version: '3.8'

services:
  inference-pipeline:
    build:
      context: .
      dockerfile: Dockerfile
      target: production
    container_name: inference-pipeline
    ports:
      - "8080:8080"
    volumes:
      - ./data:/app/data
      - ~/.kaggle:/home/appuser/.kaggle:ro
    environment:
      - PYTHONPATH=/app
      - MODEL_TYPE=distilbert
      - PYTHONUNBUFFERED=1
      - OMP_NUM_THREADS=4
      - MKL_NUM_THREADS=4
    command: python -m inference_pipeline.main serve --host 0.0.0.0 --port 8080
    healthcheck:
      test: ["CMD", "python", "/app/scripts/health_check.py", "--host", "localhost", "--port", "8080"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 60s
    restart: unless-stopped
    mem_limit: 2g
    cpus: '2.0'
    
  prometheus:
    image: prom/prometheus:latest
    container_name: inference-prometheus
    ports:
      - "9090:9090"
    volumes:
      - ./monitoring/prometheus.yml:/etc/prometheus/prometheus.yml
      - prometheus-data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.path=/prometheus'
      - '--web.console.libraries=/etc/prometheus/console_libraries'
      - '--web.console.templates=/etc/prometheus/consoles'
    restart: unless-stopped
    profiles:
      - monitoring
    
  grafana:
    image: grafana/grafana:latest
    container_name: inference-grafana
    ports:
      - "3000:3000"
    environment:
      - GF_SECURITY_ADMIN_PASSWORD=admin
    volumes:
      - grafana-data:/var/lib/grafana
    restart: unless-stopped
    profiles:
      - monitoring

  # Optional: Batch processing mode
  batch-processor:
    build:
      context: .
      dockerfile: Dockerfile
      target: production
    container_name: inference-batch
    volumes:
      - ./data:/app/data
      - ~/.kaggle:/home/appuser/.kaggle:ro
    environment:
      - PYTHONPATH=/app
      - OMP_NUM_THREADS=4
      - MKL_NUM_THREADS=4
    command: >
      sh -c "python -m inference_pipeline.main run 
             --input data/input.csv 
             --output data/output.csv 
             --model-type vader 
             --batch-size 1000"
    restart: "no"
    mem_limit: 2g
    cpus: '2.0'
    profiles:
      - batch

volumes:
  data:
    driver: local
  prometheus-data:
  grafana-data:

networks:
  default:
    name: inference-network