAG_OUTBOX_RETRY_INTERVAL_SEC:-300}
      - RAG_OUTBOX_RETRY_BATCH_LIMIT=${RAG_OUTBOX_RETRY_BATCH_LIMIT:-50}
    volumes:
      - ./data:/app/data
    depends_on:
      ollama-init:
        condition: service_completed_successfully
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
    restart: unless-stopped
    # Long warm-shutdown: no Celery task_time_limit is configured in-repo, and
    # embedding/indexing a large document can run for many minutes.
    stop_grace_period: 3600s
    healthcheck:
      test: ["CMD", "python", "-m", "tasks.worker_health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

  # Plan §4.6: Celery beat schedules escalation outbox retry (failed inbox
  # deliveries). One replica only — multiple beat processes duplicate schedules.
  # Operator/cron alternative: python scripts/outbox_retry.py
  worker-beat:
    build: .
    command:
      - celery
      - -A
      - tasks.celery_app:celery_app
      - beat
      - --loglevel=INFO
      - --pidfile=
      - --schedule=/tmp/celerybeat-schedule
    env_file:
      - .env
    environment:
      - RAG_ENV=development
      - DATABASE_URL=postgresql://rag:${POSTGRES_PASSWORD:-rag_dev_password}@postgres:5432/rag_assistant
      - REDIS_URL=redis://redis:6379/0
      - RAG_OUTBOX_RETRY_BEAT=${RAG_OUTBOX_RETRY_BEAT:-true}
      - RAG_OUTBOX_RETRY_INTERVAL_SEC=${RAG_OUTBOX_RETRY_INTERVAL_SEC:-300}
      - RAG_OUTBOX_RETRY_BATCH_LIMIT=${RAG_OUTBOX_RETRY_BATCH_LIMIT:-50}
    depends_on:
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
      worker:
        condition: service_started
    restart: unless-stopped

volumes:
  ollama_data:
  pgdata:
  redisdata: