docker-compose.yml

# QuantMind OSS Edition - Docker Compose
# 单镜像部署，所有服务运行在一个容器内

name: quantmind

services:
  # 1. 数据库 (PostgreSQL)
  db:
    # 使用官方标签；部署脚本配置的 Docker registry mirror 负责国内加速。
    # 离线镜像包也以此标签导入，避免 Compose 再次联网拉取。
    image: postgres:15-alpine
    container_name: quantmind-db
    shm_size: 256mb
    environment:
      - POSTGRES_DB=${DB_NAME:-quantmind}
      - POSTGRES_USER=${DB_USER:-quantmind}
      - POSTGRES_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - TZ=Asia/Shanghai
    volumes:
      - postgres-data:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U $${POSTGRES_USER} -d $${POSTGRES_DB}"]
      interval: 10s
      timeout: 5s
      retries: 5
    restart: unless-stopped
    ports:
      - "5432:5432"
    networks:
      - quantmind-net

  # 2. 缓存 (Redis) - 统一实例
  redis:
    image: redis:7-alpine
    container_name: quantmind-redis
    ports:
      - "6379:6379"
    command: redis-server --appendonly yes --maxmemory 512mb --maxmemory-policy allkeys-lru --databases 16 --bind 0.0.0.0
    volumes:
      - redis-data:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 5
    restart: unless-stopped
    networks:
      - quantmind-net

  # 3. QuantMind 核心服务 (单镜像运行所有后端服务)
  quantmind:
    image: quantmind-oss:latest
    build:
      context: .
      dockerfile: docker/Dockerfile.oss
      args:
        TORCH_DEVICE: ${TORCH_DEVICE:-skip}
        TORCH_CPU_INDEX_URL: ${TORCH_CPU_INDEX_URL:-https://mirrors.aliyun.com/pytorch-wheels/cpu}
    platform: linux/amd64
    # GPU 直通：宿主机已安装 nvidia-container-toolkit 时可开启
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: all
    #           capabilities: [gpu]
    container_name: quantmind
    depends_on:
      db:
        condition: service_healthy
      redis:
        condition: service_healthy
    ports:
      - "8000:8000"  # API Gateway
      - "8001:8001"  # Engine (Qlib回测)
      - "8002:8002"  # Trade (交易服务)
      - "8003:8003"  # Stream (实时行情)
    environment:
      # 应用配置
      - APP_EDITION=oss
      - APP_ENV=production
      - SERVICE_MODE=all
      - TZ=Asia/Shanghai
      - PYTHONPATH=/app
      - HOST_PROJECT_PATH=${HOST_PROJECT_PATH:-/opt/quantmind}

      # 安全配置
      - SECRET_KEY=${SECRET_KEY:-changeme-generate-a-random-secret}
      - JWT_SECRET_KEY=${JWT_SECRET_KEY:-changeme-generate-a-random-secret}

      # 数据库配置
      - DB_DRIVER=asyncpg
      - DB_HOST=db
      - DB_PORT=5432
      - DB_NAME=${DB_NAME:-quantmind}
      - DB_USER=${DB_USER:-quantmind}
      - DB_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - DATABASE_URL=postgresql+asyncpg://${DB_USER:-quantmind}:${DB_PASSWORD:-quantmind2026}@db:5432/${DB_NAME:-quantmind}

      # psycopg2 直连配置 (news.py _pg_conn 等同步连接使用)
      - POSTGRES_HOST=db
      - POSTGRES_PORT=5432
      - POSTGRES_USER=${DB_USER:-quantmind}
      - POSTGRES_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - POSTGRES_DB=${DB_NAME:-quantmind}

      # 远程数据源配置（用于从官方数据库拉取增量数据）
      - SOURCE_DATABASE_URL=${SOURCE_DATABASE_URL:-postgresql://user:pass@localhost:5432/quantmind}

      # Redis 配置
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - REDIS_PASSWORD=

      # 存储配置 (本地存储)
      - STORAGE_MODE=local
      - STORAGE_ROOT=/data

      # 训练期间资源保护：训练开始时 stop 非保护容器以释放内存，结束后自动恢复。
      # 受保护前缀（按容器名前缀匹配，不会被停止）：quantmind* 和 qm-train-* 自身，
      # 加上 ragflow 套件依赖的基础设施（redis/mysql/es/minio）。
      - TRAINING_PAUSE_OTHERS=true
      - TRAINING_PROTECTED_NAME_PREFIXES=quantmind,qm-train-,docker-redis-,docker-mysql-,docker-es01-,docker-minio-

      # 融资融券股票池文件路径（容器内）
      - MARGIN_STOCK_POOL_PATH=/data/融资融券.json

      # 服务端口
      - API_PORT=8000
      - ENGINE_PORT=8001
      - TRADE_PORT=8002
      - STREAM_PORT=8003

      # 服务 Workers 配置
      # AI-IDE 执行日志状态当前保存在 engine worker 进程内存中，
      # OSS 部署需保持单 worker，避免 /start 与 /execute/logs/{job_id}
      # 落到不同 worker 时出现 "Job not found"。
      # all 模式会先由 main_oss 创建服务进程；这里再启用 Uvicorn
      # 多 worker 会形成嵌套 spawn，导致 API/交易进程反复退出。
      - API_WORKERS=${API_WORKERS:-1}
      - ENGINE_WORKERS=${ENGINE_WORKERS:-1}
      - TRADE_WORKERS=${TRADE_WORKERS:-1}
      - STREAM_WORKERS=${STREAM_WORKERS:-1}

      # 服务间通信 (OSS单镜像模式使用localhost)
      - TRADE_SERVICE_URL=http://127.0.0.1:8002
      - ENGINE_SERVICE_URL=http://127.0.0.1:8001
      - AI_IDE_SERVICE_URL=http://127.0.0.1:8001
      - STREAM_SERVICE_URL=http://127.0.0.1:8003
      - MARKET_DATA_SERVICE_URL=http://127.0.0.1:8003
      - STRATEGY_SERVICE_URL=http://127.0.0.1:8001
      - PORTFOLIO_SERVICE_URL=http://127.0.0.1:8002
      - USER_SERVICE_URL=http://127.0.0.1:8002
      - REAL_TRADING_SERVICE_URL=http://127.0.0.1:8002

      # 通达信(TDX)交易桥配置（Windows 桥地址/token，与 Windows 侧 setup.bat 一致）
      - TDX_BRIDGE_URL=${TDX_BRIDGE_URL:-http://192.168.31.31:8550}
      - TDX_BRIDGE_TOKEN=${TDX_BRIDGE_TOKEN:-}
      - ENABLE_TDX_PUSH=${ENABLE_TDX_PUSH:-false}
      - ENABLE_REAL_TRADING=${ENABLE_REAL_TRADING:-false}
      - REAL_BROKER_TYPE=${REAL_BROKER_TYPE:-tdx}

      # 调试配置
      - DEBUG=false
      - LOG_LEVEL=INFO

      # 策略模板目录
      - STRATEGY_TEMPLATES_DIR=/app/strategy_templates

      # 回测结果存储目录（使用共享卷）
      - QLIB_BACKTEST_RESULT_DIR=/data/backtest_results
      - QLIB_BACKTEST_KERNELS=1

      # 统一报告根目录（技能中心）：复盘/市场分析/个股研究/股票池/投研报告
      - QM_REPORTS_DIR=/data/reports
      - TRADING_AGENTS_RESULTS_DIR=/data/reports/trading_agents

      # 本地 Docker 训练编排（模型训练）
      - INTERNAL_CALL_SECRET=${INTERNAL_CALL_SECRET:-changeme-internal-secret}
      - TRAINING_IMAGE=${TRAINING_IMAGE:-quantmind-trainer:latest}
      - TRAINING_DOCKER_NETWORK=${TRAINING_DOCKER_NETWORK:-quantmind_quantmind-net}
      - HOST_PROJECT_PATH=${HOST_PROJECT_PATH:-/opt/quantmind}
      - QUANTMIND_API_BASE_URL=${QUANTMIND_API_BASE_URL:-http://quantmind:8000}

      # AutoDL 远程训练节点
      - TRAINING_AUTODL_HOST=${TRAINING_AUTODL_HOST:-}
      - TRAINING_AUTODL_SSH_PORT=${TRAINING_AUTODL_SSH_PORT:-22}
      - TRAINING_AUTODL_USER=${TRAINING_AUTODL_USER:-root}
      - TRAINING_AUTODL_SSH_PASSWORD=${TRAINING_AUTODL_SSH_PASSWORD:-}
      - TRAINING_AUTODL_SSH_KEY=${TRAINING_AUTODL_SSH_KEY:-}
      - TRAINING_AUTODL_WORK_DIR=${TRAINING_AUTODL_WORK_DIR:-/workspace}
      - TRAINING_AUTODL_DOCKER_IMAGE=${TRAINING_AUTODL_DOCKER_IMAGE:-quantmind-trainer:latest}
      - TRAINING_AUTODL_NODE_NAME=${TRAINING_AUTODL_NODE_NAME:-AutoDL GPU}
      - TRAINING_AUTODL_GPUS=${TRAINING_AUTODL_GPUS:-all}
      - TRAINING_MASTER_HOST=${TRAINING_MASTER_HOST:-}

      # Qwen/DashScope API (用于 AI 策略功能)
      # 使用 mock key 允许服务启动，用户需在个人中心配置真实 API key
      - DASHSCOPE_API_KEY=${DASHSCOPE_API_KEY:-mock-api-key-not-configured}
      - QWEN_API_KEY=${QWEN_API_KEY:-mock-api-key-not-configured}
      # DeepSeek API (用于 TradingAgents 等)
      - DEEPSEEK_API_KEY=${DEEPSEEK_API_KEY:-your-deepseek-api-key}
      - DEEPSEEK_BASE_URL=${DEEPSEEK_BASE_URL:-https://api.deepseek.com}
      - DEEPSEEK_MODEL=${DEEPSEEK_MODEL:-deepseek-chat}
      # 选股意图解析复用 AI_IDE LLM 配置（OpenAI 兼容）
      - AI_IDE_LLM_API_KEY=${AI_IDE_LLM_API_KEY:-mock-api-key-not-configured}
      - AI_IDE_LLM_BASE_URL=${AI_IDE_LLM_BASE_URL:-https://token-plan-cn.xiaomimimo.com/v1}

      # AI-IDE Chat LLM 配置（OpenAI 兼容；默认 MiMo）
      - AI_IDE_LLM_BASE_URL=${AI_IDE_LLM_BASE_URL:-https://token-plan-cn.xiaomimimo.com/v1}
      - AI_IDE_LLM_MODEL=${AI_IDE_LLM_MODEL:-mimo-v2.5-pro}
      - AI_IDE_LLM_API_KEY=${AI_IDE_LLM_API_KEY:-mock-api-key-not-configured}

      # RD-Agent / AlphaAgent 使用 OpenAI 兼容 API（复用 AI_IDE 配置）
      # 模型名称由 .env 的 AI_IDE_LLM_MODEL 决定；默认 mimo-v2.5-pro
      - CHAT_MODEL=${AI_IDE_LLM_MODEL:-mimo-v2.5-pro}
      # AlphaAgent reasoning_flag 默认 True → 用 reasoning_model；
      # 如果为空则发 model="" 导致 400。镜像 CHAT_MODEL 即可。
      - REASONING_MODEL=${AI_IDE_LLM_MODEL:-mimo-v2.5-pro}
      # 关闭 streaming — reasoning 模型的 reasoning_content chunk 导致 OpenAI SDK JSON 解析失败
      - CHAT_STREAM=false
      - OPENAI_API_KEY=${AI_IDE_LLM_API_KEY:-mock-api-key-not-configured}
      - OPENAI_API_BASE=${AI_IDE_LLM_BASE_URL:-https://token-plan-cn.xiaomimimo.com/v1}
      # AlphaAgent LLM 重试配置（502/429 等临时错误需更多重试）
      - MAX_RETRY=30
      - RETRY_WAIT_SECONDS=5
      # AlphaAgent reads OPENAI_BASE_URL (modern SDK convention); duplicate
      # OPENAI_API_BASE here so its factor-mining subprocess hits the right host.
      - OPENAI_BASE_URL=${AI_IDE_LLM_BASE_URL:-https://token-plan-cn.xiaomimimo.com/v1}

      # AlphaAgent 因子去重用 embedding；MiMo 不提供 embedding，走硅基流动
      - EMBEDDING_API_KEY=${EMBEDDING_API_KEY:-mock-api-key-not-configured}
      - EMBEDDING_BASE_URL=${EMBEDDING_BASE_URL:-https://api.siliconflow.cn/v1}
      - EMBEDDING_MODEL=${EMBEDDING_MODEL:-BAAI/bge-m3}

      # QuantDB SDK（付费高质量数据源：K线/Tick/财报/估值/315维AI因子）
      - QUANTDB_API_KEY=${QUANTDB_API_KEY:-}
      # QuantDB 本地数据中枢（A 股优先数据源）
      - QM_QUANTDB_DATA_DIR=/data/quantdb
      # 市场分析快照目录（脚本 compute.py 产出后经 scp 部署到宿主机 ./data/market-analysis）
      - QM_MARKET_SNAPSHOT_DIR=/data/market-analysis
      # QuantUS / QuantHK 本地数据中枢（美股/港股，Yahoo Finance 摄取）
      - QM_QUANTUS_DATA_DIR=/data/quantus
      - QM_QUANTHK_DATA_DIR=/data/quanthk
      # QuantBC 本地数据中枢（区块链/加密货币，Binance 摄取）
      - QM_QUANTBC_DATA_DIR=/data/quantbc
      # QuantFutures 本地数据中枢（期货/贵金属，akshare 摄取）
      - QM_QUANTFUTURES_DATA_DIR=/data/quantfutures
      # 生产环境屏蔽区块链（CRYPTO）市场：默认 false 关闭 crypto 适配器，开发可用 ENABLE_CRYPTO=true 开启
      - ENABLE_CRYPTO=${ENABLE_CRYPTO:-false}

      # AlphaAgent 因子编码专家系统提示词
      - ALPHA_AGENT_SYSTEM_PROMPT=${ALPHA_AGENT_SYSTEM_PROMPT:-}

      # 市值单位配置：数据库存储单位为元，1亿=100,000,000
      - AI_STRATEGY_TOTAL_MV_PER_YI=100000000.0

      # QuantBot 代理（QwenPaw 机器人集成）
      # qwenpaw 容器已加入同一网络，通过 http://qwenpaw:8088 访问
      - COPAW_BASE_URL=${COPAW_BASE_URL:-http://qwenpaw:8088}
      - QWENPAW_BASE_URL=${QWENPAW_BASE_URL:-http://qwenpaw:8088}
      # QwenPaw 上游认证信息（按需配置）
      - QWENPAW_AUTH_USERNAME=${QWENPAW_AUTH_USERNAME:-}
      - QWENPAW_AUTH_PASSWORD=${QWENPAW_AUTH_PASSWORD:-}
      # QwenPaw 共享文件目录
      - QWENPAW_SHARED_FILES_DIR=/qwenpaw-shared

      # Huntly 资讯聚合 (后端代理调用)
      - HUNTLY_BASE_URL=${HUNTLY_BASE_URL:-http://quantmind-huntly}
      - HUNTLY_USERNAME=${HUNTLY_USERNAME:-changeme}
      - HUNTLY_PASSWORD=${HUNTLY_PASSWORD:-changeme}

      # 资讯情绪识别 (FinBERT 中文金融模型, 本地离线权重)
      - NEWS_USE_FINBERT=true
      - FINBERT_ZH_MODEL=/app/models/finbert-zh-base

      # 数据平台 - 第三方离线数据源根目录
      - QM_TDX_DIR=${QM_TDX_DIR:-/data/third_party/tdx}
      - QM_INVESTMENT_DATA_DIR=${QM_INVESTMENT_DATA_DIR:-/data/third_party/investment_data}
      - QM_SIMONLIN_A_DIR=${QM_SIMONLIN_A_DIR:-/data/third_party/simonlin_a_stock}
    volumes:
      # 运行时数据
      - ./data:/data
      # AI 模型文件 (按需下载)
      - ./models:/app/models
      # Qlib 历史数据 (按需下载)
      - ./db:/app/db
      # 日志
      - ./logs:/app/logs
      # 后端代码 (外挂载解决热更新问题)
      - ./backend:/app/backend
      - ./config:/app/config
      - ./scripts:/app/scripts
      # AlphaAgent 包源码（容器内 Python 通过 PYTHONPATH=/app 直接 import）
      - ./alphaagent:/app/alphaagent
      # TradingAgents-Astock 多Agent投研框架
      - ./TradingAgents-astock:/app/TradingAgents-astock
      # 训练脚本（用于 Docker-in-Docker 启动训练容器）
      - ./docker/training:/app/docker/training
      # 策略模板
      - ./strategy_templates:/app/strategy_templates
      # 用户股票池文件 (AI策略生成)
      - ./user_pools_local:/app/user_pools_local
      # RD-Agent conda shim：把 rdagent4qlib 环境映射到容器当前 python（容器内手动改会被重建覆盖）
      - ./docker/conda-shim:/usr/local/bin/conda:ro
      # litellm+pydantic 兼容补丁（因子挖掘 LLM 调用用，见 docker/litellm_sitecustomize.py）
      - ./docker/litellm_sitecustomize.py:/usr/local/lib/python3.10/site-packages/sitecustomize.py:ro
      # RD-Agent 已 pip install 到镜像中，不再需要运行时挂载
      # 允许 API 容器调用宿主机 Docker 启动训练容器
      - /var/run/docker.sock:/var/run/docker.sock
      # QwenPaw 共享文件目录
      - qwenpaw-shared:/qwenpaw-shared
    # TradingAgents 源码已随镜像提供；将其加入模块路径，避免每次启动
    # 都访问 PyPI 安装依赖（离线/受限网络会一直卡在此步骤）。
    command: >
      sh -c "PYTHONPATH=/app/TradingAgents-astock:$$PYTHONPATH python backend/main_oss.py"
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: all
    #           capabilities: [gpu]
    restart: unless-stopped
    networks:
      - quantmind-net

  # 4. Celery Worker (Qlib 回测任务)
  celery-worker:
    image: quantmind-oss:latest
    platform: linux/amd64
    container_name: quantmind-celery
    depends_on:
      db:
        condition: service_healthy
      redis:
        condition: service_healthy
    environment:
      # 应用配置
      - APP_EDITION=oss
      - APP_ENV=production
      - TZ=Asia/Shanghai
      - PYTHONPATH=/app

      # 数据库配置
      - DB_DRIVER=asyncpg
      - DB_HOST=db
      - DB_PORT=5432
      - DB_NAME=${DB_NAME:-quantmind}
      - DB_USER=${DB_USER:-quantmind}
      - DB_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - DATABASE_URL=postgresql+asyncpg://${DB_USER:-quantmind}:${DB_PASSWORD:-quantmind2026}@db:5432/${DB_NAME:-quantmind}

      # Redis 配置
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - REDIS_PASSWORD=

      # 存储配置
      - STORAGE_MODE=local
      - STORAGE_ROOT=/data

      # 融资融券股票池文件路径（容器内）
      - MARGIN_STOCK_POOL_PATH=/data/融资融券.json

      # 策略模板目录
      - STRATEGY_TEMPLATES_DIR=/app/strategy_templates

      # 回测结果存储目录（使用共享卷）
      - QLIB_BACKTEST_RESULT_DIR=/data/backtest_results
      - QLIB_BACKTEST_KERNELS=1

      # 统一报告根目录（技能中心）：复盘/市场分析/个股研究/股票池/投研报告
      - QM_REPORTS_DIR=/data/reports
      - TRADING_AGENTS_RESULTS_DIR=/data/reports/trading_agents

      # 本地 Docker 训练编排（模型训练）
      - INTERNAL_CALL_SECRET=${INTERNAL_CALL_SECRET:-changeme-internal-secret}
      - TRAINING_IMAGE=${TRAINING_IMAGE:-quantmind-trainer:latest}
      - TRAINING_DOCKER_NETWORK=${TRAINING_DOCKER_NETWORK:-quantmind_quantmind-net}
      - HOST_PROJECT_PATH=${HOST_PROJECT_PATH:-/opt/quantmind}
      - QUANTMIND_API_BASE_URL=${QUANTMIND_API_BASE_URL:-http://quantmind:8000}

      # AutoDL 远程训练节点
      - TRAINING_AUTODL_HOST=${TRAINING_AUTODL_HOST:-}
      - TRAINING_AUTODL_SSH_PORT=${TRAINING_AUTODL_SSH_PORT:-22}
      - TRAINING_AUTODL_USER=${TRAINING_AUTODL_USER:-root}
      - TRAINING_AUTODL_SSH_PASSWORD=${TRAINING_AUTODL_SSH_PASSWORD:-}
      - TRAINING_AUTODL_SSH_KEY=${TRAINING_AUTODL_SSH_KEY:-}
      - TRAINING_AUTODL_WORK_DIR=${TRAINING_AUTODL_WORK_DIR:-/workspace}
      - TRAINING_AUTODL_DOCKER_IMAGE=${TRAINING_AUTODL_DOCKER_IMAGE:-quantmind-trainer:latest}
      - TRAINING_AUTODL_NODE_NAME=${TRAINING_AUTODL_NODE_NAME:-AutoDL GPU}
      - TRAINING_AUTODL_GPUS=${TRAINING_AUTODL_GPUS:-all}
      - TRAINING_MASTER_HOST=${TRAINING_MASTER_HOST:-}

      # 回测进度回调地址（OSS 模式使用 quantmind 容器名）
      - BACKTEST_PROGRESS_URL=http://quantmind:8001

      # Qwen/DashScope API (mock key 允许启动，用户需配置真实 key)
      - DASHSCOPE_API_KEY=${DASHSCOPE_API_KEY:-mock-api-key-not-configured}
      - QWEN_API_KEY=${QWEN_API_KEY:-mock-api-key-not-configured}
      # DeepSeek API (用于 TradingAgents 等)
      - DEEPSEEK_API_KEY=${DEEPSEEK_API_KEY:-your-deepseek-api-key}
      - DEEPSEEK_BASE_URL=${DEEPSEEK_BASE_URL:-https://api.deepseek.com}
      - DEEPSEEK_MODEL=${DEEPSEEK_MODEL:-deepseek-chat}

      # AI-IDE Chat LLM 配置（OpenAI 兼容；默认 MiMo）
      - AI_IDE_LLM_BASE_URL=${AI_IDE_LLM_BASE_URL:-https://token-plan-cn.xiaomimimo.com/v1}
      - AI_IDE_LLM_MODEL=${AI_IDE_LLM_MODEL:-mimo-v2.5-pro}
      - AI_IDE_LLM_API_KEY=${AI_IDE_LLM_API_KEY:-mock-api-key-not-configured}

      # QuantDB SDK（数据同步 + 特征快照生成）
      - QUANTDB_API_KEY=${QUANTDB_API_KEY:-}
      - QM_QUANTDB_DATA_DIR=/data/quantdb
      # QuantDB 同步状态库（避免容器内空库导致全量重下）
      - QUANTDB_STATE_DIR=/data/quantdb/.sync_state

      # Huntly 资讯 enrich (news_enrich_recent 任务在 worker 执行, 需要凭据)
      - HUNTLY_BASE_URL=${HUNTLY_BASE_URL:-http://quantmind-huntly}
      - HUNTLY_USERNAME=${HUNTLY_USERNAME:-changeme}
      - HUNTLY_PASSWORD=${HUNTLY_PASSWORD:-changeme}
      # enricher 直连 PG 写入 news_article_enrichment
      - POSTGRES_HOST=db
      - POSTGRES_PORT=5432
      - POSTGRES_USER=${DB_USER:-quantmind}
      - POSTGRES_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - POSTGRES_DB=${DB_NAME:-quantmind}

      # 资讯情绪识别 (FinBERT 中文金融模型, 本地离线权重)
      - NEWS_USE_FINBERT=true
      - FINBERT_ZH_MODEL=/app/models/finbert-zh-base

      # 数据同步定时任务
      - DAILY_SYNC_ENABLED=true

      # 通达信(TDX)交易桥配置（自动推理完成后推送 Top N 选股到通达信）
      - TDX_BRIDGE_URL=${TDX_BRIDGE_URL:-http://192.168.31.31:8550}
      - TDX_BRIDGE_TOKEN=${TDX_BRIDGE_TOKEN:-}
      - ENABLE_TDX_PUSH=${ENABLE_TDX_PUSH:-false}
      - TDX_ACCOUNT=${TDX_ACCOUNT:-}
      - TDX_ACCOUNT_TYPE=${TDX_ACCOUNT_TYPE:-stock}
    volumes:
      # 运行时数据
      - ./data:/data
      # AI 模型文件
      - ./models:/app/models
      # Qlib 历史数据
      - ./db:/app/db
      # 日志
      - ./logs:/app/logs
      # 后端代码
      - ./backend:/app/backend
      - ./config:/app/config
      - ./scripts:/app/scripts
      # 策略模板
      - ./strategy_templates:/app/strategy_templates
      # 用户股票池文件 (AI策略生成)
      - ./user_pools_local:/app/user_pools_local
    command: celery -A backend.services.engine.qlib_app.celery_config:celery_app worker -Q ${QLIB_CELERY_QUEUE:-qlib_backtest_srv} --loglevel=info --concurrency=4
    healthcheck:
      test: ["CMD-SHELL", "pgrep -f 'celery.*worker' || exit 1"]
      interval: 30s
      timeout: 10s
      retries: 3
    restart: unless-stopped
    networks:
      - quantmind-net

  # 4b. Celery beat (定时任务调度: 资讯 enrich / matcher 重载等)
  celery-beat:
    image: quantmind-oss:latest
    platform: linux/amd64
    container_name: quantmind-celery-beat
    depends_on:
      db:
        condition: service_healthy
      redis:
        condition: service_healthy
    environment:
      - APP_EDITION=oss
      - APP_ENV=production
      - TZ=Asia/Shanghai
      - PYTHONPATH=/app
      - DB_HOST=db
      - DB_PORT=5432
      - DB_NAME=${DB_NAME:-quantmind}
      - DB_USER=${DB_USER:-quantmind}
      - DB_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - POSTGRES_HOST=db
      - POSTGRES_USER=${DB_USER:-quantmind}
      - POSTGRES_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - POSTGRES_DB=${DB_NAME:-quantmind}
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - REDIS_PASSWORD=
      - HUNTLY_BASE_URL=http://quantmind-huntly
      - HUNTLY_USERNAME=${HUNTLY_USERNAME:-changeme}
      - HUNTLY_PASSWORD=${HUNTLY_PASSWORD:-changeme}
      - NEWS_ENRICH_ENABLED=true
      - NEWS_ENRICH_INTERVAL_SEC=60
      - NEWS_MATCHER_RELOAD_SEC=600

      # QuantDB SDK（数据同步调度需要）
      - QUANTDB_API_KEY=${QUANTDB_API_KEY:-}
      - QM_QUANTDB_DATA_DIR=/data/quantdb
      - QUANTDB_STATE_DIR=/data/quantdb/.sync_state

      # 数据同步定时任务
      - DAILY_SYNC_ENABLED=true
    volumes:
      - ./backend:/app/backend
      - ./config:/app/config
      - ./scripts:/app/scripts
      - ./logs:/app/logs
      - ./data:/data
      - ./db:/app/db
    command: celery -A backend.services.engine.qlib_app.celery_config:celery_app beat --loglevel=info --schedule=/tmp/celerybeat-schedule
    healthcheck:
      test: ["CMD-SHELL", "pgrep -f 'celery.*beat' || exit 1"]
      interval: 30s
      timeout: 10s
      retries: 3
    restart: unless-stopped
    networks:
      - quantmind-net

  # 5a. Data Gateway (多数据源金融数据网关: akshare/eastmoney)
  data-gateway:
    image: quantmind-data-gateway:latest
    build:
      context: .
      dockerfile: docker/Dockerfile.data-gateway
    container_name: quantmind-data-gateway
    ports:
      - "${DATA_GATEWAY_PORT:-8004}:8004"
    environment:
      - TZ=Asia/Shanghai
      - PYTHONPATH=/app
    healthcheck:
      test: ["CMD-SHELL", "curl -f http://localhost:8004/health || exit 1"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 20s
    restart: unless-stopped
    networks:
      - quantmind-net

  # 5c. Dashboard (OpenBB + Streamlit 数据看板)
  dashboard:
    image: quantmind-dashboard:latest
    build:
      context: .
      dockerfile: docker/Dockerfile.dashboard
    container_name: quantmind-dashboard
    depends_on:
      db:
        condition: service_healthy
      redis:
        condition: service_healthy
    ports:
      - "${DASHBOARD_PORT:-8501}:8501"
    environment:
      - TZ=Asia/Shanghai
      - DB_HOST=db
      - DB_PORT=5432
      - DB_NAME=${DB_NAME:-quantmind}
      - DB_USER=${DB_USER:-quantmind}
      - DB_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - HUNTLY_BASE_URL=http://quantmind-huntly
      - HUNTLY_USERNAME=${HUNTLY_USERNAME:-changeme}
      - HUNTLY_PASSWORD=${HUNTLY_PASSWORD:-changeme}
    volumes:
      - ./data:/data
      - ./db:/app/db
      - ./dashboard:/app/dashboard
    healthcheck:
      test: ["CMD-SHELL", "curl -f http://localhost:8501/_stcore/health >/dev/null 2>&1 || exit 1"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 30s
    restart: unless-stopped
    networks:
      - quantmind-net

  # 6. Huntly (财经资讯/RSS/Twitter 聚合订阅源)
  # 提供新闻流 + 可结构化的财务事件源
  # 前端通过 QuantMind News 代理访问 (零 401 风险)
  huntly:
    image: lcomplete/huntly:latest
    container_name: quantmind-huntly
    ports:
      - "${HUNTLY_PORT:-8090}:80"
    volumes:
      - ./data/huntly:/data
    environment:
      - TZ=Asia/Shanghai
      # 默认管理员账号 (首次启动自动创建)
      - HUNTLY_DEFAULT_USERNAME=${HUNTLY_USERNAME:-changeme}
      - HUNTLY_DEFAULT_PASSWORD=${HUNTLY_PASSWORD:-changeme}
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://127.0.0.1/api/health >/dev/null 2>&1 || exit 1"]
      interval: 30s
      timeout: 5s
      retries: 5
      start_period: 60s
    restart: unless-stopped
    networks:
      - quantmind-net

  # 6b. RSSHub (为 Huntly 提供 RSS 源生成 — Twitter/微博/雪球/财经站点等)
  # Huntly 端订阅地址示例：http://rsshub:1200/twitter/user/<handle>
  rsshub:
    image: diygod/rsshub:latest
    container_name: quantmind-rsshub
    ports:
      - "${RSSHUB_PORT:-1200}:1200"
    environment:
      - NODE_ENV=production
      - TZ=Asia/Shanghai
      - CACHE_TYPE=redis
      - REDIS_URL=redis://redis:6379/3
      - CACHE_EXPIRE=300
      - CACHE_CONTENT_EXPIRE=3600
      - REQUEST_RETRY=2
      - LISTEN_INADDR_ANY=1
      - PUPPETEER_SKIP_DOWNLOAD=true
      # 可选：填入 .env 后可启用 Twitter/微博等需登录的源
      # - TWITTER_USERNAME=${TWITTER_USERNAME:-}
      # - TWITTER_PASSWORD=${TWITTER_PASSWORD:-}
    depends_on:
      redis:
        condition: service_healthy
    healthcheck:
      # 新版 diygod/rsshub 镜像不含 wget，改用 node fetch 探测
      test: ["CMD-SHELL", "node -e \"fetch('http://127.0.0.1:1200/healthz').then(r=>{if(!r.ok)process.exit(1)}).catch(()=>process.exit(1))\""]
      interval: 30s
      timeout: 5s
      retries: 5
      start_period: 30s
    restart: unless-stopped
    networks:
      - quantmind-net

  # 7. QwenPaw (QuantBot 聊天机器人)
  qwenpaw:
    # Docker registry mirror 由部署脚本统一配置。
    image: agentscope/qwenpaw:latest
    container_name: qwenpaw
    ports:
      - "${QWENPAW_PORT:-8088}:8088"
    networks:
      quantmind-net:
        aliases:
          - copaw
    environment:
      # QuantMind 代码库和数据库访问（与 quantmind 容器一致）
      - PYTHONPATH=/app
      - DB_DRIVER=asyncpg
      - DB_HOST=db
      - DB_PORT=5432
      - DB_NAME=${DB_NAME:-quantmind}
      - DB_USER=${DB_USER:-quantmind}
      - DB_PASSWORD=${DB_PASSWORD:-quantmind2026}
      - DATABASE_URL=postgresql+asyncpg://${DB_USER:-quantmind}:${DB_PASSWORD:-quantmind2026}@db:5432/${DB_NAME:-quantmind}
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - REDIS_PASSWORD=
      - STORAGE_MODE=local
      - STORAGE_ROOT=/data
      - MARGIN_STOCK_POOL_PATH=/data/融资融券.json
      - STRATEGY_TEMPLATES_DIR=/app/strategy_templates
      - QLIB_BACKTEST_RESULT_DIR=/data/backtest_results
      - QLIB_BACKTEST_KERNELS=1
      - INTERNAL_CALL_SECRET=${INTERNAL_CALL_SECRET:-changeme-internal-secret}
      - QUANTMIND_API_BASE_URL=http://quantmind:8000
      - QUANTMIND_ENGINE_URL=http://quantmind:8001
      - QUANTMIND_USER_ID=qwenpaw
      - QWENPAW_SHARED_FILES_DIR=/qwenpaw-shared
      - QWENPAW_CORS_ORIGINS=http://192.168.31.68:3080,http://localhost:3080,http://127.0.0.1:3080,http://192.168.31.68:8088,http://localhost:8088,http://127.0.0.1:8088
      # 远程数据源（用于数据同步脚本）
      - SOURCE_DATABASE_URL=${SOURCE_DATABASE_URL:-postgresql://user:pass@localhost:5432/quantmind}
      - SYNC_MAX_DAYS=30
      - TZ=Asia/Shanghai
    volumes:
      # QwenPaw 自有数据
      - qwenpaw-data:/app/working
      - qwenpaw-secrets:/app/working.secret
      - qwenpaw-backups:/app/working.backups
      # QuantMind 代码库（只读，避免误修改）
      - ./backend:/app/backend:ro
      - ./config:/app/config:ro
      - ./scripts:/app/scripts:ro
      - ./strategy_templates:/app/strategy_templates:ro
      - ./CLAUDE.md:/app/CLAUDE.md:ro
      # 项目根目录（只读，供 Agent 查找本地文件/技能 zip 等）
      - ./:/quantmind:ro
      # RD-Agent 已 pip install 到镜像中，不再需要运行时挂载
      # QuantMind 数据（读写）
      - ./data:/data
      - ./db:/app/db
      - ./models:/app/models
      - ./logs:/app/logs
      - ./user_pools_local:/app/user_pools_local
      # 共享文件目录
      - qwenpaw-shared:/qwenpaw-shared
      # 允许调用宿主机 Docker（启动训练容器等）
      - /var/run/docker.sock:/var/run/docker.sock
    restart: unless-stopped

  # ── IB Gateway（盈透证券官方安装包 + IBC 自动登录，社区标准镜像）─────────
  # 在 .env 配置 IB_ACCOUNT / IB_PASSWORD / IB_TRADING_MODE(paper|live)。
  # 4001=实盘 API，4002=模拟 API，5901=VNC（首次登录/2FA 用）。
  # 安全提示：端口仅限局域网/内网使用，切勿暴露公网。
  ib-gateway:
    image: ghcr.io/gnzsnz/ib-gateway:latest
    container_name: ib-gateway
    environment:
      IB_ACCOUNT: ${IB_ACCOUNT:-}
      IB_PASSWORD: ${IB_PASSWORD:-}
      TRADING_MODE: ${IB_TRADING_MODE:-paper}
      VNC_SERVER_PASSWORD: ${IB_VNC_PASSWORD:-quantmind-vnc}
      TWOFA_TIMEOUT_ACTION: restart
    ports:
      - "4001:4001"
      - "4002:4002"
      - "127.0.0.1:15901:5901"
    restart: unless-stopped
    networks:
      - quantmind-net

networks:
  quantmind-net:
    driver: bridge

volumes:
  postgres-data:
  redis-data:
  qwenpaw-data:
  qwenpaw-secrets:
  qwenpaw-backups:
  qwenpaw-shared: