version: '3.8'

services:
  auramind-api:
    build: .
    ports:
      - "8000:8000"
    environment:
      - MODEL_VARIANT=270m
      - MAX_CONCURRENT_REQUESTS=10
      - DEVICE=auto
    volumes:
      - ./models:/app/models
      - ./logs:/app/logs
    restart: unless-stopped
    
  auramind-lightweight:
    build: .
    ports:
      - "8001:8000"
    environment:
      - MODEL_VARIANT=90m
      - MAX_CONCURRENT_REQUESTS=20
      - DEVICE=cpu
    volumes:
      - ./models:/app/models
      - ./logs:/app/logs
    restart: unless-stopped