# Foundry Swarm - Docker Compose for local development/testing
#
# Usage:
#   docker-compose up -d                    # Start coordinator + 2 workers
#   docker-compose up -d --scale worker=5   # Start with 5 workers
#   docker-compose logs -f coordinator      # Watch coordinator logs
#   docker-compose down                     # Stop all

version: "3.8"

services:
  coordinator:
    build:
      context: ../..
      dockerfile: swarm/docker/Dockerfile.coordinator
    container_name: swarm-coordinator
    ports:
      - "7100:7100"
    volumes:
      - ./config:/app/config:ro
      - checkpoints:/app/checkpoints
      - logs:/app/logs
    environment:
      - SW_PORT=7100
      - SW_MIN_WORKERS=2
      - SW_LEARNING_RATE=1e-4
      - SW_TOTAL_STEPS=100000
      - SW_CHECKPOINT_EVERY=1000
      - SW_HEARTBEAT_TIMEOUT=60
    networks:
      - swarm-net
    healthcheck:
      test: ["CMD", "nc", "-z", "localhost", "7100"]
      interval: 30s
      timeout: 10s
      retries: 3
    restart: unless-stopped

  worker:
    build:
      context: ../..
      dockerfile: swarm/docker/Dockerfile.worker
    depends_on:
      coordinator:
        condition: service_healthy
    volumes:
      - ./config:/app/config:ro
      - training-data:/app/data:ro
      - logs:/app/logs
    environment:
      - SW_COORDINATOR_HOST=coordinator
      - SW_COORDINATOR_PORT=7100
      - SW_DATA_PATH=/app/data
      - SW_BATCH_SIZE=0          # Auto-detect
      - SW_NUM_THREADS=0         # Auto-detect
      - SW_SHARD_INDEX=-1        # Auto-assign
    networks:
      - swarm-net
    deploy:
      replicas: 2
      resources:
        limits:
          cpus: "2"
          memory: 4G
        reservations:
          cpus: "0.5"
          memory: 512M
    restart: unless-stopped

networks:
  swarm-net:
    driver: bridge

volumes:
  checkpoints:
    driver: local
  logs:
    driver: local
  training-data:
    driver: local