services:
  ir-studio:
    build:
      context: .
      dockerfile: Dockerfile
    container_name: ir-studio
    restart: unless-stopped
    ports:
      - "8000:8000"
    environment:
      - HOST=0.0.0.0
      - PORT=8000
      - CORS_ORIGINS=*
      - ONNX_NUM_THREADS=4
      - MAX_CACHE_ENTRIES=2000
      # Optional external model & index directories:
      # - BGE_MODEL_PATH=/app/models/bge/model_quantized.onnx
      # - BGE_TOKENIZER_PATH=/app/models/bge/tokenizer.json
      # - SPLADE_MODEL_PATH=/app/models/splade/model.onnx
      # - SPLADE_TOKENIZER_PATH=/app/models/splade/tokenizer.json
      # - FAISS_INDEX_DIR=/app/indexes/nfcorpus.bge-base-en-v1.5
    volumes:
      # Mount models or precomputed indexes from host if available:
      - ./indexes:/app/indexes:ro
      - ./models:/app/models:ro
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/api/health"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s