qlib/docs/diagrams/qlib-dataflow.json

{
  "schema_version": 1,
  "diagram_type": "dataflow",
  "meta": {
    "title": "A股量化研究平台 · 数据流水",
    "quality_profile": "showcase",
    "viewBox": [
      1300,
      693
    ],
    "views": [
      {
        "id": "main-path",
        "label": "主研究数据流",
        "focus": [
          "tushare",
          "provider",
          "parquet",
          "dataset",
          "factor",
          "backtest",
          "experiment",
          "api",
          "web"
        ],
        "note": "行情从 Tushare 一路流转到可复现的实验与可视化。"
      },
      {
        "id": "future-free",
        "label": "防未来函数口径",
        "focus": [
          "provider",
          "sqlite",
          "parquet",
          "dataset"
        ],
        "note": "财务数据按 announce_date 控制可见性，查询一律带 as_of_date。"
      },
      {
        "id": "agent-path",
        "label": "AI 研究助手消费",
        "focus": [
          "experiment",
          "api",
          "web"
        ],
        "note": "结果经 API 提供给 Web UI；AI Agent（Phase 5）仅经受控 Tool 经 API 消费研究数据。"
      }
    ]
  },
  "stages": [
    {
      "label": "数据接入"
    },
    {
      "label": "数据落地"
    },
    {
      "label": "特征构建"
    },
    {
      "label": "研究计算"
    },
    {
      "label": "回测 · 实验 · 消费"
    }
  ],
  "nodes": [
    {
      "id": "tushare",
      "type": "external",
      "label": "Tushare",
      "sublabel": "日线 / 财务 / 交易日历",
      "tag": "首选",
      "stage": 0,
      "row": 0,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "provider",
      "type": "backend",
      "label": "Data Provider",
      "sublabel": "归一化 / 校验 / 审计",
      "tag": "MarketDataProvider",
      "stage": 0,
      "row": 2,
      "width": 130,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "sina",
      "type": "external",
      "label": "新浪财经",
      "sublabel": "缺失兜底 / 交叉验证",
      "tag": "备用",
      "stage": 0,
      "row": 4,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "sqlite",
      "type": "database",
      "label": "SQLite",
      "sublabel": "股票 / 日历 / 因子定义 / Job",
      "tag": "业务库",
      "stage": 1,
      "row": 0,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "parquet",
      "type": "database",
      "label": "Parquet",
      "sublabel": "历史行情 / 复权 / 财务时序",
      "tag": "时序库",
      "stage": 1,
      "row": 2,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "dataset",
      "type": "database",
      "label": "Qlib Dataset",
      "sublabel": "data/qlib 特征集",
      "tag": "研究数据",
      "stage": 2,
      "row": 2,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "factor",
      "type": "backend",
      "label": "因子研究",
      "sublabel": "Alpha158 / 自定义因子",
      "tag": "IC · 分层测试",
      "stage": 3,
      "row": 0,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "model",
      "type": "backend",
      "label": "模型训练",
      "sublabel": "LightGBM",
      "tag": "预测信号",
      "stage": 3,
      "row": 2,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "backtest",
      "type": "backend",
      "label": "选股回测",
      "sublabel": "调仓 / 成本 / 涨跌停 / 停牌",
      "tag": "低频策略",
      "stage": 4,
      "row": 0,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "experiment",
      "type": "database",
      "label": "Experiment",
      "sublabel": "参数 / 结果 / 数据与代码版本",
      "tag": "可复现",
      "stage": 4,
      "row": 1,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "api",
      "type": "backend",
      "label": "FastAPI API",
      "sublabel": "业务 API（Pydantic DTO）",
      "tag": "REST / SSE",
      "stage": 4,
      "row": 3,
      "height": 36,
      "yOffset": -2
    },
    {
      "id": "web",
      "type": "frontend",
      "label": "Web UI",
      "sublabel": "Next.js · Lightweight Charts",
      "tag": "结果可视化",
      "stage": 4,
      "row": 4,
      "height": 36,
      "yOffset": -2
    }
  ],
  "flows": [
    {
      "id": "raw-tushare",
      "from": "tushare",
      "to": "provider",
      "label": "行情 / 财务 raw",
      "variant": "emphasis"
    },
    {
      "id": "raw-sina",
      "from": "sina",
      "to": "provider",
      "label": "兜底 raw",
      "variant": "dashed"
    },
    {
      "id": "to-sqlite",
      "from": "provider",
      "to": "sqlite",
      "label": "股票 / 日历 / 元数据",
      "labelAt": [
        212,
        256
      ]
    },
    {
      "id": "to-parquet",
      "from": "provider",
      "to": "parquet",
      "label": "行情 / 财务时序",
      "route": "straight",
      "labelAt": [
        212,
        426
      ]
    },
    {
      "id": "calendar-to-dataset",
      "from": "sqlite",
      "to": "dataset",
      "label": "交易日历 / 股票池",
      "labelAt": [
        422.5,
        276
      ]
    },
    {
      "id": "series-to-dataset",
      "from": "parquet",
      "to": "dataset",
      "label": "行情特征输入",
      "route": "straight",
      "labelAt": [
        452,
        426
      ]
    },
    {
      "id": "dataset-to-factor",
      "from": "dataset",
      "to": "factor",
      "label": "特征矩阵",
      "variant": "emphasis",
      "labelAt": [
        637.5,
        268
      ]
    },
    {
      "id": "dataset-to-model",
      "from": "dataset",
      "to": "model",
      "label": "训练 / 预测特征",
      "route": "straight",
      "labelAt": [
        610,
        426
      ]
    },
    {
      "id": "factor-signal",
      "from": "factor",
      "to": "backtest",
      "label": "因子信号"
    },
    {
      "id": "model-signal",
      "from": "model",
      "to": "backtest",
      "label": "预测信号",
      "labelAt": [
        852.5,
        298
      ]
    },
    {
      "id": "factor-archive",
      "from": "factor",
      "to": "experiment",
      "label": "因子测试归档",
      "fromSide": "top",
      "toSide": "right",
      "via": [
        [
          745,
          98
        ],
        [
          1060,
          98
        ],
        [
          1060,
          258
        ]
      ],
      "labelAt": [
        1060,
        176
      ]
    },
    {
      "id": "backtest-archive",
      "from": "backtest",
      "to": "experiment",
      "label": "回测结果归档",
      "variant": "emphasis"
    },
    {
      "id": "query-result",
      "from": "experiment",
      "to": "api",
      "label": "结果查询 DTO"
    },
    {
      "id": "to-web",
      "from": "api",
      "to": "web",
      "label": "REST / SSE"
    }
  ],
  "cards": [
    {
      "dot": "emerald",
      "title": "主研究数据流",
      "items": [
        "Tushare → Data Provider（归一化 / 校验 / 审计）→ SQLite 与 Parquet 分流",
        "Qlib Dataset 汇总日历与行情 → 因子 / 模型 → 回测 → Experiment 全量归档",
        "每条拉取记录 source / 行数 / 时间，来源可追溯"
      ]
    },
    {
      "dot": "violet",
      "title": "防未来函数口径",
      "items": [
        "财务数据区分 report_date / announce_date，仅 announce_date 后可见",
        "研究查询一律带 as_of_date，禁止未来成分 / 行业 / 财务数据",
        "新浪仅作缺失兜底与交叉验证，禁止静默切换"
      ]
    },
    {
      "dot": "amber",
      "title": "结果消费",
      "items": [
        "Experiment → FastAPI API → Web UI（REST / SSE）可视化",
        "AI Agent（Phase 5）仅经受控 Tool 经 API 读取研究数据与实验",
        "历史实验含数据与代码版本，可一键复跑"
      ]
    }
  ]
}