feat: 数据治理 — 入库约束+元数据卡片+编码清洗+审计看板

backend/app/api/alert_rules.py

@@ -260,7 +260,7 @@ def run_all_alert_checks(db: Session = Depends(get_db)):
             
             value = latest_value.actual_value
             period = latest_value.period
-            params = rule.params or {}
+            import json; params = json.loads(rule.params) if isinstance(rule.params, str) else (rule.params or {})
             
             alert_level = None
             alert_message = None
@@ -534,7 +534,7 @@ def _check_forecast_alerts(db: Session) -> int:
                 continue
 
             # 检查预测值是否超限
-            params = rule.params or {}
+            import json; params = json.loads(rule.params) if isinstance(rule.params, str) else (rule.params or {})
             params["kpi"] = kpi
             for forecast in latest_forecasts:
                 value = forecast.predicted_cash

backend/app/api/data_quality.py

@@ -253,10 +253,72 @@ def quality_stats(db: Session = Depends(get_db)):
         if cnt:
             type_counts[t] = cnt
 
+    # ── 数据审计看板统计 ──
+    # KPI完整度评分
+    all_kpis = db.query(KPIDefinition).filter(KPIDefinition.status == "active").all()
+    total = len(all_kpis)
+    complete_kpis = 0
+    missing_metadata_count = 0
+    missing_data_count = 0
+    stale_data_count = 0
+
+    from datetime import datetime, timedelta
+    six_months_ago = datetime.now() - timedelta(days=180)
+
+    for kpi in all_kpis:
+        # 元数据完整度检查
+        has_meta = all([
+            kpi.formula and kpi.formula.strip(),
+            kpi.data_source and kpi.data_source.strip(),
+            kpi.data_owner and kpi.data_owner.strip(),
+            kpi.unit and kpi.unit.strip(),
+            kpi.target_value is not None,
+        ])
+        if has_meta:
+            complete_kpis += 1
+        else:
+            missing_metadata_count += 1
+
+        # 数据缺失检查（是否有实际值）
+        val = db.query(KPIValue).filter(
+            KPIValue.kpi_id == kpi.id,
+            KPIValue.actual_value.isnot(None),
+        ).first()
+        if not val:
+            missing_data_count += 1
+
+        # 超30天未更新预警
+        latest_val = db.query(KPIValue).filter(
+            KPIValue.kpi_id == kpi.id,
+            KPIValue.actual_value.isnot(None),
+        ).order_by(KPIValue.period.desc()).first()
+        if latest_val and latest_val.calculated_at:
+            if latest_val.calculated_at < six_months_ago:
+                stale_data_count += 1
+
+    completeness_score = round(complete_kpis / total * 100, 1) if total > 0 else 0
+    missing_rate = round(missing_data_count / total * 100, 1) if total > 0 else 0
+
     return {
         "total_kpis": total_kpis,
         "total_logs": total_logs,
         "open_logs": open_logs,
         "severity_counts": severity_counts,
         "type_counts": type_counts,
+        # 数据审计看板
+        "completeness": {
+            "score": completeness_score,
+            "complete": complete_kpis,
+            "total": total,
+            "missing_metadata": missing_metadata_count,
+        },
+        "data_missing": {
+            "count": missing_data_count,
+            "rate": missing_rate,
+            "total": total,
+        },
+        "stale_data": {
+            "count": stale_data_count,
+            "threshold_days": 180,
+        },
     }

backend/app/api/kpis.py

@@ -406,12 +406,43 @@ def get_kpi(kpi_id: int, db: Session = Depends(get_db)):
     return kpi_to_dict(kpi)
 
 
+def _validate_kpi_data(data: dict, is_update: bool = False):
+    """数据治理：入库必检 + 元数据校验"""
+    errors = []
+
+    # 规则1: target_value 必填
+    tv = data.get("target_value")
+    if tv is None or (isinstance(tv, (int, float)) and tv < 0 and not is_update):
+        if not is_update or "target_value" in data:
+            if tv is None:
+                errors.append("目标值(target_value)不能为空")
+
+    # 规则1: unit 必填
+    unit = data.get("unit")
+    if not unit or (isinstance(unit, str) and unit.strip() == ""):
+        if not is_update or "unit" in data:
+            errors.append("单位(unit)不能为空")
+
+    # 规则2: 元数据必填 — formula/data_source/data_owner
+    for field, label in [("formula", "计算公式"), ("data_source", "数据来源"), ("data_owner", "数据责任人")]:
+        val = data.get(field)
+        if not val or (isinstance(val, str) and val.strip() == ""):
+            if not is_update or field in data:
+                errors.append(f"元数据字段'{label}'({field})不能为空")
+
+    return errors
+
+
 @router.post("")
 def create_kpi(data: dict, db: Session = Depends(get_db), user=WRITE_ROLES):
     # 检查编码唯一性
     existing = db.query(KPIDefinition).filter(KPIDefinition.kpi_code == data.get("kpi_code", "")).first()
     if existing:
         raise HTTPException(400, f"KPI编码 {data['kpi_code']} 已存在")
+    # 数据治理校验
+    errs = _validate_kpi_data(data, is_update=False)
+    if errs:
+        raise HTTPException(422, detail={"message": "数据校验不通过", "errors": errs})
     kpi = KPIDefinition(**data)
     db.add(kpi)
     db.commit()
@@ -425,6 +456,10 @@ def update_kpi(kpi_id: int, data: dict, db: Session = Depends(get_db), user=WRIT
     kpi = db.query(KPIDefinition).filter(KPIDefinition.id == kpi_id).first()
     if not kpi:
         raise HTTPException(404, "KPI不存在")
+    # 数据治理校验（更新时只检查传了但为空的字段）
+    errs = _validate_kpi_data(data, is_update=True)
+    if errs:
+        raise HTTPException(422, detail={"message": "数据校验不通过", "errors": errs})
     for k, v in data.items():
         if hasattr(kpi, k) and v is not None:
             setattr(kpi, k, v)

backend/app/models/__init__.py

@@ -59,6 +59,8 @@ class KPIDefinition(Base):
     formula_desc = Column(String(500), nullable=True, comment="公式说明")
     data_source_type = Column(String(20), default="manual", comment="erp/business/excel/manual")
     data_source_config = Column(JSON, nullable=True, comment="数据源配置")
+    data_source = Column(String(500), default="待补充", comment="数据来源")
+    data_owner = Column(String(100), default="待指定", comment="数据责任人")
     frequency = Column(String(20), default="monthly", comment="daily/weekly/monthly/quarterly/yearly")
     unit = Column(String(50), default="%", comment="单位")
     target_value = Column(Float, nullable=True, comment="目标值")

backend/scripts/cleanup_kpi_encoding.py

@@ -0,0 +1,126 @@
+"""数据治理：编码规范清洗 — 检查KPI编码前缀与维度一致性 + 修复误分类"""
+import pymysql
+import os
+import logging
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger("encoding-cleanup")
+
+DB_USER = os.getenv("CMA_DB_USER", "cma_user")
+DB_PASS = os.getenv("CMA_DB_PASS", "cma_pass_2026")
+DB_HOST = os.getenv("CMA_DB_HOST", "127.0.0.1")
+DB_PORT = int(os.getenv("CMA_DB_PORT", "3306"))
+DB_NAME = os.getenv("CMA_DB_NAME", "cma")
+
+conn = pymysql.connect(
+    host=DB_HOST, port=DB_PORT, user=DB_USER, password=DB_PASS, database=DB_NAME,
+    charset="utf8mb4", cursorclass=pymysql.cursors.DictCursor,
+)
+cur = conn.cursor()
+
+# 编码前缀 → 正确维度映射
+PREFIX_DIM_MAP = {
+    "F_": "finance",
+    "C_": "customer",
+    "P_": "process",
+    "L_": "learning",
+}
+
+# 已知误分类修复（编码 → 正确维度）
+KNOWN_FIXES = {
+    "F_QUALITY_RATE": "process",   # 产品合格率 → 流程层
+    "F_REWORK_RATE": "process",    # 返工率 → 流程层
+    "P_COST_CUT": "process",       # 招待费砍半 → 流程层
+    "P_TRAIN_PASS": "process",     # Model C考核通过 → 流程层
+}
+
+
+def run():
+    logger.info("=== 编码规范清洗 开始 ===")
+
+    cur.execute("SELECT id, entity_id, kpi_code, kpi_name, dimension FROM kpi_definitions WHERE status = 'active'")
+    kpis = cur.fetchall()
+
+    issues = []
+    fixes_applied = 0
+
+    for kpi in kpis:
+        kpi_code = kpi["kpi_code"]
+        current_dim = kpi["dimension"]
+        entity_id = kpi["entity_id"]
+
+        # 检查前缀
+        prefix = kpi_code[:2] if len(kpi_code) >= 2 else ""
+        expected_dim = PREFIX_DIM_MAP.get(prefix)
+
+        if expected_dim and current_dim != expected_dim:
+            # 先检查是否在已知修复列表
+            correct_dim = KNOWN_FIXES.get(kpi_code, expected_dim)
+            issues.append({
+                "kpi_code": kpi_code,
+                "kpi_name": kpi["kpi_name"],
+                "current_dim": current_dim,
+                "expected_dim": correct_dim,
+                "prefix": prefix,
+                "entity_id": entity_id,
+            })
+
+            if kpi_code in KNOWN_FIXES:
+                logger.info(f"  🔧 修复: {kpi_code} ({kpi['kpi_name']}) {current_dim} → {correct_dim} (entity={entity_id})")
+                cur.execute(
+                    "UPDATE kpi_definitions SET dimension = %s WHERE id = %s",
+                    (correct_dim, kpi["id"]),
+                )
+                fixes_applied += 1
+
+    conn.commit()
+
+    # 输出报告
+    logger.info(f"\n=== 清洗报告 ===")
+    logger.info(f"  检查KPI总数: {len(kpis)}")
+    logger.info(f"  编码-维度不一致: {len(issues)}")
+    logger.info(f"  已自动修复: {fixes_applied}")
+
+    if issues:
+        logger.info(f"\n  不一致详情:")
+        for i, iss in enumerate(issues, 1):
+            status = "✅ 已修复" if iss["kpi_code"] in KNOWN_FIXES else "⚠️ 需人工确认"
+            logger.info(f"    {i}. {iss['kpi_code']} ({iss['kpi_name']}) "
+                        f"当前维度={iss['current_dim']}, 期望维度={iss['expected_dim']} [{status}]")
+
+    # 检查未命名规范问题
+    logger.info(f"\n  编码前缀统计:")
+    for prefix, dim in PREFIX_DIM_MAP.items():
+        cur.execute("SELECT COUNT(*) as cnt FROM kpi_definitions WHERE kpi_code LIKE %s AND status='active'", (f"{prefix}%",))
+        row = cur.fetchone()
+        cnt = row["cnt"] if row else 0
+        logger.info(f"    {prefix} → {dim}: {cnt} 个KPI")
+
+    # 检查前缀不匹配编码
+    cur.execute("""
+        SELECT kpi_code, dimension FROM kpi_definitions
+        WHERE status='active'
+        AND (
+            (kpi_code LIKE 'F_%' AND dimension != 'finance')
+            OR (kpi_code LIKE 'C_%' AND dimension != 'customer')
+            OR (kpi_code LIKE 'P_%' AND dimension != 'process')
+            OR (kpi_code LIKE 'L_%' AND dimension != 'learning')
+        )
+    """)
+    remaining = cur.fetchall()
+    if remaining:
+        logger.warning(f"\n  ⚠️ 仍有 {len(remaining)} 个KPI编码前缀与维度不匹配:")
+        for r in remaining:
+            logger.warning(f"    {r['kpi_code']} → {r['dimension']}")
+    else:
+        logger.info(f"\n  ✅ 所有KPI编码前缀与维度一致！")
+
+    logger.info("\n=== 编码规范清洗 完成 ===")
+
+
+if __name__ == "__main__":
+    try:
+        run()
+    finally:
+        cur.close()
+        conn.close()

backend/scripts/migrate_data_governance.py

@@ -0,0 +1,105 @@
+"""数据治理 migration: 入库必检约束 + 元数据字段补充"""
+import pymysql
+import os
+import logging
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger("data-governance")
+
+DB_USER = os.getenv("CMA_DB_USER", "cma_user")
+DB_PASS = os.getenv("CMA_DB_PASS", "cma_pass_2026")
+DB_HOST = os.getenv("CMA_DB_HOST", "127.0.0.1")
+DB_PORT = int(os.getenv("CMA_DB_PORT", "3306"))
+DB_NAME = os.getenv("CMA_DB_NAME", "cma")
+
+conn = pymysql.connect(
+    host=DB_HOST, port=DB_PORT, user=DB_USER, password=DB_PASS, database=DB_NAME,
+    charset="utf8mb4", cursorclass=pymysql.cursors.DictCursor,
+)
+cur = conn.cursor()
+
+
+def run():
+    logger.info("=== 数据治理 Migration 开始 ===")
+
+    # ── 0. 添加缺失列 ──
+    logger.info("[步骤0] 检查并补充缺失列...")
+
+    for col_name, col_def in [
+        ("data_source", "ALTER TABLE kpi_definitions ADD COLUMN data_source VARCHAR(500) DEFAULT NULL COMMENT '数据来源' AFTER data_source_type"),
+        ("data_owner", "ALTER TABLE kpi_definitions ADD COLUMN data_owner VARCHAR(100) DEFAULT NULL COMMENT '数据责任人' AFTER data_source"),
+    ]:
+        cur.execute("SHOW COLUMNS FROM kpi_definitions LIKE %s", (col_name,))
+        if not cur.fetchone():
+            logger.info(f"  添加 {col_name} 列...")
+            cur.execute(col_def)
+            conn.commit()
+            logger.info(f"  {col_name} 列已添加")
+        else:
+            logger.info(f"  {col_name} 列已存在")
+
+    # ── 1. 修复空值 ──
+    logger.info("[步骤1] 修复空值...")
+
+    for col, default, label in [
+        ("target_value", "0", "NULL target_value"),
+        ("unit", "'-'", "NULL/empty unit"),
+        ("formula", "'待补充'", "NULL/empty formula"),
+        ("data_source", "'待补充'", "NULL/empty data_source"),
+        ("data_owner", "'待指定'", "NULL/empty data_owner"),
+    ]:
+        if col in ("target_value",):
+            r = cur.execute(f"SELECT COUNT(*) as cnt FROM kpi_definitions WHERE {col} IS NULL")
+        else:
+            r = cur.execute(f"SELECT COUNT(*) as cnt FROM kpi_definitions WHERE {col} IS NULL OR {col} = ''")
+        row = cur.fetchone()
+        cnt = row["cnt"] if row else 0
+        logger.info(f"  {label}: {cnt} 条")
+        if cnt > 0:
+            if col in ("target_value",):
+                cur.execute(f"UPDATE kpi_definitions SET {col} = {default} WHERE {col} IS NULL")
+            else:
+                cur.execute(f"UPDATE kpi_definitions SET {col} = {default} WHERE {col} IS NULL OR {col} = ''")
+            logger.info(f"  已修复 {cur.rowcount} 条")
+
+    conn.commit()
+
+    # ── 2. 修改列约束为 NOT NULL ──
+    logger.info("[步骤2] 修改列约束...")
+
+    alters = [
+        ("target_value", "DECIMAL(15,2) NOT NULL DEFAULT 0"),
+        ("unit", "VARCHAR(50) NOT NULL DEFAULT '-'"),
+        ("formula", "TEXT NOT NULL"),
+        ("data_source", "VARCHAR(500) NOT NULL DEFAULT '待补充'"),
+        ("data_owner", "VARCHAR(100) NOT NULL DEFAULT '待指定'"),
+    ]
+    for col, col_type in alters:
+        col_comment = {
+            "target_value": "目标值", "unit": "单位", "formula": "计算公式",
+            "data_source": "数据来源", "data_owner": "数据责任人",
+        }[col]
+        try:
+            cur.execute(f"ALTER TABLE kpi_definitions MODIFY {col} {col_type} COMMENT '{col_comment}'")
+            logger.info(f"  {col} → {col_type}")
+        except Exception as e:
+            logger.warning(f"  {col} 修改失败: {e}")
+
+    conn.commit()
+
+    # ── 3. 验证 ──
+    logger.info("[步骤3] 验证约束...")
+    cur.execute("DESCRIBE kpi_definitions")
+    for col in cur.fetchall():
+        if col['Field'] in ('target_value', 'unit', 'formula', 'data_source', 'data_owner'):
+            logger.info(f"  {col['Field']}: Null={col['Null']}, Default={col['Default']}, Type={col['Type']}")
+
+    logger.info("=== 数据治理 Migration 完成 ===")
+
+
+if __name__ == "__main__":
+    try:
+        run()
+    finally:
+        cur.close()
+        conn.close()