deployment/docker-compose-base-gpu.yml

services:
  es01:
    container_name: yusys-rag-es-01
    image: docker.elastic.co/elasticsearch/elasticsearch:${STACK_VERSION}
    volumes:
      - esdata01:/usr/share/elasticsearch/data
    ports:
      - ${ES_PORT}:9200
    environment:
      - node.name=es01
      - ELASTIC_PASSWORD=${ELASTIC_PASSWORD}
      - bootstrap.memory_lock=false
      - discovery.type=single-node
      - xpack.security.enabled=true
      - xpack.security.http.ssl.enabled=false
      - xpack.security.transport.ssl.enabled=false
      - TZ=${TIMEZONE}
    mem_limit: ${MEM_LIMIT}
    ulimits:
      memlock:
        soft: -1
        hard: -1
    healthcheck:
      test: ["CMD-SHELL", "curl http://localhost:9200"]
      interval: 10s
      timeout: 10s
      retries: 120
    networks:
      - yusys-rag
    restart: always

  mysql:
    image: docker.byai.uk/library/mysql:5.7.44
    container_name: yusys-rag-mysql
    environment:
      - MYSQL_ROOT_PASSWORD=${MYSQL_PASSWORD}
      - TZ=${TIMEZONE}
    command:
      --max_connections=1000
      --character-set-server=utf8mb4
      --collation-server=utf8mb4_general_ci
      --tls_version="TLSv1.2,TLSv1.3"
      --init-file /data/application/init.sql
    ports:
      - ${MYSQL_PORT}:3306
    volumes:
      - mysql_data:/var/lib/mysql
      - ./init.sql:/data/application/init.sql
    networks:
      - yusys-rag
    healthcheck:
      test: ["CMD", "mysqladmin" ,"ping", "-uroot", "-p${MYSQL_PASSWORD}"]
      interval: 10s
      timeout: 10s
      retries: 3
    restart: always

  minio:
    image: quay.io/minio/minio:RELEASE.2024-09-22T00-33-43Z
    container_name: yusys-rag-minio
    command: server --console-address ":9001" /data
    ports:
      - ${MINIO_PORT}:9000
      - ${MINIO_CONSOLE_PORT}:9001
    environment:
      - MINIO_ROOT_USER=${MINIO_USER}
      - MINIO_ROOT_PASSWORD=${MINIO_PASSWORD}
      - TZ=${TIMEZONE}
    volumes:
      - minio_data:/data
    networks:
      - yusys-rag
    restart: always

  redis:
    image: docker.byai.uk/library/redis:7.2.4
    container_name: yusys-rag-redis
    command: redis-server --requirepass ${REDIS_PASSWORD} --maxmemory 128mb --maxmemory-policy allkeys-lru
    ports:
      - ${REDIS_PORT}:6379
    volumes:
      - redis_data:/data
    networks:
      - yusys-rag
    restart: always

  gotenberg:
    image: docker.byai.uk/gotenberg/gotenberg:8.9.1
    command:
      - gotenberg
      - '--api-port=3000'
      - '--api-timeout=60s'
      - '--api-disable-health-check-logging'
      - '--libreoffice-restart-after=10'
      - '--libreoffice-auto-start'
      - '--libreoffice-start-timeout=10s'
      - '--libreoffice-max-queue-size=50'
      - '--gotenberg-graceful-shutdown-duration=30s'
    expose:
      - "3000"
    deploy:
      replicas: 2
      endpoint_mode: dnsrr
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:3000/health"]
      interval: 10s
      timeout: 1s
      retries: 3
    user: "1001"
    networks:
      - yusys-rag
    restart: always

  unstructured-api:
    image: downloads.unstructured.io/unstructured-io/unstructured-api:0.0.80
    expose:
      - "8000"
    volumes:
      - ./huggingface:/home/notebook-user/.cache/huggingface
    deploy:
      replicas: 4
      endpoint_mode: dnsrr
    networks:
      - yusys-rag
    restart: always

  llm-infinity:
    image: docker.byai.uk/michaelf34/infinity:0.0.55
    container_name: yusys-rag-infinity
    command: ["v2"]
    environment:
      - HF_ENDPOINT=https://hf-mirror.com
      - INFINITY_PORT=8781
      - INFINITY_MODEL_ID=BAAI/bge-reranker-v2-m3;BAAI/bge-large-zh;
    ports:
      - "8781:8781"
    volumes:
      - ./huggingface:/app/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: ['4']
            # count: all
            capabilities: [gpu]
    networks:
      - yusys-rag
    restart: always

  doc-parser:
    image: 192.168.107.2:5000/yusyscloud/llm-doc-parser:dev-32
    container_name: yusys-rag-doc
    environment:
      - UNSTRUCTURED_API_URL=http://unstructured-api:8000/general/v0/general
      - SPLIT_PDF_CONCURRENCY_LEVEL=15
    ports:
      - "8023:8023"
    volumes:
      - ./huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: ['4']
            #count: all
            capabilities: [gpu]
    networks:
      - yusys-rag
    restart: always

volumes:
  esdata01:
    driver: local
  mysql_data:
    driver: local
  minio_data:
    driver: local
  redis_data:
    driver: local

networks:
  yusys-rag:
    driver: bridge