docker-compose_ai.yml

services:
  redis.ai:
    image: redis:latest
    container_name: cometa.redis.ai
    volumes:
      - ./redis/redis.conf:/usr/local/etc/redis/redis.conf
      - ./redis/start.sh:/usr/local/bin/start.sh
      - ./data/redis/data:/data
      - ./data/redis/certs:/share/certs
    ports:
      - "6379:6379"
    networks:
      - testing
    healthcheck:
      test: [ "CMD","redis-cli", "--tls", "--cacert", "/share/certs/ca-cert.pem", "--raw", "incr", "ping"]
    restart: always
    command: sh /usr/local/bin/start.sh

  # Use this command to have mutiple containers of ollama.ai
  # docker-compose -f docker-compose_ai.yml up --scale ollama.ai=3
  ollama.ai:
    build: ./cometa_ai
    # container_name: cometa.ollama.ai
    expose:
      - "8083"
    volumes:
      - ./cometa_ai:/app
      - ./data/ollama:/home/cometa/.ollama/models
      - ./data/redis/certs:/share/certs
    working_dir: /app 
    # entrypoint: ./start.sh 
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all  # Or specify a number like 1 if you want to limit GPU usage to a single GPU
              capabilities: [gpu]
    
    runtime: nvidia  # Use NVIDIA's runtime
    depends_on:
      - redis.ai
    environment:
      OLLAMA_HOST: "0.0.0.0:8083"
      # IMAGE_ANALYZER_MODEL_NAME: "llava:13b" #Not working
      NVIDIA_VISIBLE_DEVICES: all
      NVIDIA_DRIVER_CAPABILITIES: compute,utility
      NAME: ollama
      REDIS_HOST: "cometa.redis.ai"
      REDIS_PORT: 6379
      REDIS_DB: 0
      REDIS_DB_TSL_SSL_ENABLED: Yes
      REDIS_NUMBER_OF_WORKERS: 1
    networks:
      - testing
    restart: always

networks:
  testing:
    driver: "bridge"

volumes:
  redis_data: