docker-compose.cuda.yml

services:
  ollama:
    image: hwdsl2/ollama-server:cuda
    container_name: ollama
    restart: always
    # ports:
    #   - "11434:11434/tcp"  # Uncomment for direct access to Ollama
    volumes:
      - ollama-data:/var/lib/ollama
      - ollama-shared:/var/lib/ollama-shared
      # - ./ollama.env:/ollama.env:ro  # optional: custom config
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://127.0.0.1:11434/"]
      interval: 5s
      timeout: 3s
      start_period: 60s
      retries: 3
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  db:
    image: postgres:18
    container_name: litellm-db
    restart: always
    environment:
      POSTGRES_USER: litellm
      POSTGRES_PASSWORD: litellm
      POSTGRES_DB: litellm
    volumes:
      - litellm-db:/var/lib/postgresql
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U litellm"]
      interval: 5s
      timeout: 5s
      retries: 5

  litellm:
    image: hwdsl2/litellm-server
    container_name: litellm
    restart: always
    ports:
      - "4000:4000/tcp"
    environment:
      - LITELLM_OLLAMA_BASE_URL=http://ollama:11434
      - LITELLM_MCP_URL=http://mcp:3000/mcp
      - LITELLM_DATABASE_URL=postgresql://litellm:litellm@db:5432/litellm
    volumes:
      - litellm-data:/etc/litellm
      - ollama-shared:/var/lib/ollama-shared:ro
      - mcp-shared:/var/lib/mcp-shared:ro
      - litellm-shared:/var/lib/litellm-shared
      # - ./litellm.env:/litellm.env:ro  # optional: custom config
    depends_on:
      ollama:
        condition: service_healthy
      mcp:
        condition: service_healthy
      db:
        condition: service_healthy

  embeddings:
    image: hwdsl2/embeddings-server
    container_name: embeddings
    restart: always
    ports:
      - "127.0.0.1:8000:8000/tcp"
    volumes:
      - embeddings-data:/var/lib/embeddings
      # - ./embed.env:/embed.env:ro  # optional: custom config

  whisper:
    image: hwdsl2/whisper-server:cuda
    container_name: whisper
    restart: always
    ports:
      - "127.0.0.1:9000:9000/tcp"
    volumes:
      - whisper-data:/var/lib/whisper
      # - ./whisper.env:/whisper.env:ro  # optional: custom config
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  mcp:
    image: hwdsl2/mcp-gateway
    container_name: mcp
    restart: always
    # ports:
    #   - "3000:3000/tcp"  # Uncomment for direct MCP client access (e.g. Claude Desktop, Cursor)
    volumes:
      - mcp-data:/var/lib/mcp
      - mcp-shared:/var/lib/mcp-shared
      # - ./mcp.env:/mcp.env:ro  # optional: custom config
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://127.0.0.1:3000/health"]
      interval: 5s
      timeout: 3s
      start_period: 60s
      retries: 3

  anythingllm:
    image: mintplexlabs/anythingllm
    container_name: anythingllm
    restart: always
    ports:
      - "3001:3001/tcp"  # For a host-based reverse proxy, change to "127.0.0.1:3001:3001/tcp"
    environment:
      - STORAGE_DIR=/app/server/storage
      - LLM_PROVIDER=generic-openai
      - GENERIC_OPEN_AI_BASE_PATH=http://litellm:4000/v1
      - GENERIC_OPEN_AI_MODEL_PREF=ollama/llama3.2:3b
      - GENERIC_OPEN_AI_MODEL_TOKEN_LIMIT=131072
      - EMBEDDING_ENGINE=native
      - DISABLE_TELEMETRY=true
    volumes:
      - anythingllm-data:/app/server/storage
      - litellm-shared:/var/lib/litellm-shared:ro
      - ./chat-ui-bootstrap.sh:/usr/local/bin/chat-ui-bootstrap.sh:ro
    entrypoint: ["/bin/bash", "/usr/local/bin/chat-ui-bootstrap.sh"]
    # cap_add:
    #   - SYS_ADMIN  # Uncomment if you need AnythingLLM's built-in web scraping (uses Chromium)
    depends_on:
      - litellm

  # docling:  # Uncomment to enable document parsing (PDF, DOCX, etc.); also uncomment docling-data volume below
  #   image: hwdsl2/docling-server:cuda
  #   container_name: docling
  #   restart: always
  #   ports:
  #     - "127.0.0.1:5001:5001/tcp"
  #   volumes:
  #     - docling-data:/var/lib/docling
  #     # - ./docling.env:/docling.env:ro  # optional: custom config
  #   deploy:
  #     resources:
  #       reservations:
  #         devices:
  #           - driver: nvidia
  #             count: 1
  #             capabilities: [gpu]

  # kokoro:  # Uncomment to enable text-to-speech (TTS); also uncomment kokoro-data volume below
  #   image: hwdsl2/kokoro-server:cuda
  #   container_name: kokoro
  #   restart: always
  #   ports:
  #     - "127.0.0.1:8880:8880/tcp"
  #   volumes:
  #     - kokoro-data:/var/lib/kokoro
  #     # - ./kokoro.env:/kokoro.env:ro  # optional: custom config
  #   deploy:
  #     resources:
  #       reservations:
  #         devices:
  #           - driver: nvidia
  #             count: 1
  #             capabilities: [gpu]

  # whisper-live:  # Uncomment to enable real-time speech-to-text (WebSocket); also uncomment whisper-live-data volume below
  #   image: hwdsl2/whisper-live-server:cuda
  #   container_name: whisper-live
  #   restart: always
  #   ports:
  #     - "127.0.0.1:9090:9090/tcp"  # WebSocket
  #     - "127.0.0.1:8001:8000/tcp"  # REST API (host port 8001 avoids conflict with embeddings)
  #   volumes:
  #     - whisper-live-data:/var/lib/whisper-live
  #     # - ./whisper-live.env:/whisper-live.env:ro  # optional: custom config
  #   deploy:
  #     resources:
  #       reservations:
  #         devices:
  #           - driver: nvidia
  #             count: 1
  #             capabilities: [gpu]

volumes:
  ollama-data:
    name: ollama-data
  ollama-shared:
    name: ollama-shared
  litellm-data:
    name: litellm-data
  embeddings-data:
    name: embeddings-data
  whisper-data:
    name: whisper-data
  mcp-data:
    name: mcp-data
  mcp-shared:
    name: mcp-shared
  litellm-shared:
    name: litellm-shared
  anythingllm-data:
    name: anythingllm-data
  litellm-db:
    name: litellm-db
  # docling-data:
  #   name: docling-data
  # kokoro-data:
  #   name: kokoro-data
  # whisper-live-data:
  #   name: whisper-live-data