services:
  # PostgreSQL Database Service
  database:
    image: postgres:16-alpine
    container_name: dashboard-database
    restart: unless-stopped
    environment:
      - POSTGRES_USER=${DB_USER:-dashboard_user}
      - POSTGRES_PASSWORD=${DB_PASSWORD:-change_this_password}
      - POSTGRES_DB=${DB_NAME:-dashboard_db}
      - PGDATA=/var/lib/postgresql/data/pgdata
    volumes:
      # Persist database data
      - db-data:/var/lib/postgresql/data
    networks:
      - dashboard-network
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${DB_USER:-dashboard_user} -d ${DB_NAME:-dashboard_db}"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    # Server: SSH tunnel / DBeaver via localhost:15432 (avoids host Postgres on 5432).
    # Mac dev: add docker-compose.override.yml from docker-compose.override.local.example for :5432.
    ports:
      - "127.0.0.1:15432:5432"

  # Python RAG Microservice
  rag:
    build:
      context: ./AI Insight
      dockerfile: Dockerfile
    container_name: dashboard-rag
    restart: unless-stopped
    ports:
      - "5005:5002"
    environment:
      - RAG_PORT=5002
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
      - dashboard-network

  # Backend API Service
  backend:
    build:
      context: ./backend
      dockerfile: Dockerfile
    container_name: dashboard-backend
    restart: unless-stopped
    ports:
      - "127.0.0.1:15001:5001"
    environment:
      - NODE_ENV=production
      - PORT=5001
      - DB_HOST=database
      - DB_PORT=5432
      - DB_USER=${DB_USER:-dashboard_user}
      - DB_PASSWORD=${DB_PASSWORD:-change_this_password}
      - DB_NAME=${DB_NAME:-dashboard_db}
      - DB_SSL=false
      - RAG_SERVICE_URL=http://rag:5002
      - LM_STUDIO_BASE_URL=http://llm:11434
      - LLM_MODEL_NAME=${LLM_MODEL_NAME:-deepseek-r1:8b}
    extra_hosts:
      - "host.docker.internal:host-gateway"
    depends_on:
      database:
        condition: service_healthy
      rag:
        condition: service_started
      llm:
        condition: service_started
    networks:
      - dashboard-network
    healthcheck:
      test: ["CMD", "node", "-e", "require('http').get('http://localhost:5001/health', (r) => {process.exit(r.statusCode === 200 ? 0 : 1)})"]
      interval: 30s
      timeout: 3s
      retries: 3
      start_period: 40s

  # Frontend Service
  frontend:
    build:
      context: .
      dockerfile: Dockerfile
      args:
        - CHICKEN_COUNTING_API_KEY=${VITE_CHICKEN_COUNTING_API_KEY:-}
    container_name: dashboard-frontend
    restart: unless-stopped
    ports:
      - "5002:80"
    depends_on:
      backend:
        condition: service_healthy
    networks:
      - dashboard-network
    healthcheck:
      test: ["CMD", "wget", "--quiet", "--tries=1", "--spider", "http://localhost:80/health"]
      interval: 30s
      timeout: 3s
      retries: 3
      start_period: 10s

  # Local LLM Server (Ollama)
  llm:
    image: ollama/ollama:latest
    container_name: dashboard-llm
    restart: unless-stopped
    ports:
      - "11434:11434"
    volumes:
      - ollama-data:/root/.ollama
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
      - dashboard-network
    # Untuk Windows dengan GPU Nvidia (Docker): Install Nvidia Container Toolkit, lalu uncomment baris di bawah:
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: all
    #           capabilities: [gpu]

  # Sidecar helper to auto-download the LLM model
  llm-model-downloader:
    image: curlimages/curl:latest
    container_name: dashboard-llm-downloader
    restart: "no"
    depends_on:
      - llm
    environment:
      - LLM_MODEL_NAME=${LLM_MODEL_NAME:-deepseek-r1:8b}
    entrypoint: ["/bin/sh", "-c"]
    command:
      - |
        echo "Waiting for Ollama service to start..."
        until curl -s http://llm:11434/api/tags > /dev/null 2>&1; do
          sleep 3
        done
        echo "Ollama is ready. Checking if model $$LLM_MODEL_NAME is pulled..."
        if curl -s http://llm:11434/api/tags | grep -q "$$LLM_MODEL_NAME"; then
          echo "Model $$LLM_MODEL_NAME is already available. Skipping download."
        else
          echo "Model $$LLM_MODEL_NAME not found. Initiating auto-pull (this may take a few minutes)..."
          curl -X POST http://llm:11434/api/pull -d "{\"name\": \"$$LLM_MODEL_NAME\"}"
          echo "Model $$LLM_MODEL_NAME successfully pulled!"
        fi
    networks:
      - dashboard-network

# Networks
networks:
  dashboard-network:
    driver: bridge

# Volumes
volumes:
  db-data:
    driver: local
    # Using named volume - Docker will manage it automatically
    # Data will persist in Docker's volume directory
  ollama-data:
    driver: local