version: "3.9"

services:
  ollama:
    image: ollama/ollama:0.4.2
    container_name: aimnote-ollama
    restart: unless-stopped
    ports:
      - "${OLLAMA_PORT:-11434}:11434"
    volumes:
      - ollama_data:/root/.ollama
    environment:
      # 为了让模型常驻内存，否则频繁调度时会被逐出
      - OLLAMA_KEEP_ALIVE=${OLLAMA_KEEP_ALIVE:-24h}
    healthcheck:
      test: ["CMD", "curl", "-f", "http://127.0.0.1:11434/api/version"]
      interval: 30s
      timeout: 10s
      retries: 5
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: ["gpu"]
              count: "${OLLAMA_GPU_COUNT:-0}"
        limits:
          cpus: "${OLLAMA_CPU_LIMIT:-4}"
          memory: "${OLLAMA_MEMORY_LIMIT:-16g}"

  ollama-model-init:
    image: curlimages/curl:8.11.1
    container_name: aimnote-ollama-init
    depends_on:
      ollama:
        condition: service_started
    restart: "no"
    entrypoint: ["/bin/sh", "-c"]
    # 通过 OpenAI 兼容 API 主动预拉取 qwen3-embedding 与 qwen2.5
    command: |
      until curl -fsS http://ollama:11434/api/version >/dev/null 2>&1; do
        echo "等待 Ollama 服务启动..."
        sleep 2
      done
      for model in "qwen3-embedding:8b" "qwen2.5:14b"; do
        echo "准备拉取 $model"
        curl -fsS -X POST http://ollama:11434/api/pull \
          -H "Content-Type: application/json" \
          -d "{\"name\":\"$model\"}" || exit 1
      done
      echo "模型拉取完成"

volumes:
  ollama_data:
    driver: local