docker-compose.yml

# DeepSupport OS + RAGLab (sibling ../RAGLab) — all-in-Docker stack.
# UI :5173 (nginx → api) · API host :18000 · RAGLab host :18001 · Qdrant :6333
# Internal: api → http://raglab:8000  (same compose network)

services:
  raglab-qdrant:
    image: qdrant/qdrant:v1.13.2
    ports:
      - "6333:6333"
      - "6334:6334"
    volumes:
      # Reuse the sibling RAGLab project's qdrant volume (raglab_qdrant_data) so the
      # already-ingested vectors (kb=deepsupport / huawei) are available without re-ingest.
      # Fresh machines: run `cd ../RAGLab && docker compose up -d qdrant` once (creates the
      # volume), or re-ingest via scripts/ingest_to_raglab.py.
      - raglab_qdrant_data:/qdrant/storage
    restart: unless-stopped

  raglab:
    build:
      # Slim context — avoid GB-scale data/models/*.whl from RAGLab root
      context: ../RAGLab/backend
      dockerfile: Dockerfile.context
    ports:
      # Host 8000/8001 often in Windows Hyper-V excluded ranges
      - "18001:8000"
    env_file:
      - ../RAGLab/.env
    environment:
      QDRANT_URL: http://raglab-qdrant:6333
      QDRANT_PATH: /data/qdrant
      DATABASE_URL: sqlite:////data/raglab.db
      UPLOAD_DIR: /data/uploads
      HUAWEI_KB_PATH: /data/huawei_kb
      HUAWEI_QA_PATH: /data/huawei_kb/qa.jsonl
      CORPUS_QA_PATH: /data/huawei_kb/qa.jsonl
      EMBEDDING_MODEL_PATH: /models/bge-small-zh-v1.5
      RERANKER_MODEL_PATH: /models/bge-reranker-v2-m3
      EMBEDDING_DEVICE: ${EMBEDDING_DEVICE:-cpu}
      RERANKER_DEVICE: ${RERANKER_DEVICE:-cpu}
      CORS_ORIGINS: http://localhost:5173,http://127.0.0.1:5173,http://localhost:18080,http://127.0.0.1:18080
      HF_HOME: /cache/huggingface
    volumes:
      - ../RAGLab/data:/data
      # NOTE: model dirs must be REAL directories — NTFS junctions (e.g. pointing into
      # %USERPROFILE%\.cache\huggingface) resolve to host-only paths inside the container
      # and SentenceTransformer fails with "Path ... not found".
      - ../RAGLab/models:/models:ro
      - raglab_hf_cache:/cache/huggingface
    depends_on:
      - raglab-qdrant
    # Alias "backend" so RAGLab frontend nginx.conf (proxy_pass http://backend:8000) works
    networks:
      default:
        aliases:
          - backend
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://127.0.0.1:8000/api/health"]
      interval: 15s
      timeout: 10s
      retries: 24
      start_period: 240s

  raglab-frontend:
    build:
      context: ../RAGLab/frontend
      dockerfile: Dockerfile.context
    ports:
      - "18080:80"
    depends_on:
      raglab:
        condition: service_healthy
    restart: unless-stopped

  api:
    build:
      context: ./backend
    ports:
      - "18000:8000"
    env_file:
      - .env
    volumes:
      - ./data:/app/data
      - ./workspace:/app/workspace
      - ./skills:/app/skills
      - ./config:/app/config
      - ./memory:/app/memory
    environment:
      - DATABASE_URL=sqlite:///data/deepsupport.db
      - WORKSPACE_DIR=workspace
      - MCP_SERVERS_CONFIG=config/mcp_servers.json
      - API_HOST=0.0.0.0
      # Same-compose DNS — do not use host.docker.internal for RAGLab
      - RAGLAB_BASE_URL=http://raglab:8000
      - OLLAMA_BASE_URL=http://host.docker.internal:11434/v1
    extra_hosts:
      - "host.docker.internal:host-gateway"
    depends_on:
      raglab:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health')"]
      interval: 10s
      timeout: 5s
      retries: 6
      start_period: 20s

  frontend:
    build:
      context: ./frontend
    ports:
      - "5173:80"
    depends_on:
      api:
        condition: service_healthy
    restart: unless-stopped

volumes:
  raglab_qdrant_data:
    external: true
    name: raglab_qdrant_data
  raglab_hf_cache: