Kokoro-FastAPI/docker/gpu/docker-compose.yml

name: InstaVoice
services:
 server1:
    build:
      context: ../..
      dockerfile: docker/gpu/Dockerfile
    volumes:
      - ../../api:/app/api
    ports:
      - "8880:8880"
    environment:
      - PYTHONPATH=/app:/app/api
      - USE_GPU=true
      - USE_ONNX=false
      - PYTHONUNBUFFERED=1
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

 server2:
    build:
      context: ../..
      dockerfile: docker/gpu/Dockerfile
    volumes:
      - ../../api:/app/api
    ports:
      - "8880:8880"
    environment:
      - PYTHONPATH=/app:/app/api
      - USE_GPU=true
      - USE_ONNX=false
      - PYTHONUNBUFFERED=1
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

 server3:
    build:
      context: ../..
      dockerfile: docker/gpu/Dockerfile
    volumes:
      - ../../api:/app/api
    ports:
      - "8880:8880"
    environment:
      - PYTHONPATH=/app:/app/api
      - USE_GPU=true
      - USE_ONNX=false
      - PYTHONUNBUFFERED=1
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

 
nginx:
    image: nginx:alpine
    ports:
      - "80:80" # Expose port 80 on the host machine
    volumes:
      - ./nginx.conf:/etc/nginx/nginx.conf # Load custom NGINX configuration
    depends_on:
      - server3
      - server1
      - server2