docker-compose.yml

# Docker Compose configuration for Abstracts Explorer
# Compatible with Docker Compose v2+ and Podman Compose

services:
  # Nginx reverse proxy with SSL termination
  # Listens on ports 80 (redirect to HTTPS) and 443 (HTTPS).
  #
  # REQUIRED before starting: create a ./certs/ directory containing:
  #   cert.pem  — your SSL certificate (or full chain)
  #   key.pem   — your SSL private key
  # The nginx container will fail to start if either file is missing.
  nginx:
    image: nginx:stable-alpine
    container_name: abstracts-nginx
    ports:
      - "80:80"
      - "443:443"
    volumes:
      - ./nginx/nginx.conf:/etc/nginx/nginx.conf:ro
      - ./certs:/etc/nginx/certs:ro
    depends_on:
      abstracts-explorer:
        condition: service_healthy
    restart: unless-stopped

  # Main Abstracts Explorer application
  # Port 5000 is intentionally not exposed to the host; all external traffic
  # is routed through the nginx reverse proxy above.
  abstracts-explorer:
    image: ghcr.io/thawn/abstracts-explorer:latest
    container_name: abstracts-explorer
    volumes:
      # Persist data directory
      - abstracts-data:/app/data
    environment:
      # Data dir must be the same as assigned to the volume, otherwise downloaded data will not persist
      - DATA_DIR=/app/data

      # Database configuration - using PostgreSQL
      - PAPER_DB=postgresql://abstracts:abstracts_password@postgres:5432/abstracts
      
      # Embeddings configuration - using ChromaDB service
      - EMBEDDING_DB=http://chromadb:8000
      - COLLECTION_NAME=papers
      
      # LLM Backend configuration
      # The auth token for the LLM backend is read from the environment variable LLM_BACKEND_AUTH_TOKEN,
      # which must be set in the environment or .env file.
      # Option 1: Use external service (e.g., blablador)
      - LLM_BACKEND_URL=https://api.helmholtz-blablador.fz-juelich.de
      - LLM_BACKEND_AUTH_TOKEN=${LLM_BACKEND_AUTH_TOKEN}

      # Option 2: Use host's LM Studio (requires host.docker.internal or host network)
      # - LLM_BACKEND_URL=http://host.docker.internal:1234
      
      # Models
      # For Blablador
      - CHAT_MODEL=alias-code
      - EMBEDDING_MODEL=alias-qwen3-8b-embeddings

      # For host's LM Studio we recommend
      # - CHAT_MODEL=gemma-3-4b-it-qat
      # - EMBEDDING_MODEL=text-embedding-qwen3-embedding-4b

      # RAG settings
      - MAX_CONTEXT_PAPERS=5
      - ENABLE_QUERY_REWRITING=true
      - QUERY_SIMILARITY_THRESHOLD=0.7
      
      # Logging configuration
      # Set log level: DEBUG, INFO, WARNING (default), ERROR, CRITICAL
      - LOG_LEVEL=WARNING

      # Registry settings (optional)
      # Used by: abstracts-explorer registry upload/download
      # GITHUB_TOKEN: PAT with read:packages (download) or write:packages (upload) scope
      # REGISTRY_REPOSITORY: default OCI repository, e.g. ghcr.io/thawn/abstracts-data
      - GITHUB_TOKEN=${GITHUB_TOKEN:-}
      - REGISTRY_REPOSITORY=${REGISTRY_REPOSITORY:-}

      # Reverse-proxy hop counts for werkzeug ProxyFix.
      # 1 = trust one proxy layer (default, matches the nginx service above).
      # Set to 0 to disable, or increase for additional proxy layers.
      - PROXY_X_FOR=1
      - PROXY_X_PROTO=1
      - PROXY_X_HOST=1
      - PROXY_X_PREFIX=1
    
    # Use host network mode for easier access to host's LM Studio
    # Uncomment if using LM Studio on host
    # network_mode: host
    
    # Depends on database services
    depends_on:
      chromadb:
        condition: service_healthy
      postgres:
        condition: service_healthy
    
    restart: unless-stopped
    
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:5000/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 10s

  # ChromaDB vector database service
  chromadb:
    image: chromadb/chroma:latest
    container_name: abstracts-chromadb
    volumes:
      - chromadb-data:/data
    environment:
      - IS_PERSISTENT=TRUE
      - ANONYMIZED_TELEMETRY=FALSE
    restart: unless-stopped
    
    healthcheck:
      test: ["CMD-SHELL", "timeout 1 bash -c '</dev/tcp/localhost/8000' || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s

  # PostgreSQL database service
  postgres:
    image: postgres:16-alpine
    container_name: abstracts-postgres
    volumes:
      - postgres-data:/var/lib/postgresql/data
    environment:
      - POSTGRES_DB=abstracts
      - POSTGRES_USER=abstracts
      - POSTGRES_PASSWORD=abstracts_password
      # Change password in production!
    restart: unless-stopped
    
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U abstracts"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s

volumes:
  # Named volumes for persistent data
  abstracts-data:
    driver: local
  chromadb-data:
    driver: local
  postgres-data:
    driver: local

networks:
  default:
    name: abstracts-network