services:
  haf_squid:
    image: ubuntu/squid:latest
    container_name: haf_squid
    volumes:
      - ./docker/squid/squid.conf:/etc/squid/squid.conf:ro
      - ./docker/squid/whitelist.txt:/etc/squid/whitelist.txt:ro
      - ./docker/squid/keywordblock.txt:/etc/squid/keywordblock.txt:ro
      - ./docker/squid/errors/:/etc/squid/errors/:ro
    restart: unless-stopped
    networks:
      - haf_squid_net
      - haf_public_net


  # # # # # #
  # Hermes  #
  # # # # # #
  haf_hermes:
    image: nousresearch/hermes-agent:latest
    container_name: haf_hermes
    restart: unless-stopped
    command: gateway run
    environment:
      API_SERVER_ENABLED:      true
      API_SERVER_HOST:         0.0.0.0
      API_SERVER_KEY:          ${HERMES_API_SERVER_KEY}
      API_SERVER_CORS_ORIGINS: '*'
      HERMES_DASHBOARD:        1
    env_file:
      - ".env"
    ports:
      - "8642:8642"
      - "9119:9119"
    networks:
      - haf_llama_net
      - haf_public_net
    depends_on:
      - haf_llama-vulkan
    volumes:
      - ./data/hermes/config/open/:/opt/data
      - ./data/hermes/data/open/:/opt/data/workspace
    deploy:
      resources:
        limits:
          memory: 4G
          cpus: "2.0"


  haf_hermes-restricted:
    image: nousresearch/hermes-agent:latest
    container_name: haf_hermes-restricted
    environment:
      HTTP_PROXY:  http://haf_squid:3128
      HTTPS_PROXY: http://haf_squid:3128
      NO_PROXY:    localhost,127.0.0.1,haf_llama-vulkan
      API_SERVER_ENABLED:      true
      API_SERVER_HOST:         0.0.0.0
      API_SERVER_KEY:          ${HERMES_API_SERVER_KEY}
      API_SERVER_CORS_ORIGINS: '*'
      HERMES_DASHBOARD:        1
    env_file:
      - ".env"
    restart: unless-stopped
    networks:
      - haf_squid_net
      - haf_llama_net
    depends_on:
      - haf_squid
      - haf_llama-vulkan
    volumes:
      - ./data/hermes/config/restricted/:/opt/data
      - ./data/hermes/data/restricted/:/opt/data/workspace
    deploy:
      resources:
        limits:
          memory: 4G
          cpus: "2.0"

  # # # # #
  # Llama #
  # # # # #
  haf_llama-vulkan:
    image: ghcr.io/ggml-org/llama.cpp:server-vulkan
    container_name: haf_llama-vulkan
    ports:
      - "8080:8080"
    volumes:
      - ./data/models:/models
    devices:
      - /dev/dri:/dev/dri
    group_add:
      - video
    command: >
      -m /models/${LLAMA_MODEL}
      --host 0.0.0.0
      --port 8080
      ${LLAMA_ARGS}
    restart: unless-stopped
    networks:
      - haf_llama_net


networks:
  haf_llama_net:
    internal: true
  haf_squid_net:
    internal: true
  haf_public_net: