fix(deploy): pnpm 版本断言不一致时先 corepack 自愈，仍不一致则仍 fail-closed exit 1

- MapCanvas: 层聚焦与回全景联动；右栏停用改自绘可拖详情浮层；目标树默认收起+localStorage 持久化；底部跨层关系摘要条；聚焦态只绘聚焦层同层连线
- StrategyLayer: 层标题可点击聚焦并标「聚焦中」；非聚焦层压缩为 ≤48px 单行摘要（层名—达成率/目标数/mini 色条）
- MapSummaryBar: 汇总条压缩 ≤44px、5 项同排；窄屏无横向溢出
- MapEditBar: 新增「全景 / L1–L4」聚焦切换 chip（内联 SVG，无 emoji）
- 只读渲染层改动：不写 canvas_data / map_connections，无新增写链路调用

backend/app/api/alert_rules.py

@@ -2,7 +2,7 @@
 预警规则智能化 — 任务6
 后端组件: alert_rules 模型 + API + 预警引擎
 """
-from fastapi import APIRouter, Depends, HTTPException, Query
+from fastapi import APIRouter, Body, Depends, HTTPException, Query
 from sqlalchemy.orm import Session
 from sqlalchemy import text, Column, Integer, String, Text, Float, DateTime, JSON, Boolean, func
 from typing import Optional, List
@@ -26,6 +26,8 @@ from app.utils.cash_thresholds import (
     CASH_GAP_RULE_TYPE,
     validate_params as validate_cash_gap_params,
 )
+# 应收账龄逾期预警（T1 · 2026-10-20）：计算/落库单源在服务层，阈值全走 alert_rules.params
+from app.services import ar_overdue_alert
 
 logger = logging.getLogger("alert_rules")
 
@@ -38,6 +40,7 @@ VALID_RULE_TYPES = (
     "forecast_deviation",   # 预测偏差（升级2b 2026-08-25）
     "customer_concentration",  # 客户集中度（P2 2026-09-10）
     "cash_gap_threshold",   # 资金缺口阈值配置（G1 2026-10-17，配置行，不进预警生成链）
+    "ar_overdue",       # 应收账龄逾期（T1 2026-10-20：90天以上占比 / 逐户逾期额 / 逾期户数环比）
 )
 
 # ============================================================
@@ -142,6 +145,11 @@ def create_alert_rule(data: dict, db: Session = Depends(get_db), user=Depends(re
         err = validate_cash_gap_params(data.get("params"))
         if err:
             raise HTTPException(400, err)
+    # T1（2026-10-20）：ar_overdue 参数只校结构（阈值/责任人全量由 DB 行给，代码不设默认值）
+    if rule_type == ar_overdue_alert.RULE_TYPE:
+        err = ar_overdue_alert.validate_params(data.get("params"))
+        if err:
+            raise HTTPException(400, err)
 
     rule = AlertRule(
         entity_id=entity_id,
@@ -180,6 +188,12 @@ def update_alert_rule(rule_id: int, data: dict, db: Session = Depends(get_db), e
         if err:
             db.rollback()
             raise HTTPException(400, err)
+    # T1（2026-10-20）：ar_overdue 改参同样只校结构（阈值可现场改，无需发版）
+    if rule.rule_type == ar_overdue_alert.RULE_TYPE:
+        err = ar_overdue_alert.validate_params(rule.params)
+        if err:
+            db.rollback()
+            raise HTTPException(400, err)
     db.commit()
     db.refresh(rule)
     return {"data": {c.name: getattr(rule, c.name) for c in AlertRule.__table__.columns}}
@@ -307,6 +321,31 @@ def run_all_alert_checks(db: Session = Depends(get_db), entity_id: int = Depends
             # 不进 KPIValue 比价链（否则未来若绑定到有值的 KPI 会误生成预警）。
             if rule.rule_type == CASH_GAP_RULE_TYPE:
                 continue
+            # T1（2026-10-20）：应收账龄逾期 —— 与 POST /run-ar-overdue 同一计算函数，
+            # 不走 KPIValue 比价链（F_AR_002 无本期 kpi_values 也不得报错）。
+            if rule.rule_type == ar_overdue_alert.RULE_TYPE:
+                rp = rule.params
+                if isinstance(rp, str):
+                    rp = json.loads(rp) if rp else {}
+                rp = rp or {}
+                try:
+                    ar_overdue_alert.ensure_assignee_configured(rp)
+                    res = ar_overdue_alert.compute(db, entity_id, None, rp)
+                except (ar_overdue_alert.AssigneeMissing, ar_overdue_alert.ParamsInvalid) as exc:
+                    logger.warning("ar_overdue 规则跳过（rule_id=%s）: %s", rule.id, exc)
+                    continue
+                if res.get("skipped_reason"):
+                    logger.info("ar_overdue 规则无数据跳过（rule_id=%s）: %s",
+                                rule.id, res["skipped_reason"])
+                    continue
+                try:
+                    w = ar_overdue_alert.write_alerts(db, entity_id, rule.kpi_id, res)
+                except Exception:  # noqa: BLE001 — 单规则失败不拖垮整轮检查
+                    db.rollback()
+                    logger.exception("ar_overdue 落库失败（rule_id=%s）", rule.id)
+                    continue
+                alerts_generated += w["inserted"]
+                continue
             kpi = kpi_cache.get(rule.kpi_id)
             if kpi is None:
                 kpi = db.query(KPIDefinition).filter(KPIDefinition.id == rule.kpi_id).first()
@@ -316,7 +355,7 @@ def run_all_alert_checks(db: Session = Depends(get_db), entity_id: int = Depends
             if not kpi or kpi.entity_id != entity_id:
                 continue
 
-            import json; params = json.loads(rule.params) if isinstance(rule.params, str) else (rule.params or {})
+            params = json.loads(rule.params) if isinstance(rule.params, str) else (rule.params or {})
 
             # 客户集中度规则（P2 2026-09-10）：不依赖同期KPIValue，直接看客户收入占比
             if rule.rule_type == "customer_concentration":
@@ -655,7 +694,7 @@ def _check_forecast_alerts(db: Session, entity_id: int = 1) -> int:
                 continue
 
             # 检查预测值是否超限
-            import json; params = json.loads(rule.params) if isinstance(rule.params, str) else (rule.params or {})
+            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
@@ -979,3 +1018,136 @@ def run_customer_concentration_check(
             continue
     db.commit()
     return {"message": f"客户集中度检查完成: {generated}条", "generated": generated}
+
+
+# ============================================================
+# T1（2026-10-20）：应收账龄逾期预警 —— 运行入口 + 手工触发
+# ============================================================
+def _resolve_ar_kpi(db: Session, entity_id: int, params: dict, rule, override_kpi_id):
+    """解析 ar_overdue 绑定的 KPI：规则行 → 显式 override（仅 dry-run）→ params.kpi_code。"""
+    if rule is not None and rule.kpi_id:
+        kpi = db.query(KPIDefinition).filter(
+            KPIDefinition.id == rule.kpi_id, KPIDefinition.entity_id == entity_id).first()
+        return (kpi.id if kpi else rule.kpi_id), (kpi.kpi_code if kpi else None), "alert_rules.kpi_id"
+    if override_kpi_id:
+        kpi = db.query(KPIDefinition).filter(
+            KPIDefinition.id == override_kpi_id, KPIDefinition.entity_id == entity_id).first()
+        if not kpi:
+            raise HTTPException(400, f"KPI 不存在或不属于当前账套: kpi_id={override_kpi_id}")
+        return kpi.id, kpi.kpi_code, "query.kpi_id(dry-run override)"
+    code = (params or {}).get("kpi_code")
+    if code:
+        kpi = db.query(KPIDefinition).filter(
+            KPIDefinition.kpi_code == code, KPIDefinition.entity_id == entity_id).first()
+        if kpi:
+            return kpi.id, kpi.kpi_code, f"params.kpi_code={code}"
+        return None, code, f"params.kpi_code={code}(库内不存在)"
+    return None, None, "unresolved"
+
+
+@router.post("/run-ar-overdue")
+def run_ar_overdue_check(
+    period: Optional[str] = Query(None, description="账龄基准期 YYYY-MM；缺省=表内 MAX(period)"),
+    dry_run: bool = Query(True, description="默认 true：只算不写。false 走既有写通道插入 kpi_alerts"),
+    kpi_id: Optional[int] = Query(None, description="dry-run 专用：临时绑定 KPI（正式以规则行 kpi_id 为准）"),
+    data: Optional[dict] = Body(None, description="dry-run 专用：{'params': {...}} 覆盖规则参数（证明参数化，不碰 DB）"),
+    db: Session = Depends(get_db),
+    entity_id: int = Depends(get_entity_id),
+):
+    """应收账龄逾期预警运行（T1）。
+
+    参数优先级：dry-run body 覆盖 > `alert_rules` 行（rule_type='ar_overdue'）> 无规则 ⇒ 200 skipped。
+    阈值/责任人/排除名单全部来自参数；`assignee` 缺失 ⇒ 400（fail-closed）。
+    `dry_run=false` 时禁止任何参数覆盖（防绕过 DB 配置），且只 INSERT 不 UPDATE/DELETE。
+    """
+    if is_excluded_entity(db, entity_id):
+        return {"message": f"账套 {entity_id} 为 demo 演示账套，按 R1 不参与预警生成",
+                "generated": 0, "skipped": True, "reason": "R1_demo_entity_excluded"}
+
+    override_params = None
+    if isinstance(data, dict):
+        override_params = data.get("params") if isinstance(data.get("params"), dict) else data.get("params_override")
+    if (override_params is not None or kpi_id is not None) and not dry_run:
+        raise HTTPException(
+            400, "dry_run=false 时禁止参数/KPI 覆盖：阈值与绑定必须来自 alert_rules 行（防绕过 DB 配置）")
+
+    rule = db.query(AlertRule).filter(
+        AlertRule.entity_id == entity_id,
+        AlertRule.rule_type == ar_overdue_alert.RULE_TYPE,
+        AlertRule.enabled == 1,
+    ).order_by(AlertRule.id).first()
+
+    params_source = "dry_run.body.params"
+    if override_params is not None:
+        params = override_params
+    elif rule is not None:
+        params = rule.params
+        if isinstance(params, str):
+            params = json.loads(params) if params else {}
+        params = params or {}
+        params_source = f"alert_rules#{rule.id}"
+    else:
+        return {
+            "dry_run": dry_run, "entity_id": entity_id, "period": period,
+            "skipped": True, "skipped_reason": "no_ar_overdue_rule",
+            "hint": "无 rule_type='ar_overdue' 规则行；阶段B 用 backend/scripts/seed_ar_overdue_rule.py 播种，"
+                    "或 dry-run 时在 body 传 {'params': {...}} 演练",
+        }
+
+    err = ar_overdue_alert.validate_params(params)
+    if err:
+        raise HTTPException(400, err)
+    try:
+        ar_overdue_alert.ensure_assignee_configured(params)
+    except ar_overdue_alert.AssigneeMissing as e:
+        raise HTTPException(400, str(e))
+
+    resolved_kpi_id, kpi_code, binding = _resolve_ar_kpi(db, entity_id, params, rule, kpi_id)
+
+    try:
+        result = ar_overdue_alert.compute(db, entity_id, period, params)
+    except ar_overdue_alert.ParamsInvalid as e:
+        raise HTTPException(400, str(e))
+
+    hits = ar_overdue_alert.hits(result)
+    blockers = []
+    if resolved_kpi_id is None:
+        blockers.append("kpi_id_unresolved:%s" % (kpi_code or "-"))
+
+    payload = {
+        "dry_run": dry_run,
+        "entity_id": entity_id,
+        "period": result.get("period"),
+        "caliber": result.get("caliber"),
+        "params_source": params_source,
+        "kpi_id": resolved_kpi_id,
+        "kpi_code": kpi_code,
+        "kpi_binding": binding,
+        "skipped_reason": result.get("skipped_reason"),
+        "metrics": result.get("metrics"),
+        "aggregates": result.get("aggregates"),
+        "per_customer": result.get("per_customer"),
+        "skipped": result.get("skipped"),
+        "hit_count": len(hits),
+        "notes": result.get("notes"),
+    }
+    if result.get("skipped_reason"):
+        payload.update({"would_insert": 0, "write_blockers": blockers})
+        return payload
+
+    if dry_run:
+        payload.update({"would_insert": 0 if blockers else ar_overdue_alert.count_would_insert(
+            db, entity_id, resolved_kpi_id, result), "write_blockers": blockers})
+        return payload
+
+    if blockers:
+        raise HTTPException(400, "无法落库：" + "; ".join(blockers))
+    written = ar_overdue_alert.write_alerts(db, entity_id, resolved_kpi_id, result)
+    log_op(db, "run_ar_overdue_alert", target_type="alert", target_id=resolved_kpi_id,
+           detail={"period": result.get("period"), "caliber": result.get("caliber"),
+                   "hit": len(hits), **written},
+           entity_id=entity_id)
+    db.commit()
+    payload.update({"would_insert": written["inserted"], "write_blockers": [], **written,
+                    "generated": written["inserted"]})
+    return payload

backend/app/api/ar_aging.py

@@ -0,0 +1,467 @@
+"""应收账龄看板 API（T2 · 2026-10-20）— prefix `/api/cma/ar-aging`
+
+数据源：`hanke_ar_aging`（逐户·逐期·三口径分列，财务口径见 `app.utils.ar_caliber`）。
+
+纪律（全部硬约束）：
+1. **每条查询都带 `entity_id` 过滤**（防两企业混报）；本模块只读，无 ORM 写模型/无迁移。
+2. 三口径**禁合计**：`caliber=all` 或未知字面值 → 400；默认口径 = `借方应收户口径`。
+3. 缺期显「该期无数据」：`has_data=false` 且金额字段为 `null`（**禁补 0**）。
+4. `opening_amt`（期初）**只展示**，单独返回并标 `non_aging_base=true`。
+5. 表不存在/无权限 ⇒ `available=false` + 原因，**HTTP 200 不 500**。
+"""
+from __future__ import annotations
+
+import csv
+import io
+import logging
+from datetime import datetime
+from decimal import Decimal, ROUND_HALF_UP
+from typing import Any, Dict, List, Optional
+
+from fastapi import APIRouter, Depends, HTTPException, Query
+from sqlalchemy import text
+from sqlalchemy.orm import Session
+
+from app.database import get_db
+from app.deps import get_entity_id
+from app.auth_middleware import require_role
+from app.utils import ar_caliber
+
+logger = logging.getLogger("ar_aging")
+
+router = APIRouter(
+    prefix="/api/cma/ar-aging",
+    tags=["应收账龄看板"],
+    dependencies=[Depends(require_role("ceo", "finance", "business"))],
+)
+
+_CENT = Decimal("0.01")
+_TABLE = ar_caliber.AR_AGING_TABLE
+_BUCKET_COLS = ar_caliber.BUCKET_COLUMNS
+_SORTABLE = {
+    "unit_code", "unit_name", "total_amt", "overdue_amt", "opening_amt", "oldest_days",
+    *_BUCKET_COLS,
+}
+
+
+# ============================================================
+# 公共helper
+# ============================================================
+def _resolve_caliber(caliber: Optional[str]) -> str:
+    try:
+        return ar_caliber.validate_caliber(caliber, allow_none=True)
+    except ar_caliber.CaliberError as e:
+        raise HTTPException(400, str(e))
+
+
+def _dec(v: Any) -> Decimal:
+    return Decimal("0") if v is None else Decimal(str(v))
+
+
+def _money(v: Any) -> Optional[float]:
+    if v is None:
+        return None
+    return float(_dec(v).quantize(_CENT, rounding=ROUND_HALF_UP))
+
+
+def _money_str(v: Any) -> str:
+    return "{:.2f}".format(_dec(v).quantize(_CENT, rounding=ROUND_HALF_UP))
+
+
+def _available(db: Session) -> tuple[bool, Optional[str]]:
+    try:
+        db.execute(text("SELECT 1 FROM %s LIMIT 1" % _TABLE)).first()
+        return True, None
+    except Exception as e:  # noqa: BLE001 — 表缺失/无权限一律降级
+        db.rollback()
+        logger.warning("hanke_ar_aging 不可用: %s", e)
+        return False, "表 %s 不可用（不存在或无 SELECT 权限）: %s" % (_TABLE, str(e).split("\n")[0][:180])
+
+
+def _periods(db: Session, entity_id: int) -> List[str]:
+    rows = db.execute(
+        text("SELECT DISTINCT period FROM %s WHERE entity_id = :e ORDER BY period" % _TABLE),
+        {"e": entity_id},
+    ).all()
+    return [r[0] for r in rows]
+
+
+def _month_add(period: str, delta: int) -> str:
+    y, m = int(period[:4]), int(period[5:7])
+    idx = y * 12 + (m - 1) + delta
+    return "%04d-%02d" % (idx // 12, idx % 12 + 1)
+
+
+def _month_range(start: str, end: str) -> List[str]:
+    if start > end:
+        return []
+    out, cur = [], start
+    while cur <= end:
+        out.append(cur)
+        cur = _month_add(cur, 1)
+    return out
+
+
+def _rows(db: Session, entity_id: int, caliber: str, period: str) -> List[Dict[str, Any]]:
+    sql = (
+        "SELECT id, period, caliber, unit_code, unit_name, opening_amt, total_amt, %s "
+        "FROM %s WHERE entity_id = :e AND caliber = :c AND period = :p "
+        "ORDER BY total_amt DESC, unit_code ASC"
+        % (", ".join(_BUCKET_COLS), _TABLE)
+    )
+    return [dict(r) for r in db.execute(
+        text(sql), {"e": entity_id, "c": caliber, "p": period}).mappings().all()]
+
+
+def _overdue_amt(row: Dict[str, Any]) -> Decimal:
+    """逾期额（>30 天）= total_amt − b_0_30；负档位不 ABS/不 COALESCE。"""
+    return _dec(row.get("total_amt")) - _dec(row.get("b_0_30"))
+
+
+def _oldest_days(row: Dict[str, Any]) -> int:
+    days = 0
+    for col, lower in ar_caliber.OVERDUE_BUCKETS:
+        if _dec(row.get(col)) != 0:
+            days = max(days, lower)
+    return days
+
+
+def _thresholds(db: Session, entity_id: int) -> Dict[str, Any]:
+    """看板阈值标注来源：`alert_rules(rule_type='ar_overdue')` 行（无规则 ⇒ 全 null，前端显「阈值未配置」）。
+
+    只读、只取展示用；判定仍以 `app.services.ar_overdue_alert` 单源为准。
+    """
+    from app.api.alert_rules import AlertRule
+    rule = db.query(AlertRule).filter(
+        AlertRule.entity_id == entity_id,
+        AlertRule.rule_type == "ar_overdue",
+        AlertRule.enabled == 1,
+    ).order_by(AlertRule.id).first()
+    if not rule:
+        return {"source": None, "share_90plus_red_gt": None,
+                "share_90plus_yellow_min": None, "share_90plus_yellow_max": None}
+    params = rule.params or {}
+    red = params.get("red") or {}
+    yellow = params.get("yellow") or {}
+    return {
+        "source": "alert_rules#%s" % rule.id,
+        "share_90plus_red_gt": red.get("share_90plus_gt"),
+        "share_90plus_yellow_min": yellow.get("share_90plus_min"),
+        "share_90plus_yellow_max": yellow.get("share_90plus_max"),
+    }
+
+
+def _enrich(row: Dict[str, Any]) -> Dict[str, Any]:
+    total = _dec(row.get("total_amt"))
+    od = _overdue_amt(row)
+    o90 = sum((_dec(row.get(c)) for c in ar_caliber.NINETY_PLUS_COLUMNS), Decimal("0"))
+    out = {
+        "unit_code": row.get("unit_code"),
+        "unit_name": row.get("unit_name"),
+        "opening_amt": _money(row.get("opening_amt")),
+        "total_amt": _money(total),
+        "buckets": {col: _money(row.get(col)) for col in _BUCKET_COLS},
+        "overdue_amt": _money(od),
+        "overdue_90plus": _money(o90),
+        "oldest_days": _oldest_days(row),
+        "pct_of_unit": (round(float(od / total), 6) if total > 0 else None),
+        "is_overdue": od > 0,
+    }
+    return out
+
+
+# ============================================================
+# GET /meta
+# ============================================================
+@router.get("/meta")
+def get_meta(db: Session = Depends(get_db), entity_id: int = Depends(get_entity_id)):
+    """看板元信息：可用性、口径全集、可用期间、最新期、入表时刻、来源件。"""
+    ok, reason = _available(db)
+    base = {
+        "entity_id": entity_id,
+        "table": _TABLE,
+        "calibers": list(ar_caliber.CALIBERS),
+        "default_caliber": ar_caliber.DEFAULT_CALIBER,
+        "available": ok,
+        "reason": reason,
+        "periods_available": [],
+        "latest_period": None,
+        "case_insensitive_note": "口径字面值逐字匹配；贷方口径为 `贷方户单列(净额)`",
+    }
+    if not ok:
+        return base
+
+    periods = _periods(db, entity_id)
+    base["periods_available"] = periods
+    base["latest_period"] = periods[-1] if periods else None
+    row = db.execute(
+        text("SELECT COUNT(*) n, MAX(loaded_at) loaded, MIN(period) pmin, MAX(period) pmax "
+             "FROM %s WHERE entity_id = :e" % _TABLE),
+        {"e": entity_id},
+    ).mappings().first()
+    loaded = row["loaded"]
+    base.update({
+        "row_count": int(row["n"] or 0),
+        "loaded_at": (loaded.isoformat(sep=" ") if hasattr(loaded, "isoformat") else (str(loaded) if loaded else None)),
+        "period_min": row["pmin"],
+        "period_max": row["pmax"],
+    })
+    src = db.execute(
+        text("SELECT DISTINCT source_file FROM %s WHERE entity_id = :e ORDER BY source_file" % _TABLE),
+        {"e": entity_id},
+    ).all()
+    base["source_files"] = [r[0] for r in src]
+    if not periods:
+        base["reason"] = "表内该账套无数据"
+    return base
+
+
+# ============================================================
+# GET /trend
+# ============================================================
+@router.get("/trend")
+def get_trend(
+    caliber: Optional[str] = Query(None),
+    start: Optional[str] = Query(None, description="起始期 YYYY-MM"),
+    end: Optional[str] = Query(None, description="结束期 YYYY-MM（缺省=表内最新期与当前月取大）"),
+    db: Session = Depends(get_db),
+    entity_id: int = Depends(get_entity_id),
+):
+    """逐期趋势（总额/户数/六档）；无数据期 has_data=false 且金额 null（禁补 0）。"""
+    cal = _resolve_caliber(caliber)
+    ok, reason = _available(db)
+    if not ok:
+        return {"available": False, "reason": reason, "entity_id": entity_id,
+                "caliber": cal, "points": []}
+
+    periods = _periods(db, entity_id)
+    if not periods:
+        return {"available": True, "entity_id": entity_id, "caliber": cal,
+                "periods_available": [], "points": [], "reason": "表内该账套无数据"}
+
+    now_period = datetime.now().strftime("%Y-%m")
+    start_p = start or periods[0]
+    end_p = end or max(periods[-1], now_period)
+    for p in (start_p, end_p):
+        if len(p) != 7 or p[4] != "-" or not (p[:4].isdigit() and p[5:].isdigit()):
+            raise HTTPException(400, f"期间格式非法（要求 YYYY-MM）: {p}")
+
+    points = []
+    for p in _month_range(start_p, end_p):
+        rows = _rows(db, entity_id, cal, p)
+        if not rows:
+            points.append({
+                "period": p, "has_data": False,
+                "total_amt": None, "customers": None, "overdue_amt": None,
+                "buckets": {col: None for col in _BUCKET_COLS},
+                "note": "该期无数据",
+            })
+            continue
+        total = sum((_dec(r.get("total_amt")) for r in rows), Decimal("0"))
+        points.append({
+            "period": p,
+            "has_data": True,
+            "total_amt": _money(total),
+            "customers": len(rows),
+            "overdue_amt": _money(sum((_overdue_amt(r) for r in rows), Decimal("0"))),
+            "buckets": {col: _money(sum((_dec(r.get(col)) for r in rows), Decimal("0")))
+                        for col in _BUCKET_COLS},
+            "note": None,
+        })
+    return {
+        "available": True, "entity_id": entity_id, "caliber": cal,
+        "start": start_p, "end": end_p,
+        "periods_available": periods,
+        "points": points,
+    }
+
+
+# ============================================================
+# GET /overview
+# ============================================================
+@router.get("/overview")
+def get_overview(
+    period: Optional[str] = Query(None),
+    caliber: Optional[str] = Query(None),
+    db: Session = Depends(get_db),
+    entity_id: int = Depends(get_entity_id),
+):
+    """期间概览：合计/户数/六档/逾期额/90天以上占比/环比；opening_amt 单列且标非账龄基数。"""
+    cal = _resolve_caliber(caliber)
+    th = _thresholds(db, entity_id)
+    ok, reason = _available(db)
+    if not ok:
+        return {"available": False, "reason": reason, "entity_id": entity_id,
+                "caliber": cal, "period": period, "has_data": False, "thresholds": th}
+
+    periods = _periods(db, entity_id)
+    if not periods:
+        return {"available": True, "entity_id": entity_id, "caliber": cal,
+                "period": period, "has_data": False, "reason": "表内该账套无数据",
+                "thresholds": th}
+    p = period or periods[-1]
+    rows = _rows(db, entity_id, cal, p)
+    if not rows:
+        return {
+            "available": True, "entity_id": entity_id, "caliber": cal, "period": p,
+            "has_data": False, "reason": "该期无数据",
+            "total_amt": None, "customers": None, "overdue_amt": None,
+            "overdue_90plus": None, "share_90plus": None,
+            "buckets": {col: None for col in _BUCKET_COLS},
+            "opening_amt": None, "opening_amt_non_aging_base": True,
+            "overdue_customers": None, "prev_period": None, "prev_overdue_amt": None,
+            "overdue_amt_mom_pct": None, "overdue_customers_mom_pct": None,
+            "thresholds": th,
+        }
+
+    total = sum((_dec(r.get("total_amt")) for r in rows), Decimal("0"))
+    buckets = {col: sum((_dec(r.get(col)) for r in rows), Decimal("0")) for col in _BUCKET_COLS}
+    od = sum((_overdue_amt(r) for r in rows), Decimal("0"))
+    o90 = sum(buckets[c] for c in ar_caliber.NINETY_PLUS_COLUMNS)
+    overdue_customers = sum(1 for r in rows if _overdue_amt(r) > 0)
+
+    idx = periods.index(p) if p in periods else -1
+    prev_period = periods[idx - 1] if idx > 0 else None
+    prev_rows = _rows(db, entity_id, cal, prev_period) if prev_period else []
+    prev_od = sum((_overdue_amt(r) for r in prev_rows), Decimal("0")) if prev_rows else None
+    prev_customers = (sum(1 for r in prev_rows if _overdue_amt(r) > 0) if prev_rows else None)
+
+    def _mom(cur: Optional[Decimal], prev: Optional[Decimal]) -> Optional[float]:
+        if cur is None or prev is None or prev == 0:
+            return None
+        return round(float((cur - prev) / prev), 6)
+
+    return {
+        "available": True,
+        "entity_id": entity_id,
+        "caliber": cal,
+        "period": p,
+        "has_data": True,
+        "total_amt": _money(total),
+        "customers": len(rows),
+        "buckets": {col: _money(v) for col, v in buckets.items()},
+        "buckets_sum": _money(sum(buckets.values())),
+        "buckets_sum_matches_total": _money(sum(buckets.values())) == _money(total),
+        "overdue_amt": _money(od),
+        "overdue_90plus": _money(o90),
+        "share_90plus": (round(float(o90 / total), 6) if total > 0 else None),
+        "overdue_customers": overdue_customers,
+        # 期初只展示：禁作账龄基数、禁进分子分母（财务定死）
+        "opening_amt": _money(sum((_dec(r.get("opening_amt")) for r in rows), Decimal("0"))),
+        "opening_amt_non_aging_base": True,
+        "prev_period": prev_period,
+        "prev_has_data": bool(prev_rows),
+        "prev_overdue_amt": _money(prev_od) if prev_od is not None else None,
+        "prev_overdue_customers": prev_customers,
+        "overdue_amt_mom_pct": _mom(od, prev_od),
+        "overdue_customers_mom_pct": (
+            _mom(Decimal(overdue_customers), Decimal(prev_customers))
+            if prev_customers is not None else None),
+        "mom_note": None if prev_rows else "上期无数据，不计算环比",
+        "thresholds": th,
+    }
+
+
+# ============================================================
+# GET /customers
+# ============================================================
+@router.get("/customers")
+def get_customers(
+    period: Optional[str] = Query(None),
+    caliber: Optional[str] = Query(None),
+    overdue_only: bool = Query(False),
+    sort: str = Query("overdue_amt"),
+    order: str = Query("desc"),
+    db: Session = Depends(get_db),
+    entity_id: int = Depends(get_entity_id),
+):
+    """逐户明细（六档 + 逾期额 + 最老档），服务端排序。"""
+    cal = _resolve_caliber(caliber)
+    if sort not in _SORTABLE:
+        raise HTTPException(400, f"sort 不支持: {sort}（可选 {sorted(_SORTABLE)}）")
+    if order not in ("asc", "desc"):
+        raise HTTPException(400, "order 仅接受 asc/desc")
+    ok, reason = _available(db)
+    if not ok:
+        return {"available": False, "reason": reason, "entity_id": entity_id,
+                "caliber": cal, "period": period, "has_data": False, "items": [], "total": 0}
+
+    periods = _periods(db, entity_id)
+    if not periods:
+        return {"available": True, "entity_id": entity_id, "caliber": cal,
+                "period": period, "has_data": False, "items": [], "total": 0}
+    p = period or periods[-1]
+    rows = _rows(db, entity_id, cal, p)
+    if not rows:
+        return {"available": True, "entity_id": entity_id, "caliber": cal, "period": p,
+                "has_data": False, "reason": "该期无数据", "items": [], "total": 0}
+
+    items = [_enrich(r) for r in rows]
+    if overdue_only:
+        items = [i for i in items if i["is_overdue"]]
+    items.sort(key=lambda i: (i.get(sort) is None, i.get(sort)), reverse=(order == "desc"))
+    return {
+        "available": True, "entity_id": entity_id, "caliber": cal, "period": p,
+        "has_data": True, "total": len(items), "sort": sort, "order": order,
+        "overdue_only": overdue_only, "items": items,
+    }
+
+
+# ============================================================
+# GET /customers/export
+# ============================================================
+@router.get("/customers/export")
+def export_customers(
+    period: Optional[str] = Query(None),
+    caliber: Optional[str] = Query(None),
+    overdue_only: bool = Query(False),
+    db: Session = Depends(get_db),
+    entity_id: int = Depends(get_entity_id),
+):
+    """逐户明细 CSV（UTF-8 BOM，Excel 直开）；首段写「截止期/口径/导出时刻」。"""
+    from urllib.parse import quote
+    from starlette.responses import Response
+
+    cal = _resolve_caliber(caliber)
+    ok, reason = _available(db)
+    if not ok:
+        raise HTTPException(503, reason or "数据表不可用")
+    periods = _periods(db, entity_id)
+    if not periods:
+        raise HTTPException(404, "表内该账套无数据")
+    p = period or periods[-1]
+    rows = _rows(db, entity_id, cal, p)
+    if not rows:
+        raise HTTPException(404, f"该期无数据: {p}")
+
+    items = [_enrich(r) for r in rows]
+    if overdue_only:
+        items = [i for i in items if i["is_overdue"]]
+    items.sort(key=lambda i: (i["overdue_amt"] is None, -(i["overdue_amt"] or 0)))
+
+    buf = io.StringIO()
+    w = csv.writer(buf)
+    w.writerow(["截止期", p, "口径", cal,
+                "导出时刻", datetime.now().strftime("%Y-%m-%d %H:%M:%S")])
+    w.writerow(["单位编码", "单位名称", "期初（非账龄基数）", "期末应收合计",
+                "0-30", "31-60", "61-90", "90-120", "120-150", "150+",
+                "逾期额(>30天)", "90天以上", "最老逾期档(天起)", "逾期占该户%", "是否逾期"])
+    for i in items:
+        w.writerow([
+            i["unit_code"], i["unit_name"], _money_str(i["opening_amt"]),
+            _money_str(i["total_amt"]),
+            *[_money_str(i["buckets"][c]) for c in _BUCKET_COLS],
+            _money_str(i["overdue_amt"]), _money_str(i["overdue_90plus"]),
+            i["oldest_days"],
+            ("" if i["pct_of_unit"] is None else "%.2f" % (i["pct_of_unit"] * 100)),
+            "是" if i["is_overdue"] else "否",
+        ])
+
+    content = "\ufeff" + buf.getvalue()
+    filename = "应收账龄_%s_%s.csv" % (p, cal)
+    return Response(
+        content=content.encode("utf-8"),
+        media_type="text/csv; charset=utf-8",
+        headers={"Content-Disposition": "attachment; filename=\"ar_aging_%s.csv\"; filename*=UTF-8''%s"
+                                         % (p, quote(filename))},
+    )

backend/app/api/charter.py

@@ -0,0 +1,227 @@
+"""战略基座 — 战略章程 API（/api/cma/charter）
+
+任务：`project-20261021-cma-strategy-charter-p0-fullstack`（P0 战略基座）
+------------------------------------------------------------------
+契约：
+  GET /api/cma/charter  读当前账套章程（无行 → 200 + filled=false 空态，**不造默认/示例值**）
+  PUT /api/cma/charter  写当前账套章程（upsert 单行；字段白名单 + 值域校验）
+
+隔离（三道）：
+  1. 账套一律取 `deps.get_entity_id` —— token 优先，显式传别的账套 → 403（平台统一语义）；
+  2. body 里出现 entity_id 且与解析结果不一致 → 403（防 Bot/脚本越权注入）；
+  3. 读写都带 `entity_id` 过滤条件（不靠唯一键兜底），跨账套读不到对方行。
+
+白名单（派单要求「白名单」）：
+  · 字段白名单 CHARTER_FIELDS —— 出现白名单外字段直接 400，不静默忽略；
+  · entity_id 不可改（由登录账套决定，手写一律拒绝）；
+  · 不可空：mission / vision / vision_year / vision_measurable / values_text / value_items；
+  · vision_year 须 4 位年份；value_items 须为 `[{name, meaning}]` 数组（每条可作裁决依据）。
+
+鉴权：GET 已登录即可（含 business 旁观）；PUT 限 ceo / finance（战略口径归经营层）。
+"""
+from datetime import datetime
+
+from fastapi import APIRouter, Depends, HTTPException
+from sqlalchemy.orm import Session
+
+from app.auth_middleware import require_auth, require_role
+from app.database import get_db
+from app.deps import get_entity_id
+from app.models import StrategyCharter, User
+from app.models.strategy_charter import (
+    MISSION_MAX_LEN,
+    VALUE_ITEM_MEANING_MAX_LEN,
+    VALUE_ITEM_NAME_MAX_LEN,
+    VALUE_ITEMS_MAX,
+    VALUES_TEXT_MAX_LEN,
+    VISION_MAX_LEN,
+    VISION_MEASURABLE_MAX_LEN,
+    VISION_YEAR_RE,
+)
+
+import re
+
+router = APIRouter(
+    prefix="/api/cma/charter",
+    tags=["战略基座"],
+    dependencies=[Depends(require_auth)],
+)
+
+# 可写字段白名单（entity_id / 审计字段不接受客户端输入）
+CHARTER_FIELDS = (
+    "mission",
+    "vision",
+    "vision_year",
+    "vision_measurable",
+    "values_text",
+    "value_items",
+)
+
+REQUIRED_FIELDS = CHARTER_FIELDS  # 六项均为必填（vision_year / vision_measurable 派单点名不可空）
+
+
+def _bad(msg: str) -> HTTPException:
+    return HTTPException(status_code=400, detail=msg)
+
+
+def _to_dict(row: StrategyCharter) -> dict:
+    return {
+        "id": row.id,
+        "entity_id": row.entity_id,
+        "mission": row.mission,
+        "vision": row.vision,
+        "vision_year": row.vision_year,
+        "vision_measurable": row.vision_measurable,
+        "values_text": row.values_text,
+        "value_items": row.value_items or [],
+        "updated_by": row.updated_by,
+        "updated_by_name": row.updated_by_name,
+        "created_at": row.created_at.isoformat(sep=" ", timespec="seconds") if row.created_at else None,
+        "updated_at": row.updated_at.isoformat(sep=" ", timespec="seconds") if row.updated_at else None,
+    }
+
+
+def _empty_state(entity_id: int) -> dict:
+    """空态：库里没有章程行。
+
+    明确返回 filled=false —— 前端据此显示「尚未填写」并给录入入口，
+    **不用任何默认/示例文案填充**（派单红线：真实数据非 Demo）。
+    """
+    return {"entity_id": entity_id, "filled": False, "data": None}
+
+
+@router.get("")
+def get_charter(
+    entity_id: int = Depends(get_entity_id),
+    db: Session = Depends(get_db),
+):
+    """读取当前账套的战略章程（无行 → filled=false）。"""
+    row = (
+        db.query(StrategyCharter)
+        .filter(StrategyCharter.entity_id == entity_id)
+        .first()
+    )
+    if not row:
+        return _empty_state(entity_id)
+    payload = _empty_state(entity_id)
+    payload["filled"] = True
+    payload["data"] = _to_dict(row)
+    return payload
+
+
+def _clean_text(raw, field: str, max_len: int) -> str:
+    if raw is None:
+        raise _bad(f"缺少必填字段 {field}")
+    if not isinstance(raw, str):
+        raise _bad(f"字段 {field} 必须是字符串")
+    val = raw.strip()
+    if not val:
+        raise _bad(f"字段 {field} 不能为空")
+    if len(val) > max_len:
+        raise _bad(f"字段 {field} 超长（最多 {max_len} 字符，当前 {len(val)}）")
+    return val
+
+
+def _clean_value_items(raw):
+    """价值观条目：[{name, meaning}]，每条须可作决策冲突时的裁决依据。"""
+    if raw is None:
+        raise _bad("缺少必填字段 value_items（价值观条目数组）")
+    if not isinstance(raw, list):
+        raise _bad("value_items 必须是数组 [{name, meaning}]")
+    if not raw:
+        raise _bad("value_items 至少需要 1 条（价值观须可作裁决依据）")
+    if len(raw) > VALUE_ITEMS_MAX:
+        raise _bad(f"value_items 最多 {VALUE_ITEMS_MAX} 条")
+    items = []
+    for idx, item in enumerate(raw, start=1):
+        if not isinstance(item, dict):
+            raise _bad(f"value_items 第 {idx} 条必须是对象 {{name, meaning}}")
+        extra = sorted(k for k in item.keys() if k not in ("name", "meaning"))
+        if extra:
+            raise _bad(f"value_items 第 {idx} 条含非白名单键 {extra}；只允许 name / meaning")
+        name = _clean_text(item.get("name"), f"value_items[{idx}].name", VALUE_ITEM_NAME_MAX_LEN)
+        meaning = _clean_text(item.get("meaning"), f"value_items[{idx}].meaning",
+                              VALUE_ITEM_MEANING_MAX_LEN)
+        items.append({"name": name, "meaning": meaning})
+    return items
+
+
+@router.put("")
+def put_charter(
+    body: dict,
+    entity_id: int = Depends(get_entity_id),
+    db: Session = Depends(get_db),
+    user: User = Depends(require_role("ceo", "finance")),
+):
+    """写入/更新当前账套的战略章程（单行 upsert，字段白名单 + 值域校验）。"""
+    if not isinstance(body, dict) or not body:
+        raise _bad("请求体不能为空")
+
+    # ① 越权账套参数：显式 entity_id 与 token 账套不一致 → 403（不静默覆盖）
+    raw_entity = body.get("entity_id")
+    if raw_entity is not None:
+        try:
+            claimed = int(raw_entity)
+        except (TypeError, ValueError):
+            raise _bad(f"entity_id 无效: {raw_entity!r}")
+        if claimed != entity_id:
+            raise HTTPException(
+                status_code=403,
+                detail=f"无权写入企业 entity_id={claimed}（当前账套: {entity_id}）",
+            )
+
+    # ② 字段白名单：白名单外字段一律拒绝（含 entity_id / id / updated_by 等审计字段）
+    unknown = sorted(k for k in body.keys() if k not in CHARTER_FIELDS)
+    if unknown:
+        raise _bad(
+            f"非白名单字段 {unknown}；允许字段 {list(CHARTER_FIELDS)}"
+            f"（entity_id 由登录账套决定，不可手写）"
+        )
+
+    # ③ 必填校验（vision_year / vision_measurable 不可空）
+    missing = [f for f in REQUIRED_FIELDS if f not in body or body.get(f) is None]
+    if missing:
+        raise _bad(f"缺少必填字段 {missing}；vision_year 与 vision_measurable 不可空")
+
+    mission = _clean_text(body.get("mission"), "mission", MISSION_MAX_LEN)
+    vision = _clean_text(body.get("vision"), "vision", VISION_MAX_LEN)
+    values_text = _clean_text(body.get("values_text"), "values_text", VALUES_TEXT_MAX_LEN)
+
+    # ④ vision_year：4 位年份（接受 int 2030 或 "2030"）
+    raw_year = body.get("vision_year")
+    if isinstance(raw_year, bool) or not isinstance(raw_year, (int, str)):
+        raise _bad("vision_year 必须是年份（如 2030）")
+    year = str(raw_year).strip()
+    if not re.match(VISION_YEAR_RE, year):
+        raise _bad(f"vision_year 须为 4 位年份（如 2030），收到 {raw_year!r}")
+
+    measurable = _clean_text(body.get("vision_measurable"), "vision_measurable",
+                             VISION_MEASURABLE_MAX_LEN)
+    value_items = _clean_value_items(body.get("value_items"))
+
+    row = (
+        db.query(StrategyCharter)
+        .filter(StrategyCharter.entity_id == entity_id)
+        .first()
+    )
+    if row is None:
+        row = StrategyCharter(entity_id=entity_id)
+        db.add(row)
+
+    row.mission = mission
+    row.vision = vision
+    row.vision_year = year
+    row.vision_measurable = measurable
+    row.values_text = values_text
+    row.value_items = value_items
+    row.updated_by = getattr(user, "id", None)
+    row.updated_by_name = getattr(user, "name", None)
+
+    db.commit()
+    db.refresh(row)
+
+    payload = _empty_state(entity_id)
+    payload["filled"] = True
+    payload["data"] = _to_dict(row)
+    payload["message"] = "战略章程已保存"
+    return payload

backend/app/api/lead.py

@@ -1,16 +1,17 @@
 """AI获客 - 客户咨询接口"""
 from fastapi import APIRouter, HTTPException
 from pydantic import BaseModel
-import json, os, datetime, subprocess, urllib.request
+import json, os, datetime, urllib.request
 
 router = APIRouter(prefix="/api/cma/lead", tags=["AI获客"])
 
-# ai-cs 实盘分流脚本（跨 profile 调用）
-# 必须以 ai-cs 身份 + 其 HERMES_HOME/HOME 运行：脚本内用 `hermes send` 复用该 profile 的企微凭据；
-# cma-backend.service 的 Environment 为空，不显式传入会解析到非 ai-cs 的 profile（root 默认 profile）。
-AI_CS_USER = "ai-cs"
-AI_CS_HOME = "/var/lib/wecom-ai-cs/profile"
-AI_CS_ROUTING_SCRIPT = AI_CS_HOME + "/scripts/routing-live.py"
+# 客户咨询通知唯一通道：系统总线中继 8660 → 任总 DM。
+# 2026-10-22 起取代旧的 ai-cs routing-live.py（该 profile/user 已下线，runuser 必然失败且异常被吞，
+# 导致线索静默丢失通知）。中继类别按契约取 fallback（无会话兜底），company 类被中继直接拒绝。
+BUS_RELAY_URL = "http://127.0.0.1:8660/bus/send"
+BUS_KEY_FILE = "/srv/hermes/shared/.secrets/bus_relay.key"
+BOSS_USERID = "200653946"  # 任富海
+NOTIFY_LOG = "/root/leads/notify.log"
 
 class LeadRequest(BaseModel):
     name: str
@@ -18,42 +19,54 @@ class LeadRequest(BaseModel):
     requirement: str = ""
     source: str = "sxbh.ltd"
 
-def notify_wecom(name, phone, requirement, source):
-    """推送到企微全员群"""
-    msg = {
-        "msgtype": "markdown",
-        "markdown": {
-            "content": f"## 🔔 新客户咨询\n**姓名**: {name}\n**电话**: {phone}\n**需求**: {requirement}\n**来源**: {source}\n**时间**: {datetime.datetime.now().strftime('%Y-%m-%d %H:%M')}"
-        }
-    }
+def _log_notify(ok, detail):
+    """通知结果落痕（旧实现 except: pass 静默吞异常，导致线索漏通知无人知晓）"""
     try:
-        data = json.dumps(msg).encode('utf-8')
-        req = urllib.request.Request(
-            "http://127.0.0.1:8800/send",
-            data=data,
-            headers={"Content-Type": "application/json"},
-            method="POST"
-        )
-        urllib.request.urlopen(req, timeout=5)
-    except:
-        pass  # 通知失败不影响主流程
-
-def notify_boss(name, phone, requirement):
-    """额外推给任总（经 ai-cs 实盘分流脚本；失败不影响主流程）"""
-    try:
-        subprocess.run(
-            ["runuser", "-u", AI_CS_USER, "--",
-             "env", f"HOME={AI_CS_HOME}", f"HERMES_HOME={AI_CS_HOME}",
-             f"USER={AI_CS_USER}", f"LOGNAME={AI_CS_USER}",
-             "/usr/bin/python3", AI_CS_ROUTING_SCRIPT,
-             requirement or "（未填需求）",
-             f"{name} {phone}",
-             "website:sxbh.ltd"],
-            timeout=30, capture_output=True,
-        )
+        os.makedirs(os.path.dirname(NOTIFY_LOG), exist_ok=True)
+        with open(NOTIFY_LOG, "a", encoding="utf-8") as f:
+            f.write("%s %s %s\n" % (datetime.datetime.now().isoformat(timespec="seconds"),
+                                    "OK" if ok else "FAIL", detail))
     except Exception:
         pass
 
+
+def notify_boss(name, phone, requirement):
+    """客户咨询通知：经系统总线中继 8660 推任总 DM。失败不阻断主流程，但必须留痕。"""
+    now = datetime.datetime.now().strftime("%Y-%m-%d %H:%M")
+    content = (
+        "## 新客户咨询（sxbh.ltd 官网）\n"
+        f"> 姓名：{name}\n"
+        f"> 电话：{phone}\n"
+        f"> 需求：{requirement or '（未填）'}\n"
+        f"> 时间：{now}\n\n"
+        "承诺 30 分钟内联系。"
+    )
+    try:
+        with open(BUS_KEY_FILE, encoding="utf-8") as f:
+            key = f.read().strip()
+        if not key:
+            raise RuntimeError("bus key empty")
+        payload = json.dumps({
+            "touser": BOSS_USERID,
+            "msgtype": "markdown",
+            "content": content,
+            "notice_class": "fallback",
+        }, ensure_ascii=False).encode("utf-8")
+        req = urllib.request.Request(
+            BUS_RELAY_URL, data=payload,
+            headers={"Content-Type": "application/json",
+                     "X-API-Key": key,
+                     "X-Notice-Class": "fallback"},
+            method="POST",
+        )
+        with urllib.request.urlopen(req, timeout=10) as resp:
+            body = resp.read().decode("utf-8", "ignore")[:300]
+        _log_notify(True, body)
+        return True
+    except Exception as e:
+        _log_notify(False, "%s: %s" % (type(e).__name__, e))
+        return False
+
 @router.post("")
 def create_lead(data: LeadRequest):
     """接收客户咨询并保存"""
@@ -66,9 +79,8 @@ def create_lead(data: LeadRequest):
     with open(f"{log_dir}/leads.json", "a") as f:
         f.write(json.dumps(record, ensure_ascii=False) + "\n")
     
-    # 异步推送企微通知
-    notify_wecom(data.name, data.phone, data.requirement, data.source)
-    # 加性：额外推给任总 DM（口径与 ai-cs 客服分流一致）
+    # 客户咨询通知唯一通道：推任总 DM（经 ai-cs 分流）；不再推 8800 全员群
+    # （2026-10-22 裁定：8800 仅限内容类群发，客户线索含个人信息，不得进全员群）
     notify_boss(data.name, data.phone, data.requirement)
     
     return {"success": True, "message": "咨询已提交，我们将在30分钟内联系您"}

backend/app/api/maps.py

@@ -7,12 +7,13 @@ from datetime import date, datetime, timedelta
 from app.database import get_db
 from app.deps import get_entity_id
 from app.auth_middleware import require_auth, require_role
-from app.models import StrategicMap, OperationLog, MapObjective, Objective, KR, KPIDefinition, ReviewRecord, User
+from app.models import StrategicMap, OperationLog, MapObjective, Objective, KR, KPIDefinition, ReviewRecord, User, MapConnection
 from app.utils.kr_name import sanitize_kr_name
 from app.utils.kpi_value_scope import actual_only_clause
 import copy
 import json
 import logging
+import re
 
 logger = logging.getLogger(__name__)
 
@@ -507,7 +508,165 @@ def batch_delete_maps(data: dict, db: Session = Depends(get_db), entity_id: int
     return {"message": f"已删除 {deleted} 个地图", "deleted": deleted}
 
 
-# ── 连线管理 ─────────────────────────────────
+# ── 连线管理（P1 四层因果结构化：map_connections 表 ⇄ canvas_data 双写）──
+#
+# 2026-10-21 cma-map-connections-table-p1：
+#   · 生产存量节点标识是 `层:中文目标名`（如 `learning:贡献率思维全员培训`），
+#     当前前端 MapCanvas 生成的是 `层-序号`（如 `learning-0`）⇒ 两种格式都必须能解析；
+#   · 表是新增读模型（归因/校验用），canvas_data 保持原样供前端渲染 ⇒ 双写期；
+#   · 表缺失时（迁移前/回滚后）所有写路径降级为只写 canvas_data，不阻断前端。
+
+LAYER_ORDER = ("learning", "process", "customer", "finance")
+_LAYER_INDEX = {k: i for i, k in enumerate(LAYER_ORDER)}
+_NODE_COLON = re.compile(r"^([a-z_]+):(.+)$")
+_NODE_INDEX = re.compile(r"^([a-z_]+)-(\d+)$")
+
+
+def _norm_obj_name(s) -> str:
+    return re.sub(r"\s+", " ", str(s or "").strip())
+
+
+def parse_node_ref(node):
+    """节点标识解析（双格式兼容）→ dict(layer, kind, value, error)
+
+    `learning:贡献率思维全员培训` → layer=learning, kind=name, value=中文名
+    `learning-0`                → layer=learning, kind=index, value=0
+    """
+    if not isinstance(node, str) or not node.strip():
+        return {"layer": None, "kind": None, "value": None, "error": "node_empty"}
+    s = node.strip()
+    m = _NODE_COLON.match(s)
+    if m:
+        layer, name = m.group(1), m.group(2).strip()
+        if layer not in _LAYER_INDEX:
+            return {"layer": None, "kind": None, "value": None, "error": f"unknown_layer:{layer}"}
+        if not name:
+            return {"layer": layer, "kind": None, "value": None, "error": "node_name_empty"}
+        return {"layer": layer, "kind": "name", "value": name, "error": None}
+    m = _NODE_INDEX.match(s)
+    if m:
+        layer, idx = m.group(1), int(m.group(2))
+        if layer not in _LAYER_INDEX:
+            return {"layer": None, "kind": None, "value": None, "error": f"unknown_layer:{layer}"}
+        return {"layer": layer, "kind": "index", "value": idx, "error": None}
+    return {"layer": None, "kind": None, "value": None, "error": "node_format_unrecognized"}
+
+
+def _objective_index(db: Session, m: StrategicMap):
+    """(名字索引, 层内顺序索引) —— 名字索引键 = (dimension_key, 归一化名称)。"""
+    rows = db.query(MapObjective).filter(MapObjective.map_id == m.id).order_by(
+        MapObjective.sort_order, MapObjective.id).all()
+    by_name, by_layer = {}, {}
+    for o in rows:
+        by_name[(o.dimension_key, _norm_obj_name(o.name))] = o.id
+        by_layer.setdefault(o.dimension_key, []).append(o)
+    return by_name, by_layer
+
+
+def resolve_node(db: Session, m: StrategicMap, node):
+    """→ (layer, objective_id|None, note|None)
+
+    `层-序号` 的序号语义 = 前端 MapCanvas 使用的 `dimensions[层].objectives` 数组下标；
+    dimensions 缺失时退化到 map_objectives.sort_order / 层内顺序。
+    """
+    ref = parse_node_ref(node)
+    layer = ref["layer"]
+    if ref["error"]:
+        return layer, None, ref["error"]
+    by_name, by_layer = _objective_index(db, m)
+    if ref["kind"] == "name":
+        want = _norm_obj_name(ref["value"])
+        hit = by_name.get((layer, want))
+        if hit:
+            return layer, hit, None
+        return layer, None, "objective_unresolved"
+    idx = ref["value"]
+    dims = _dims_list(m.dimensions)
+    for d in dims:
+        if d.get("key") == layer:
+            objs = d.get("objectives") or []
+            if 0 <= idx < len(objs):
+                name = _norm_obj_name((objs[idx] or {}).get("name"))
+                hit = by_name.get((layer, name))
+                return layer, hit, (None if hit else "objective_unresolved")
+            return layer, None, "index_out_of_range"
+    cand = by_layer.get(layer) or []
+    for o in cand:
+        if (o.sort_order or 0) == idx:
+            return layer, o.id, "resolved_by_sort_order"
+    if 0 <= idx < len(cand):
+        return layer, cand[idx].id, "resolved_by_position"
+    return layer, None, "index_out_of_range"
+
+
+def _row_payload_from_canvas(db: Session, m: StrategicMap, c: dict, sort_order: int) -> dict:
+    frm, to = c.get("from", ""), c.get("to", "")
+    fl, f_obj, f_note = resolve_node(db, m, frm)
+    tl, t_obj, t_note = resolve_node(db, m, to)
+    return {
+        "entity_id": m.entity_id if m.entity_id is not None else 1,
+        "map_id": m.id, "from_node": frm, "to_node": to,
+        "from_layer": fl, "to_layer": tl,
+        "from_objective_id": f_obj, "to_objective_id": t_obj,
+        "effect": c.get("effect") or "positive",
+        "label": (c.get("label") or None),
+        "sort_order": sort_order,
+    }
+
+
+def _table_ready(db: Session) -> bool:
+    """新表是否可用（迁移前/回滚后为 False）——用 SAVEPOINT 隔离探测，避免污染事务。"""
+    try:
+        with db.begin_nested():
+            db.query(MapConnection).limit(1).all()
+        return True
+    except Exception:
+        return False
+
+
+def _table_rows(db: Session, m: StrategicMap):
+    # 账套隔离已在 map 查询处完成；这里只按 map_id 取行，避免历史 map.entity_id 为空时
+    # 冗余列与 map 不一致导致「表里有行却读不到」的假回落。
+    rows = db.query(MapConnection).filter(
+        MapConnection.map_id == m.id,
+    ).order_by(MapConnection.sort_order, MapConnection.id).all()
+    return rows
+
+
+def _table_upsert(db: Session, m: StrategicMap, payload: dict) -> bool:
+    """双写：失败只告警、不阻断 canvas_data 主路径（兼容期容错）。"""
+    try:
+        with db.begin_nested():
+            existing = db.query(MapConnection).filter(
+                MapConnection.map_id == payload["map_id"],
+                MapConnection.from_node == payload["from_node"],
+                MapConnection.to_node == payload["to_node"],
+            ).first()
+            if existing:
+                for k, v in payload.items():
+                    setattr(existing, k, v)
+            else:
+                db.add(MapConnection(**payload))
+            db.flush()
+        return True
+    except Exception as exc:  # 表缺失/授权不足/约束冲突 → 主路径继续
+        logger.warning("map_connections 双写失败（canvas_data 已保留）: %s: %s", type(exc).__name__, exc)
+        return False
+
+
+def _table_delete(db: Session, m: StrategicMap, from_node: str, to_node: str) -> bool:
+    try:
+        with db.begin_nested():
+            db.query(MapConnection).filter(
+                MapConnection.map_id == m.id,
+                MapConnection.from_node == from_node,
+                MapConnection.to_node == to_node,
+            ).delete(synchronize_session=False)
+        return True
+    except Exception as exc:
+        logger.warning("map_connections 删除同步失败: %s: %s", type(exc).__name__, exc)
+        return False
+
 
 def _get_connections(m: StrategicMap) -> list:
     if not m.canvas_data:
@@ -522,9 +681,228 @@ def _get_connections(m: StrategicMap) -> list:
     return m.canvas_data["connections"]
 
 
+def _objective_key_index(db: Session, m: StrategicMap):
+    """连线解析锚点：MapObjective.id → 画布渲染键 `${dim}-${idx}`（以当前 dimensions 顺序为准）
+
+    返回 (id2key, name2key, dim_len)：
+      · id2key  : 本表行 id → `${dim}-${idx}`（map_connections.from/to_objective_id 的锚点）
+      · name2key: (dim, 归一化名称) → `${dim}-${idx}`（历史 `层:中文名` 的兜底，改名即失效）
+      · dim_len : {dim: len(objectives)}（`层-序号` 越界判定）
+
+    只按名称把 dimensions 节点对回 map_objectives 行，**不读** dims JSON 里的 `_objective_id`
+    —— 那是 objectives 表（实体级目标）的 id 空间，与 map_objectives 行 id 不是一回事。
+    """
+    dims = _dims_list(m.dimensions)
+    rows = db.query(MapObjective).filter(MapObjective.map_id == m.id).order_by(
+        MapObjective.sort_order, MapObjective.id).all()
+    by_name = {}
+    for o in rows:
+        by_name[(o.dimension_key, _norm_obj_name(o.name))] = o.id
+    id2key, name2key, dim_len = {}, {}, {}
+    for d in dims:
+        dk = d.get("key") or ""
+        if not dk:
+            continue
+        objs = d.get("objectives") or []
+        dim_len[dk] = len(objs)
+        for idx, o in enumerate(objs):
+            if not isinstance(o, dict):
+                continue
+            key = f"{dk}-{idx}"
+            nm = _norm_obj_name(o.get("name"))
+            if nm:
+                name2key[(dk, nm)] = key
+            oid = by_name.get((dk, nm))
+            if oid is not None:
+                id2key[oid] = key
+    return id2key, name2key, dim_len
+
+
+def _key_for_node(node, id2key, name2key, dim_len, objective_id=None):
+    """单端节点解析 → (key, source, reason)
+
+    顺序：objective_id 锚点 → `层:中文名` → `层-序号`。
+    改名后 `objective_id` 仍指向同一节点 ⇒ key 不变、线不断；锚点失效（目标已删/历史脏 id/
+    跨 map）才回退原名字串，仍解析不到才返回 None（调用方计入 unresolved）。
+    """
+    if objective_id:
+        hit = id2key.get(objective_id)
+        if hit:
+            return hit, "objective_id", None
+    ref = parse_node_ref(node)
+    if ref["error"]:
+        return None, None, ref["error"]
+    layer = ref["layer"]
+    stale = "stale_objective_id" if objective_id else None
+    if ref["kind"] == "name":
+        want = _norm_obj_name(ref["value"])
+        hit = name2key.get((layer, want))
+        if hit:
+            return hit, "name", stale
+        return None, None, (stale or "objective_unresolved")
+    idx = ref["value"]
+    if 0 <= idx < dim_len.get(layer, 0):
+        return f"{layer}-{idx}", "index", stale
+    return None, None, "index_out_of_range"
+
+
+def _attach_keys(db: Session, m: StrategicMap, items: list):
+    """给连线条目附 `from_key/to_key`（渲染键）+ 未解析清单。
+
+    条目原地补字段：`{side}_key` / `{side}_key_source` / `{side}_key_reason`，
+    键与中文名解耦的唯一出口（前端按 key 画线，缺失回落原值）。
+    """
+    id2key, name2key, dim_len = _objective_key_index(db, m)
+    unresolved = []
+    for it in items:
+        for side in ("from", "to"):
+            key, source, reason = _key_for_node(
+                it.get(side), id2key, name2key, dim_len, it.get(f"{side}_objective_id"))
+            it[f"{side}_key"] = key
+            it[f"{side}_key_source"] = source
+            it[f"{side}_key_reason"] = reason
+            if key is None:
+                unresolved.append({
+                    "from": it.get("from"), "to": it.get("to"),
+                    "side": side, "node": it.get(side),
+                    "reason": reason, "objective_id": it.get(f"{side}_objective_id"),
+                })
+    return items, unresolved
+
+
+@router.get("/{map_id}/connections")
+def list_connections(map_id: int, db: Session = Depends(get_db), entity_id: int = Depends(get_entity_id)):
+    """读连线：表优先，表空回落 canvas_data（兼容期）。
+
+    P3：每条连线附 `from_key/to_key` = `${dim}-${idx}`（由 objective_id 在当前 dimensions
+    树反查）+ 顶层 `unresolved` 计数与清单。前端优先按 key 画线 ⇒ 目标改名不断线。
+    """
+    m = db.query(StrategicMap).filter(StrategicMap.id == map_id, StrategicMap.entity_id == entity_id).first()
+    if not m:
+        raise HTTPException(404, "战略地图不存在")
+
+    if _table_ready(db):
+        rows = _table_rows(db, m)
+        if rows:
+            items = [{
+                "from": r.from_node, "to": r.to_node, "style": "solid",
+                "effect": r.effect or "positive", "label": r.label or "",
+                "from_layer": r.from_layer, "to_layer": r.to_layer,
+                "from_objective_id": r.from_objective_id, "to_objective_id": r.to_objective_id,
+            } for r in rows]
+            items, unresolved = _attach_keys(db, m, items)
+            return {
+                "source": "table", "total": len(items),
+                "unresolved": len(unresolved), "unresolved_items": unresolved,
+                "connections": items,
+            }
+
+    conns = _get_connections(m)
+    items = [{
+        "from": c.get("from"), "to": c.get("to"),
+        "style": c.get("style") or "solid",
+        "effect": c.get("effect") or "positive",
+        "label": c.get("label") or "",
+        "from_objective_id": c.get("from_objective_id"),
+        "to_objective_id": c.get("to_objective_id"),
+    } for c in conns if isinstance(c, dict)]
+    items, unresolved = _attach_keys(db, m, items)
+    return {
+        "source": "canvas_data", "total": len(items),
+        "unresolved": len(unresolved), "unresolved_items": unresolved,
+        "connections": items,
+    }
+
+
+@router.get("/{map_id}/connections/analysis")
+def analyze_connections(map_id: int, db: Session = Depends(get_db), entity_id: int = Depends(get_entity_id)):
+    """连线只读分析（归因/复盘用）：层统计 + 跨层分布 + 断链 + 反向告警（不拦截）。
+
+    读路径：表内有行 → 读表；否则回落 canvas_data（并对回落数据现算解析结果）。
+    """
+    m = db.query(StrategicMap).filter(StrategicMap.id == map_id, StrategicMap.entity_id == entity_id).first()
+    if not m:
+        raise HTTPException(404, "战略地图不存在")
+
+    source = "canvas_data"
+    items = []
+    table_rows = _table_rows(db, m) if _table_ready(db) else []
+    if table_rows:
+        source = "table"
+        for r in table_rows:
+            items.append({
+                "from": r.from_node, "to": r.to_node,
+                "from_layer": r.from_layer, "to_layer": r.to_layer,
+                "from_objective_id": r.from_objective_id, "to_objective_id": r.to_objective_id,
+                "effect": r.effect or "positive", "label": r.label,
+            })
+    else:
+        for i, c in enumerate(_get_connections(m)):
+            if not isinstance(c, dict):
+                continue
+            row = _row_payload_from_canvas(db, m, c, i)
+            items.append({
+                "from": row["from_node"], "to": row["to_node"],
+                "from_layer": row["from_layer"], "to_layer": row["to_layer"],
+                "from_objective_id": row["from_objective_id"],
+                "to_objective_id": row["to_objective_id"],
+                "effect": row["effect"], "label": row["label"],
+            })
+
+    layer_stats = {k: {"as_from": 0, "as_to": 0} for k in LAYER_ORDER}
+    cross_layer, reversed_edges, broken, unknown = {}, [], [], []
+    same_layer = 0
+    items, unresolved = _attach_keys(db, m, items)
+    for it in items:
+        fl, tl = it.get("from_layer"), it.get("to_layer")
+        if fl in layer_stats:
+            layer_stats[fl]["as_from"] += 1
+        if tl in layer_stats:
+            layer_stats[tl]["as_to"] += 1
+        if fl not in _LAYER_INDEX or tl not in _LAYER_INDEX:
+            unknown.append({"from": it["from"], "to": it["to"],
+                            "from_layer": fl, "to_layer": tl})
+        else:
+            pair = f"{fl}->{tl}"
+            cross_layer[pair] = cross_layer.get(pair, 0) + 1
+            if fl == tl:
+                same_layer += 1
+            elif _LAYER_INDEX[tl] < _LAYER_INDEX[fl]:
+                reversed_edges.append({"from": it["from"], "to": it["to"],
+                                       "from_layer": fl, "to_layer": tl})
+        if not it.get("from_key"):
+            broken.append({"from": it["from"], "to": it["to"], "side": "from",
+                           "node": it["from"],
+                           "reason": it.get("from_key_reason") or "objective_unresolved"})
+        if not it.get("to_key"):
+            broken.append({"from": it["from"], "to": it["to"], "side": "to",
+                           "node": it["to"],
+                           "reason": it.get("to_key_reason") or "objective_unresolved"})
+
+    warnings = []
+    if reversed_edges:
+        warnings.append(f"反向连线 {len(reversed_edges)} 条（与 learning→process→customer→finance 相反，"
+                        f"仅告警不拦截）")
+    if broken:
+        warnings.append(f"断链 {len(broken)} 处（节点未能解析到 map_objectives）")
+    if unknown:
+        warnings.append(f"层前缀无法识别 {len(unknown)} 条")
+
+    return {
+        "map_id": m.id, "entity_id": m.entity_id, "source": source,
+        "layer_order": list(LAYER_ORDER), "total": len(items),
+        "layer_stats": layer_stats, "cross_layer": cross_layer,
+        "same_layer_count": same_layer,
+        "reversed": reversed_edges, "broken": broken,
+        "unknown_layer": unknown, "warnings": warnings,
+        "unresolved": len(unresolved), "unresolved_items": unresolved,
+        "items": items,
+    }
+
+
 @router.post("/{map_id}/connections")
 def add_connection(map_id: int, data: dict, db: Session = Depends(get_db), entity_id: int = Depends(get_entity_id)):
-    """新增因果连线: {"from": "learning-0", "to": "process-0"}"""
+    """新增因果连线: {"from": "learning-0", "to": "process-0"}（表 + canvas_data 双写）"""
     m = db.query(StrategicMap).filter(StrategicMap.id == map_id, StrategicMap.entity_id == entity_id).first()
     if not m:
         raise HTTPException(404, "战略地图不存在")
@@ -539,10 +917,6 @@ def add_connection(map_id: int, data: dict, db: Session = Depends(get_db), entit
     if from_id == to_id:
         raise HTTPException(400, "不能自身连线")
 
-    # 校验: 维度不能相同 (但放开允许同层连线, 仅禁止自连)
-    from_dim = from_id.rsplit("-", 1)[0]
-    to_dim = to_id.rsplit("-", 1)[0]
-
     conns = _get_connections(m)
 
     # 校验: 不能重复
@@ -558,14 +932,21 @@ def add_connection(map_id: int, data: dict, db: Session = Depends(get_db), entit
         "label": data.get("label", ""),
     })
     m.canvas_data["connections"] = conns
+    from sqlalchemy.orm.attributes import flag_modified
+    flag_modified(m, "canvas_data")   # 同上：JSON 列需显式标脏才能落库
+
+    # 双写（表缺失/失败不阻断前端主路径）
+    payload = _row_payload_from_canvas(db, m, conns[-1], len(conns) - 1)
+    db_synced = _table_upsert(db, m, payload)
+
     db.commit()
     # updated_at: 供前端采纳为新基准（并发保护 base_updated_at），避免自伤 409
-    return {"connections": conns, "updated_at": _stamp(m.updated_at)}
+    return {"connections": conns, "updated_at": _stamp(m.updated_at), "db_synced": db_synced}
 
 
 @router.delete("/{map_id}/connections")
 def delete_connection_by_key(map_id: int, data: dict, db: Session = Depends(get_db), entity_id: int = Depends(get_entity_id)):
-    """根据 from/to 删除连线"""
+    """根据 from/to 删除连线（表 + canvas_data 同步删）"""
     m = db.query(StrategicMap).filter(StrategicMap.id == map_id, StrategicMap.entity_id == entity_id).first()
     if not m:
         raise HTTPException(404, "战略地图不存在")
@@ -580,9 +961,14 @@ def delete_connection_by_key(map_id: int, data: dict, db: Session = Depends(get_
         raise HTTPException(404, "连线不存在")
 
     m.canvas_data["connections"] = new_conns
+    # JSON 列就地/换列表赋值不会被 ORM 自动标记为 dirty ⇒ 必须显式 flag_modified，
+    # 否则接口返回 200 且响应列表已减 1，但 commit 不 flush，重读仍是旧值（删连线变哑弹）。
+    from sqlalchemy.orm.attributes import flag_modified
+    flag_modified(m, "canvas_data")
+    db_synced = _table_delete(db, m, from_id, to_id)
     db.commit()
     return {"connections": new_conns, "removed": {"from": from_id, "to": to_id},
-            "updated_at": _stamp(m.updated_at)}
+            "updated_at": _stamp(m.updated_at), "db_synced": db_synced}
 
 
 # ── 版本管理 ─────────────────────────────────
@@ -645,26 +1031,98 @@ def m_to_dict(m, db: Session = None):
 
 
 def _sync_map_objectives(m, db):
-    """保存时：将dimensions JSON中的目标同步到map_objectives表"""
-    db.query(MapObjective).filter(MapObjective.map_id == m.id).delete()
-    dims = m.dimensions
-    if isinstance(dims, str):
-        dims = json.loads(dims)
-    dims = dims or []
+    """保存时：将dimensions JSON中的目标同步到map_objectives表（就地 upsert）
+
+    2026-10-22 cma-map-canvas-conn-key-rename-stable（P3）：
+    原实现是 delete+insert —— MySQL 的 AUTO_INCREMENT 计数器不随 DELETE 归零，每次保存
+    都会给同一节点换一批新行 id，`map_connections.from/to_objective_id` 随即变成野指针
+    （读侧再按 id 反查就查不到），「用 objective_id 与中文名解耦」的前提被静默破坏。
+    改为就地 upsert：行 id 全程稳定，连线锚点才真的能扛住改名。
+
+    定位顺序：dims 的 `_mo_id`（读侧由 _merge_map_objectives 回写）→ (dimension_key, 归一化名称)。
+    **不做位置兜底**（口径 B，ASK-20261023-04 任总 2026-10-23 15:13「放行 4」）：
+    未命中名字的项一律视为新目标，未被占用的存量行一律视为已删除。理由见下方 ③。
+    """
+    dims = _dims_list(m.dimensions)
+    entity_id = m.entity_id if m.entity_id is not None else 1
+    rows = db.query(MapObjective).filter(MapObjective.map_id == m.id).order_by(
+        MapObjective.sort_order, MapObjective.id).all()
+    by_id = {r.id: r for r in rows}
+    used = set()
+
+    # 计划项：按 dims 顺序展开
+    planned = []
     for dim in dims:
-        for i, obj in enumerate(dim.get("objectives", [])):
-            mo = MapObjective(
-                map_id=m.id,
-                entity_id=m.entity_id if m.entity_id is not None else 1,  # 强制继承地图企业(OpenCode审查#11)
-                dimension_key=dim.get("key", ""),
-                name=obj.get("name", ""),
-                description=obj.get("description", ""),
-                icon=obj.get("icon", "target"),
-                kpis=obj.get("kpis", []),
-                sort_order=i,
-            )
-            db.add(mo)
+        dk = dim.get("key", "")
+        for i, obj in enumerate(dim.get("objectives") or []):
+            if isinstance(obj, dict):
+                planned.append((dk, i, obj))
+
+    assigned = {}
+
+    # ① 显式锚点：客户端回带的 `_mo_id`（读侧 _merge_map_objectives 写入）优先级最高
+    for idx, (dk, i, obj) in enumerate(planned):
+        try:
+            mo_id = int(obj.get("_mo_id")) if obj.get("_mo_id") is not None else None
+        except (TypeError, ValueError):
+            mo_id = None
+        if mo_id is not None and mo_id in by_id and mo_id not in used:
+            assigned[idx] = by_id[mo_id]
+            used.add(mo_id)
+
+    # ② 名称批：**先全表跑完名字匹配，再做位置兜底**。
+    #   逐项贪心（名→序位→空位）时，改名项在名字层落空后会立刻抢占同序位行，把
+    #   「同一次保存里改名 + 同层调序」的两个目标 id 串位 —— 线静默画到别的节点上，
+    #   比断线更坏（看不见）。批处理顺序把名字仍有效的项先钉住，未命中的项再按顺序配对。
+    for idx, (dk, i, obj) in enumerate(planned):
+        if idx in assigned:
+            continue
+        want = _norm_obj_name(obj.get("name"))
+        cand = [r for r in rows
+                if r.id not in used and r.dimension_key == dk
+                and _norm_obj_name(r.name) == want]
+        if cand:
+            cand.sort(key=lambda r: (r.sort_order or 0, r.id))   # 同名重复取最靠前，结果确定
+            assigned[idx] = cand[0]
+            used.add(cand[0].id)
+
+    # ③ （口径 B · ASK-20261023-04 任总 2026-10-23 15:13「放行 4」）**不做位置兜底**。
+    #   老通道（不带 `_mo_id`）下「删一个 + 加一个」与「改名」数据上不可区分：按位置配对
+    #   在其中一种情况下必然静默错绑（新项继承被删项的锚点，线正常画着却是错的）。
+    #   错绑比断线更坏 —— 断线在读侧 `unresolved` 里可见、可计数。故：
+    #   · 名字未命中的项 → 一律当新目标插入（新行 id）；
+    #   · 未被占用的存量行 → 一律当已删除删除；
+    #   · 连线锚点由此指向被删行 ⇒ 读侧记 `stale_objective_id` 并计入 unresolved。
+    #   UI 通道（客户端回带 `_mo_id`）改名仍走 ① 精确保住锚点，本口径不影响它。
+    #   逐项贪心（名→序位→空位）的串位反证见 tests/test_conn_key_rename_stability.py
+    #   的 test_negative_greedy_matching_would_crossbind。
+
+    created = 0
+    keep = []
+    for idx, (dk, i, obj) in enumerate(planned):
+        target = assigned.get(idx)
+        if target is None:
+            target = MapObjective(map_id=m.id, entity_id=entity_id, dimension_key=dk or "finance")
+            db.add(target)
+            created += 1
+        target.entity_id = entity_id
+        target.dimension_key = dk
+        target.name = (obj.get("name") or "")[:200]
+        target.description = obj.get("description") or ""
+        target.icon = obj.get("icon") or "target"
+        target.kpis = obj.get("kpis") or []
+        target.sort_order = i
+        keep.append(target)
+
+    db.flush()                       # 新增行拿到 id 后才有 id 可比
+    deleted = 0
+    for r in rows:
+        if r.id not in used:
+            db.delete(r)
+            deleted += 1
     db.commit()
+    return {"planned": len(planned), "created": created, "deleted": deleted,
+            "kept": len(used)}
 
 
 def _merge_map_objectives(m, db):
@@ -698,6 +1156,9 @@ def _merge_map_objectives(m, db):
                     "icon": o.icon or existing.get("icon", "target"),
                     "kpis": o.kpis or existing.get("kpis", []),
                     **{k: existing[k] for k in existing if k not in ("name", "description", "icon", "kpis")},
+                    # P3：回写本表行 id（连线锚点的落点）。只在读响应里出现，不落 dimensions JSON；
+                    # 客户端原样带回保存时，_sync_map_objectives 据此就地更新、保住行 id。
+                    "_mo_id": o.id,
                 })
             dim["objectives"] = merged
     m.dimensions = dims
@@ -1025,6 +1486,7 @@ def get_map_review(map_id: int, level: Optional[str] = None, db: Session = Depen
     total_ok = 0
     total_warn = 0
     total_err = 0
+    total_hold = 0  # C1 数据待核计数（此前漏初始化 → 命中 hold_reason 即 UnboundLocalError → review 500）
     total_obj_count = 0
     focus_items = []
 
@@ -1164,6 +1626,7 @@ def get_map_review(map_id: int, level: Optional[str] = None, db: Session = Depen
             "green": total_ok,
             "yellow": total_warn,
             "red": total_err,
+            "hold": total_hold,
             "health_score": round(total_ok / total_obj_count * 100, 1) if total_obj_count > 0 else 0,
         },
         "focus_items": focus_items,

backend/app/main.py

@@ -7,7 +7,7 @@ from fastapi.responses import JSONResponse
 from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
 from dotenv import load_dotenv
 from app.database import init_db, get_db
-from app.api import auth, kpis, kpi_governance, templates, maps, dashboard, data, alerts, ai_analysis, ai_suggestions, alert_rules, users, thresholds, notifications, permissions, action_plans, alignment, org, objectives, versions, budget, cost, predict, reports, security, knowledge, bot_bridge, bot_bridge_v2, bot_suggestions, lead, tenant, customer_dashboard, deviation_push, budget_generate, knowledge_articles, kpi_causality, data_quality, bi_reports, entities, bsc_layers, okr, okr_templates, subjects, driver_budget, bot_kpis, ontology, bot_iron_law, analysis_results, expenses, cash, tax_compliance, verify, growth_quality, products, data_classification, value_sources, zero_based, derivation_rules, cash_classify, cashflow_structure, cash_vouchers_json, ext_input_ledgers, treasury, doc_center, map_completeness, map_summary, map_dimensions
+from app.api import auth, kpis, kpi_governance, templates, maps, dashboard, data, alerts, ai_analysis, ai_suggestions, alert_rules, users, thresholds, notifications, permissions, action_plans, alignment, org, objectives, versions, budget, cost, predict, reports, security, knowledge, bot_bridge, bot_bridge_v2, bot_suggestions, lead, tenant, customer_dashboard, deviation_push, budget_generate, knowledge_articles, kpi_causality, data_quality, bi_reports, entities, bsc_layers, okr, okr_templates, subjects, driver_budget, bot_kpis, ontology, bot_iron_law, analysis_results, expenses, cash, tax_compliance, verify, growth_quality, products, data_classification, value_sources, zero_based, derivation_rules, cash_classify, cashflow_structure, cash_vouchers_json, ext_input_ledgers, treasury, doc_center, map_completeness, map_summary, map_dimensions, ar_aging, charter
 from app.utils.cache import clear_all as clear_cache, delete as delete_cache
 from scripts.erp_sync import run_sync as run_erp_sync
 from app.auth_middleware import require_auth, require_auth_401
@@ -40,6 +40,10 @@ app.include_router(auth.router)
 app.include_router(kpis.router)
 app.include_router(kpi_governance.router)
 app.include_router(templates.router)
+# ⚠️ 顺序要求：完备度/摘要条路由必须先于 maps/objectives 等「/{map_id}/...」通配路由注册，
+#    否则 /maps/completeness/objectives 会被 /{map_id}/objectives 抢先匹配（map_id="completeness" → 422）。
+app.include_router(map_completeness.router)  # 战略地图数据完备度（口径 CMA-CALIBER-MAP-COMPLETENESS-20261018 v1.0）
+app.include_router(map_summary.router)  # 战略地图摘要条（收入分线/新式毛利率/管理指标 · 口径 CMA-A1 终稿）
 app.include_router(maps.router)
 app.include_router(dashboard.router)
 app.include_router(data.router)
@@ -56,8 +60,6 @@ app.include_router(action_plans.router)
 app.include_router(alignment.router)
 app.include_router(org.router)
 app.include_router(objectives.router)
-app.include_router(map_completeness.router)  # 战略地图数据完备度（口径 CMA-CALIBER-MAP-COMPLETENESS-20261018 v1.0）
-app.include_router(map_summary.router)  # 战略地图摘要条（收入分线/新式毛利率/管理指标 · 口径 CMA-A1 终稿）
 app.include_router(versions.router)
 app.include_router(budget.router)
 app.include_router(value_sources.router)
@@ -103,7 +105,9 @@ app.include_router(tax_compliance.router)
 app.include_router(verify.router)
 app.include_router(ext_input_ledgers.router)  # 外部输入台账（M2 准入3 派生，2026-09-15）
 app.include_router(treasury.router)  # 司库（资金）阶段一：账户池/头寸/内控闸门（2026-09-15）
+app.include_router(ar_aging.router)  # 应收账龄看板（T2 2026-10-20：逐户·六档·三口径禁合计）
 app.include_router(doc_center.router)  # 文档中心：md 单一来源（运行期读盘，2026-09-24）
+app.include_router(charter.router)  # 战略基座：战略章程（使命/愿景/价值观，P0 2026-10-21）
 
 @app.exception_handler(Exception)
 async def global_exception_handler(request: Request, exc: Exception):

backend/app/models/__init__.py

@@ -9,6 +9,7 @@ from app.models.driver_budget import DriverFactorTemplate, DriverFactorBudget
 from app.models.product_sales import ProductSales
 from app.models.ext_input_ledger import TaxFilingLog, HrMonthlyLedger, LicenceLedger
 from app.models.report_opening import ReportOpeningBalance
+from app.models.strategy_charter import StrategyCharter
 from app.models.treasury import (
     TreasuryAccount, TreasuryBalance, TreasuryControlRule,
     TreasuryPayment, TreasuryControlHit, TreasuryLimit,
@@ -166,6 +167,8 @@ class KPIAlert(Base):
     alert_level = Column(String(20), default="yellow", comment="green/yellow/red")
     alert_message = Column(String(500), nullable=False)
     alert_type = Column(String(30), default="actual", comment="actual/forecast — 实际值超限/预测值超限")
+    dedupe_key = Column(String(160), nullable=True,
+                        comment="生成侧幂等键 period|scope|ident；仅 alert_type=overdue 填充，历史行为 NULL")
     suggestion = Column(Text, nullable=True, comment="情景建议")
     action_plan_linked_id = Column(Integer, nullable=True, comment="关联的改善计划ID")
     status = Column(String(20), default="pending", comment="pending/processing/resolved")
@@ -363,6 +366,37 @@ class MapObjective(Base):
     created_at = Column(DateTime, server_default=func.now())
     updated_at = Column(DateTime, server_default=func.now(), onupdate=func.now())
 
+class MapConnection(Base):
+    """战略地图四层因果连线（2026-10-21 P1：canvas_data.connections → 独立表）
+
+    迁移来源 = `strategic_maps.canvas_data->connections`，双写期表与 JSON 并存
+    （JSON 保留供前端渲染；表供归因/校验/分析）。
+
+    节点标识双兼容（生产实况两种格式并存）：
+      - `dim:中文目标名`（生产存量实际格式，如 `learning:贡献率思维全员培训`）
+      - `dim-序号`（当前前端 MapCanvas 生成的格式，如 `learning-0`）
+    原始串原样保存在 from_node/to_node，解析结果落 from_layer/to_layer/from_objective_id。
+    """
+    __tablename__ = "map_connections"
+    __table_args__ = (
+        UniqueConstraint("map_id", "from_node", "to_node", name="uk_map_conn_edge"),
+    )
+    id = Column(Integer, primary_key=True, index=True)
+    entity_id = Column(Integer, default=1, index=True, comment="企业ID（冗余，便于按账套直查）")
+    map_id = Column(Integer, ForeignKey("strategic_maps.id", ondelete="CASCADE"), nullable=False, index=True, comment="关联战略地图")
+    from_node = Column(String(255), nullable=False, comment="源节点原始标识（原样保留）")
+    to_node = Column(String(255), nullable=False, comment="目标节点原始标识（原样保留）")
+    from_layer = Column(String(20), nullable=True, comment="解析层: learning/process/customer/finance")
+    to_layer = Column(String(20), nullable=True, comment="解析层: learning/process/customer/finance")
+    from_objective_id = Column(Integer, ForeignKey("map_objectives.id", ondelete="SET NULL"), nullable=True, comment="源目标ID（可解析时回填）")
+    to_objective_id = Column(Integer, ForeignKey("map_objectives.id", ondelete="SET NULL"), nullable=True, comment="目标目标ID（可解析时回填）")
+    effect = Column(String(20), default="positive", comment="positive/negative（存量缺省 positive）")
+    label = Column(String(255), nullable=True, comment="连线标签")
+    sort_order = Column(Integer, default=0, comment="画布内原始顺序（保持渲染一致）")
+    created_at = Column(DateTime, server_default=func.now())
+    updated_at = Column(DateTime, server_default=func.now(), onupdate=func.now())
+
+
 class KPICausality(Base):
     """KPI因果链 — 记录KPI间的因果关系"""
     __tablename__ = "kpi_causality"

backend/app/models/strategy_charter.py

@@ -0,0 +1,64 @@
+"""战略基座 — 战略章程（使命 / 愿景 / 价值观）
+
+任务：`project-20261021-cma-strategy-charter-p0-fullstack`（P0 战略基座）
+依据：《2026-10-21 CMA 知识库 M0 与系统实现差距报告》§四 P0 —— 知识卡
+       `CMA-P1-B-009 愿景、使命与价值观`：使命＝为何存在；愿景＝要成为什么
+       （**带年份与可验证状态**）；价值观＝不可交易的底线，是决策冲突时的裁决依据。
+
+单一事实源：每个账套（entity_id）**一条**章程行，唯一键 uq_strategy_charter_entity。
+本模块只做「表结构 + 契约常量」；读写端点见 app/api/charter.py。
+
+字段语义（派单表：语义不可减，类型可调整）：
+  mission           使命 TEXT NOT NULL
+  vision            愿景 TEXT NOT NULL
+  vision_year       愿景达成年份 VARCHAR(20) NOT NULL（**不可空**；4 位年份）
+  vision_measurable 可验证状态 TEXT NOT NULL（**不可空**；用可核对的规模/业务口径描述）
+  values_text       价值观总述 TEXT NOT NULL
+  value_items       价值观条目 JSON NOT NULL `[{name, meaning}]`
+                    —— 每条须能作"决策冲突时的裁决依据"
+
+隔离：`entity_id NOT NULL` + 唯一键以 entity_id 打头（多租户隔离 + 索引可用性）。
+禁止示例数据：本模块与迁移脚本**零 INSERT**；GET 无行返回 `filled=false` 空态。
+"""
+from sqlalchemy import Column, Integer, String, Text, JSON, DateTime, UniqueConstraint, func
+from app.database import Base
+
+# 愿景年度格式：4 位年份（不臆造业务上限；允许字符串以兼容前端输入）
+VISION_YEAR_RE = r"^\d{4}$"
+
+# 长文本上限（与列宽一致，避免 MySQL 截断静默丢字）
+MISSION_MAX_LEN = 1000
+VISION_MAX_LEN = 1000
+VISION_MEASURABLE_MAX_LEN = 1000
+VALUES_TEXT_MAX_LEN = 2000
+VALUE_ITEM_NAME_MAX_LEN = 50
+VALUE_ITEM_MEANING_MAX_LEN = 300
+VALUE_ITEMS_MAX = 20
+
+
+class StrategyCharter(Base):
+    """战略章程 — 使命/愿景（年度+可验证状态）/价值观（每账套一行）"""
+    __tablename__ = "strategy_charter"
+
+    id = Column(Integer, primary_key=True)
+    entity_id = Column(Integer, nullable=False, comment="企业ID（账套，多租户隔离；唯一键已建索引）")
+    mission = Column(Text, nullable=False, comment="使命：我们为何存在（做什么、为谁做）")
+    vision = Column(Text, nullable=False, comment="愿景：要成为什么")
+    vision_year = Column(String(20), nullable=False, comment="愿景达成年份 YYYY（不可空）")
+    vision_measurable = Column(
+        Text, nullable=False,
+        comment="可验证状态：用可核对的规模/业务口径描述（不可空）",
+    )
+    values_text = Column(Text, nullable=False, comment="价值观总述（不可交易的底线）")
+    value_items = Column(
+        JSON, nullable=False,
+        comment="价值观条目数组 [{name, meaning}]，每条可作决策冲突时的裁决依据",
+    )
+    updated_by = Column(Integer, nullable=True, comment="最后修改人 user.id")
+    updated_by_name = Column(String(50), nullable=True, comment="最后修改人姓名（快照）")
+    created_at = Column(DateTime, server_default=func.now())
+    updated_at = Column(DateTime, server_default=func.now(), onupdate=func.now())
+
+    __table_args__ = (
+        UniqueConstraint("entity_id", name="uq_strategy_charter_entity"),
+    )

backend/app/services/ar_overdue_alert.py

@@ -0,0 +1,918 @@
+"""应收账龄逾期预警（`alert_rules.rule_type='ar_overdue'`）计算与落库
+
+设计要点（任务 project-20261020-ar-overdue-alert-and-aging-board-fullstack）：
+
+* **阈值/责任人/排除名单全部来自 `alert_rules.params`（DB 行）**，本模块不出现任何
+  业务阈值数字；缺键即该判据关闭（显式失效，不回落隐式默认值）。
+* **只 INSERT**：命中写 `kpi_alerts(alert_type='overdue')`，去重键为**生成侧语义键**
+  `dedupe_key = period|scope|ident`（单户 `cust|<unit_code>`、聚合 `agg|<族名>`），
+  由唯一索引兜底；文案口径变化不再产生重复行。历史行 `dedupe_key IS NULL` 走只读反解兼容。
+  **零 UPDATE / 零 DELETE**。
+* `assignee` 必填（`assignee_by_unit[unit_code]` → `assignee_default`），两者皆空
+  ⇒ `AssigneeMissing`（调用方转 400，fail-closed）——禁落空 assignee 的行。
+* 口径：只读 `借方应收户口径`（`app.utils.ar_caliber`），负档位不抹平。
+"""
+from __future__ import annotations
+
+import calendar
+import logging
+import re
+from decimal import Decimal, ROUND_HALF_UP
+from typing import Any, Dict, List, Optional, Tuple
+
+from sqlalchemy import text
+from sqlalchemy.exc import IntegrityError
+from sqlalchemy.orm import Session
+
+from app.models import KPIAlert
+from app.utils import ar_caliber
+
+logger = logging.getLogger("ar_overdue_alert")
+
+RULE_TYPE = "ar_overdue"
+ALERT_TYPE = "overdue"
+MESSAGE_MAX = 500
+
+_CENT = Decimal("0.01")
+
+
+class AssigneeMissing(Exception):
+    """责任人未配置（fail-closed：拒绝运行，不落空 assignee 行）。"""
+
+
+class ParamsInvalid(Exception):
+    """规则参数结构非法。"""
+
+
+# ============================================================
+# 参数校验（只校结构，不设默认阈值）
+# ============================================================
+def validate_params(params: Any) -> Optional[str]:
+    """返回错误文案；None = 合法。"""
+    if params is None:
+        return "params 不能为空（ar_overdue 规则参数全量走 DB，代码不设业务默认值）"
+    if not isinstance(params, dict):
+        return "params 必须是 JSON 对象"
+    try:
+        ar_caliber.validate_caliber(params.get("caliber") or ar_caliber.DEFAULT_CALIBER)
+    except ar_caliber.CaliberError as e:
+        return str(e)
+    for key in ("red", "yellow"):
+        # 2026-10-21 阶段B：G5 新参数形状为扁平键（share_90_150_* 等），red/yellow 分组可缺省；
+        # 存在则必须是对象（防误传标量导致阈值静默失效）。
+        if params.get(key) is not None and not isinstance(params.get(key), dict):
+            return "params.%s 必须是对象（阈值分组）" % key
+    for key in ("assignee_by_unit", "action_texts"):
+        if key in params and params[key] is not None and not isinstance(params[key], dict):
+            return "params.%s 必须是对象" % key
+    if "exclude_unit_codes" in params and params["exclude_unit_codes"] is not None:
+        v = params["exclude_unit_codes"]
+        if not isinstance(v, (list, tuple)):
+            return "params.exclude_unit_codes 必须是数组（排他名单）"
+    return None
+
+
+def _params_caliber(params: Dict[str, Any]) -> str:
+    return params.get("caliber") or ar_caliber.DEFAULT_CALIBER
+
+
+def _action_text(params: Dict[str, Any], level: str) -> str:
+    texts = params.get("action_texts") or {}
+    return str(texts.get(level) or texts.get("default") or "")
+
+
+# ============================================================
+# assignee 解析（任务 §6.2 定值；全部只读 SELECT，禁止落空 assignee 行）
+#   优先级：assignee_by_unit 手工点名 → 近 3 个月主业务员 → 部门负责人
+#   两路皆失败 ⇒ 该户不生成预警，进 `skipped[]`（含原因）
+# ============================================================
+ASSIGNEE_MODES = ("salesman_then_dept", "manual_then_salesman_then_dept", "dept_only")
+
+
+def _source_cfg(params: Dict[str, Any]) -> Dict[str, Any]:
+    src = params.get("assignee_source")
+    return src if isinstance(src, dict) else {}
+
+
+def _assignee_by_unit(params: Dict[str, Any], unit_code: Optional[str]) -> Optional[str]:
+    by_unit = params.get("assignee_by_unit") or {}
+    if unit_code is None:
+        return None
+    v = by_unit.get(str(unit_code))
+    return str(v).strip() if v else None
+
+
+def ensure_assignee_configured(params: Dict[str, Any]) -> None:
+    """责任人来源整体缺失 ⇒ 400 fail-closed（§4.8 负测）。
+
+    §6.2 允许逐户派生（主业务员 → 部门负责人），因此「配了 assignee_source.mode」
+    也算已配置；派生失败的**单户**进 skipped[]，不落空 assignee 行。
+    完全没有任何来源（无 default / 无 by_unit / 无 mode）才拒绝整体运行。
+    """
+    if str(params.get("assignee_default") or "").strip():
+        return
+    if params.get("assignee_by_unit"):
+        return
+    if str(_source_cfg(params).get("mode") or "") in ASSIGNEE_MODES:
+        return
+    raise AssigneeMissing(
+        "ar_overdue 规则未配置任何责任人来源：params.assignee_default / assignee_by_unit / "
+        "assignee_source.mode 均为空，按 fail-closed 拒绝运行（禁止落空 assignee 的预警行）"
+    )
+
+
+def _has_table(db: Session, name: str) -> bool:
+    """表是否可读（方言无关，不依赖 information_schema）。"""
+    try:
+        db.execute(text("SELECT 1 FROM %s LIMIT 1" % name))
+        return True
+    except Exception:  # noqa: BLE001 — 缺表/无权限一律降级
+        return False
+
+
+def _employee_name(db: Session, emp_id: Any) -> Optional[str]:
+    """employee_dict.u_Name（只读）。缺表/无该 id ⇒ None（调用方继续降级）。"""
+    try:
+        emp = int(emp_id)
+    except (TypeError, ValueError):
+        return None
+    if emp <= 0 or not _has_table(db, ar_caliber.EMPLOYEE_DICT_TABLE):
+        return None
+    try:
+        row = db.execute(
+            text("SELECT u_Name FROM %s WHERE id = :i" % ar_caliber.EMPLOYEE_DICT_TABLE),
+            {"i": emp},
+        ).first()
+        return str(row[0]).strip() if row and row[0] else None
+    except Exception:  # noqa: BLE001
+        return None
+
+
+def _operator_from_abst(abst: Any) -> Optional[str]:
+    """单据摘要尾段制单人姓名（`…：姓名` / `…元;姓名`）——employee_dict 缺表时的同库只读替代源。
+
+    摘要里常混入金额/银行/备注（如 `…【建设银行】270000.00元;兀晨博`），因此逐段
+    取尾段后做**形状校验**：纯中文姓名、长度 ≤ 20、不含数字/括号/金额符号，否则判失败。
+    """
+    if not abst:
+        return None
+    s = str(abst).strip()
+    for sep in ("：", ";"):
+        if sep in s:
+            s = s.rsplit(sep, 1)[-1]
+    s = s.strip(" \u3000;；")
+    if not s or len(s) > 20:
+        return None
+    if any(ch.isdigit() for ch in s):
+        return None
+    if any(ch in s for ch in "【】[]()（）元,，.。:："):
+        return None
+    return s
+
+
+def _period_window(period: str, months: int) -> Tuple[str, str]:
+    """账龄期 → 业务员回看窗口 [期末前 (months-1) 个月 1 日, 期末]。"""
+    y, m = int(period[:4]), int(period[5:7])
+    ly, lm = y, m - (months - 1)
+    while lm <= 0:
+        lm += 12
+        ly -= 1
+    end_day = calendar.monthrange(y, m)[1]
+    return "%04d-%02d-01" % (ly, lm), "%04d-%02d-%02d" % (y, m, end_day)
+
+
+def _resolve_unit_ref(db: Session, entity_id: int, unit_code: Any, unit_name: Any,
+                      start: str, end: str) -> Tuple[Optional[int], str, str]:
+    """unit_code → zxt_masterbill.unit_id。返回 (unit_id, 来源, 失败原因)。"""
+    code = str(unit_code or "").strip()
+    if _has_table(db, ar_caliber.UNITS_DICT_TABLE):
+        for code_col in ("unit_code", "u_code", "code"):
+            try:
+                row = db.execute(
+                    text("SELECT unit_id FROM %s WHERE %s = :c"
+                         % (ar_caliber.UNITS_DICT_TABLE, code_col)),
+                    {"c": code},
+                ).first()
+                if row and row[0] is not None:
+                    return int(row[0]), "units_dict.%s" % code_col, ""
+            except Exception:  # noqa: BLE001 — 列名不匹配则试下一列
+                continue
+    if code.isdigit():
+        return int(code), "numeric_code", ""
+    if unit_name:
+        try:
+            row = db.execute(
+                text("SELECT unit_id, COUNT(*) AS n FROM %s WHERE entity_id = :e "
+                     "AND abst LIKE :kw AND bill_date >= :s AND bill_date <= :d "
+                     "GROUP BY unit_id ORDER BY n DESC LIMIT 1" % ar_caliber.ZXT_MASTERBILL_TABLE),
+                {"e": entity_id, "kw": "%%%s%%" % unit_name, "s": start, "d": end},
+            ).first()
+            if row and row[0] is not None:
+                return int(row[0]), "abst_name_match", ""
+        except Exception:  # noqa: BLE001
+            pass
+    return None, "", "unit_link_unresolved(code=%s)" % (code or "-")
+
+
+def _salesman_name(db: Session, entity_id: int, unit_id: int, start: str, end: str,
+                   allow_abst_fallback: bool) -> Tuple[Optional[str], Optional[int], str]:
+    """该户近窗主业务员（单据数最多者；同数取最近单据）。"""
+    try:
+        row = db.execute(
+            text("SELECT emp_id, COUNT(*) AS n FROM %s WHERE entity_id = :e AND unit_id = :u "
+                 "AND bill_date >= :s AND bill_date <= :d AND emp_id IS NOT NULL AND emp_id > 0 "
+                 "GROUP BY emp_id ORDER BY n DESC, MAX(bill_date) DESC LIMIT 1"
+                 % ar_caliber.ZXT_MASTERBILL_TABLE),
+            {"e": entity_id, "u": unit_id, "s": start, "d": end},
+        ).first()
+    except Exception as exc:  # noqa: BLE001
+        return None, None, "salesman_query_failed:%s" % type(exc).__name__
+    if not row or row[0] is None:
+        return None, None, "no_recent_bill(unit_id=%s)" % unit_id
+    emp_id = int(row[0])
+    name = _employee_name(db, emp_id)
+    if name:
+        return name, emp_id, ""
+    if not allow_abst_fallback:
+        return None, emp_id, "salesman_name_unresolved(emp_id=%s, employee_dict 缺表)" % emp_id
+    try:
+        row2 = db.execute(
+            text("SELECT abst FROM %s WHERE entity_id = :e AND unit_id = :u AND emp_id = :m "
+                 "AND bill_date >= :s AND bill_date <= :d ORDER BY bill_date DESC, id DESC LIMIT 1"
+                 % ar_caliber.ZXT_MASTERBILL_TABLE),
+            {"e": entity_id, "u": unit_id, "m": emp_id, "s": start, "d": end},
+        ).first()
+        name = _operator_from_abst(row2[0] if row2 else None)
+    except Exception:  # noqa: BLE001
+        name = None
+    if name:
+        return name, emp_id, ""
+    return None, emp_id, "salesman_name_unresolved(emp_id=%s, employee_dict 缺表且摘要无姓名)" % emp_id
+
+
+def _dept_assignee(db: Session, unit_code: Any,
+                   params: Dict[str, Any]) -> Tuple[Optional[str], str]:
+    """部门负责人兜底（§6.2：渠道 118 / 商务 114 / 直属酒窖 117）。
+
+    部门归属无法自动判定时**不猜**：要求 params.assignee_source.dept_by_unit /
+    dept_by_prefix / dept_default 之一给出部门键，否则返回原因。
+    """
+    src = _source_cfg(params)
+    fb = src.get("dept_fallback") or {}
+    code = str(unit_code or "")
+    key = (src.get("dept_by_unit") or {}).get(code) or src.get("dept_default")
+    if key is None:
+        for prefix, k in (src.get("dept_by_prefix") or {}).items():
+            if code.startswith(str(prefix)):
+                key = k
+                break
+    if key is None:
+        return None, "dept_unclassified"
+    if key not in fb:
+        return None, "dept_key_unmapped:%s" % key
+    name = _employee_name(db, fb[key])
+    if name:
+        return name, "dept:%s(id=%s)" % (key, fb[key])
+    return None, "dept_head_name_unresolved:%s(id=%s, employee_dict 缺表)" % (key, fb[key])
+
+
+def resolve_assignee(db: Session, entity_id: int, unit_code: Any, unit_name: Any,
+                     params: Dict[str, Any], period: str) -> Dict[str, Any]:
+    """单户 assignee 解析结果：{name, source, emp_id, reason}。name=None ⇒ 不生成该户。"""
+    manual = _assignee_by_unit(params, str(unit_code) if unit_code is not None else None)
+    if manual:
+        return {"name": manual, "source": "manual:assignee_by_unit", "emp_id": None, "reason": ""}
+
+    src = _source_cfg(params)
+    mode = str(src.get("mode") or "")
+    reasons: List[str] = []
+    if mode in ASSIGNEE_MODES:
+        months = int(src.get("lookback_months") or 3)
+        start, end = _period_window(period, months)
+        if mode != "dept_only":
+            unit_id, link, link_reason = _resolve_unit_ref(
+                db, entity_id, unit_code, unit_name, start, end)
+            if unit_id is not None:
+                name, emp_id, why = _salesman_name(
+                    db, entity_id, unit_id, start, end,
+                    bool(src.get("abst_operator_fallback", True)))
+                if name:
+                    return {"name": name, "source": "salesman:%s(emp_id=%s, %s)"
+                            % (link, emp_id, "employee_dict" if _has_table(db, ar_caliber.EMPLOYEE_DICT_TABLE)
+                               else "abst_operator"), "emp_id": emp_id, "reason": ""}
+                reasons.append(why)
+            else:
+                reasons.append(link_reason)
+        name, why = _dept_assignee(db, unit_code, params)
+        if name:
+            return {"name": name, "source": why, "emp_id": None, "reason": ""}
+        reasons.append(why)
+        return {"name": None, "source": "", "emp_id": None, "reason": "; ".join(r for r in reasons if r)}
+
+    default = str(params.get("assignee_default") or "").strip()
+    if default:
+        return {"name": default, "source": "assignee_default", "emp_id": None, "reason": ""}
+    return {"name": None, "source": "", "emp_id": None, "reason": "no_assignee_source"}
+
+
+def aggregate_assignee(params: Dict[str, Any]) -> Tuple[Optional[str], str]:
+    """聚合预警的责任人（整体口径）：assignee_default → assignee_source.aggregate_assignee。"""
+    d = str(params.get("assignee_default") or "").strip()
+    if d:
+        return d, "assignee_default"
+    a = str(_source_cfg(params).get("aggregate_assignee") or "").strip()
+    if a:
+        return a, "assignee_source.aggregate_assignee"
+    return None, "aggregate_assignee_unresolved"
+
+
+
+# ============================================================
+# 取数与计算
+# ============================================================
+def _dec(v: Any) -> Decimal:
+    if v is None:
+        return Decimal("0")
+    return Decimal(str(v))
+
+
+def _money(v: Decimal) -> float:
+    return float(v.quantize(_CENT, rounding=ROUND_HALF_UP))
+
+
+def _fmt_money(v: Decimal) -> str:
+    return "{:,.2f}".format(v.quantize(_CENT, rounding=ROUND_HALF_UP))
+
+
+def _fmt_pct(ratio: Optional[float]) -> str:
+    if ratio is None:
+        return "-"
+    return "{:.2f}".format(ratio * 100.0)
+
+
+_SELECT_COLS = (
+    "unit_code, unit_name, opening_amt, total_amt, "
+    + ", ".join(ar_caliber.BUCKET_COLUMNS)
+)
+
+
+def _periods(db: Session, entity_id: int, caliber: str) -> List[str]:
+    rows = db.execute(
+        text(
+            "SELECT DISTINCT period FROM %s WHERE entity_id = :e AND caliber = :c ORDER BY period"
+            % ar_caliber.AR_AGING_TABLE
+        ),
+        {"e": entity_id, "c": caliber},
+    ).all()
+    return [r[0] for r in rows]
+
+
+def _rows_for_period(db: Session, entity_id: int, caliber: str, period: str) -> List[Dict[str, Any]]:
+    sql = (
+        "SELECT %s FROM %s WHERE entity_id = :e AND caliber = :c AND period = :p "
+        "ORDER BY total_amt DESC, unit_code ASC" % (_SELECT_COLS, ar_caliber.AR_AGING_TABLE)
+    )
+    return [dict(r) for r in db.execute(text(sql), {"e": entity_id, "c": caliber, "p": period}).mappings().all()]
+
+
+def _oldest_overdue_days(row: Dict[str, Any]) -> int:
+    """最老非零逾期档的下界天数（31/61/91/121/151）；无逾期 = 0。"""
+    days = 0
+    for col, lower in ar_caliber.OVERDUE_BUCKETS:
+        if _dec(row.get(col)) != 0:
+            days = max(days, lower)
+    return days
+
+
+def _level_share(share: Optional[float], params: Dict[str, Any]) -> Optional[str]:
+    red = params.get("red") or {}
+    yellow = params.get("yellow") or {}
+    gt = red.get("share_90plus_gt")
+    if gt is not None and share is not None and share > float(gt):
+        return "red"
+    ymin, ymax = yellow.get("share_90plus_min"), yellow.get("share_90plus_max")
+    if share is not None and ymin is not None and ymax is not None:
+        if float(ymin) <= share <= float(ymax):
+            return "yellow"
+    return None
+
+
+def _level_mom(mom_pct: Optional[float], params: Dict[str, Any]) -> Optional[str]:
+    red = params.get("red") or {}
+    gte = red.get("count_mom_pct_gte")
+    if gte is not None and mom_pct is not None and mom_pct >= float(gte):
+        return "red"
+    return None
+
+
+def _level_customer(overdue_amt: Decimal, oldest_days: int, params: Dict[str, Any]) -> Optional[str]:
+    red = params.get("red") or {}
+    yellow = params.get("yellow") or {}
+    # 阶段B G5 扁平参数：顶层 single_overdue_amt_gt 优先（未配 red 分组时仍可判红线）
+    red_amt = red.get("single_overdue_amt_gt")
+    if red_amt is None:
+        red_amt = params.get("single_overdue_amt_gt")
+    red_min_days = red.get("single_overdue_min_days")
+    if red_amt is not None and overdue_amt > _dec(red_amt):
+        if red_min_days is None or oldest_days >= int(red_min_days):
+            return "red"
+    ymin = yellow.get("single_overdue_amt_min")
+    ymax = yellow.get("single_overdue_amt_max")
+    if ymin is not None and ymax is not None:
+        if _dec(ymin) <= overdue_amt <= _dec(ymax):
+            return "yellow"
+    return None
+
+
+def _clip(msg: str) -> str:
+    return msg if len(msg) <= MESSAGE_MAX else msg[: MESSAGE_MAX - 1] + "…"
+
+
+def compute(
+    db: Session,
+    entity_id: int,
+    period: Optional[str] = None,
+    params: Optional[Dict[str, Any]] = None,
+) -> Dict[str, Any]:
+    """计算逾期预警命中集合（纯读，不写库）。
+
+    返回:
+      {period, caliber, metrics, aggregates[], per_customer[], skipped[], skipped_reason, notes[]}
+    `skipped_reason` 非空表示本次不生成任何预警（缺期/无数据/分母非正…）。
+    `skipped[]` 记录 assignee 解析失败而未生成的户/聚合项（§6.2 禁落空 assignee）。
+    """
+    params = params or {}
+    err = validate_params(params)
+    if err:
+        raise ParamsInvalid(err)
+    caliber = _params_caliber(params)
+    notes: List[str] = []
+    out: Dict[str, Any] = {
+        "period": period,
+        "caliber": caliber,
+        "metrics": {},
+        "aggregates": [],
+        "per_customer": [],
+        "skipped": [],
+        "skipped_reason": None,
+        "notes": notes,
+    }
+
+    # ── 期间解析：默认表内 MAX(period)；缺期直接 skipped（不补 0）──
+    periods = _periods(db, entity_id, caliber)
+    if not periods:
+        out["skipped_reason"] = "no_data_in_table:%s/%s" % (entity_id, caliber)
+        return out
+    if not period:
+        period = periods[-1]
+        out["period"] = period
+        notes.append("period 未指定 → 取表内 MAX(period)=%s" % period)
+    if period not in periods:
+        out["skipped_reason"] = "no_data_for_period:%s" % period
+        return out
+
+    rows = _rows_for_period(db, entity_id, caliber, period)
+    if not rows:
+        out["skipped_reason"] = "no_data_for_period:%s" % period
+        return out
+
+    exclude = set(params.get("exclude_unit_codes") or [])
+    if exclude:
+        before = len(rows)
+        rows = [r for r in rows if str(r.get("unit_code")) not in exclude]
+        notes.append("排除名单命中 %d 行（exclude_unit_codes）" % (before - len(rows)))
+    if not rows:
+        out["skipped_reason"] = "all_rows_excluded"
+        return out
+
+    total = sum((_dec(r.get("total_amt")) for r in rows), Decimal("0"))
+    overdue_90plus = sum((_dec(r.get(c)) for r in rows for c in ar_caliber.NINETY_PLUS_COLUMNS), Decimal("0"))
+    share = (float(overdue_90plus / total) if total > 0 else None)
+    if total <= 0:
+        notes.append("分母 SUM(total_amt)=%.2f ≤ 0 → 跳过聚合占比判据（禁除零）" % float(total))
+
+    # 逐户：逾期额 = total_amt - b_0_30（等价 31-60..150+ 之和，负档位不抹平）
+    customers: List[Dict[str, Any]] = []
+    n_overdue = 0
+    for r in rows:
+        o_i = _dec(r.get("total_amt")) - _dec(r.get("b_0_30"))
+        oldest = _oldest_overdue_days(r)
+        if o_i > 0:
+            n_overdue += 1
+        level = _level_customer(o_i, oldest, params) if o_i > 0 else None
+        customers.append({"row": r, "overdue_amt": o_i, "oldest_days": oldest, "level": level})
+
+    # 环比：上一个有数据期（同口径同算法）；上期无数据 ⇒ 不算环比（文案写明，不得当 0）
+    idx = periods.index(period)
+    prev_period = periods[idx - 1] if idx > 0 else None
+    n_prev = None
+    prev_rows: List[Dict[str, Any]] = []
+    if prev_period:
+        prev_rows = _rows_for_period(db, entity_id, caliber, prev_period)
+        if exclude:
+            prev_rows = [r for r in prev_rows if str(r.get("unit_code")) not in exclude]
+        n_prev = sum(
+            1 for r in prev_rows
+            if (_dec(r.get("total_amt")) - _dec(r.get("b_0_30"))) > 0
+        )
+    mom_pct = None
+    if n_prev and n_prev > 0:
+        mom_pct = (n_overdue - n_prev) / n_prev
+    else:
+        notes.append("上期无数据（%s）→ 不计算环比" % (prev_period or "无更早期间"))
+
+    metrics = {
+        "period": period,
+        "caliber": caliber,
+        "total_amt": _money(total),
+        "overdue_90plus": _money(overdue_90plus),
+        "share_90plus": (round(share, 6) if share is not None else None),
+        "overdue_amt": _money(sum((c["overdue_amt"] for c in customers), Decimal("0"))),
+        "customers": len(rows),
+        "overdue_customers": n_overdue,
+        "prev_period": prev_period,
+        "prev_overdue_customers": n_prev,
+        "count_mom_pct": (round(mom_pct, 6) if mom_pct is not None else None),
+    }
+    out["metrics"] = metrics
+
+    # ── 聚合判据（§6.1）：90 天占比（red 键 null ⇒ 停用）/ 逾期户数环比 / 150+ 结构恶化 ──
+    agg_assignee, agg_src = aggregate_assignee(params)
+
+    def _push_agg(key: str, level: str, metrics_dict: Dict[str, Any], msg: str, action: str) -> None:
+        blocked = None if agg_assignee else agg_src
+        out["aggregates"].append({
+            "key": key,
+            "level": level,
+            "metrics": metrics_dict,
+            "alert_message": msg,
+            "suggestion": action,
+            "assignee": agg_assignee or "",
+            "assignee_source": agg_src,
+            "blocked": blocked,
+        })
+        if blocked:
+            out["skipped"].append({"scope": "aggregate", "key": key, "level": level,
+                                   "reason": blocked})
+
+    red_cfg = params.get("red") or {}
+    yellow_cfg = params.get("yellow") or {}
+    red_gt = red_cfg.get("share_90plus_gt")
+    if share is not None and red_gt is None:
+        notes.append("red.share_90plus_gt 为 null ⇒ 90 天占比红线停用（不回落 0.25）")
+
+    # 聚合判据 1：90 天以上占比（红线停用时只剩黄线 15%–25%）
+    if share is not None:
+        level = _level_share(share, params)
+        if level:
+            action = _action_text(params, level)
+            thr = ("红线 >%g%%" % (float(red_gt) * 100)) if red_gt is not None else "红线停用（DB 置 null）"
+            msg = _clip(
+                "【应收账款逾期】%s %s 90 天以上 ¥%s，占应收总额 %s%%（%s）｜建议：%s"
+                % (period, caliber, _fmt_money(overdue_90plus), _fmt_pct(share), thr,
+                   action or "见规则配置")
+            )
+            _push_agg("share_90plus", level,
+                      {"overdue_90plus": _money(overdue_90plus), "total_amt": _money(total),
+                       "share_90plus": round(share, 6)},
+                      msg, action)
+
+    # 聚合判据 2：逾期户数环比
+    level = _level_mom(mom_pct, params)
+    if level:
+        action = _action_text(params, level)
+        gte = red_cfg.get("count_mom_pct_gte")
+        msg = _clip(
+            "【应收账款逾期】%s 逾期户数 %d 户，环比 %s%%（红线 +%s%%）｜建议：%s"
+            % (period, n_overdue,
+               ("%+.2f" % (mom_pct * 100)) if mom_pct is not None else "-",
+               ("%g" % (float(gte) * 100)) if gte is not None else "-",
+               action or "见规则配置")
+        )
+        _push_agg("count_mom", level,
+                  {"overdue_customers": n_overdue, "prev_overdue_customers": n_prev,
+                   "count_mom_pct": (round(mom_pct, 6) if mom_pct is not None else None)},
+                  msg, action)
+
+    # 聚合判据 3：账龄结构恶化（G5 2026-10-21 裁定）
+    #   · 新判据（参数化）：90–150 天「前池」占比环比上升 ≥ rise_pp_min **且** 该档余额 ≥ amt_min（AND）
+    #   · 兼容旧参数：未配 share_90_150_rise_pp_min 时，沿用 150+ 占比环比上升（red.structure_worsen）
+    share_90_150 = None
+    prev_share_90_150 = None
+    amt_90_150 = None
+    prev_amt_90_150 = None
+    if total > 0:
+        amt_90_150 = sum((_dec(r.get("b_90_120")) + _dec(r.get("b_120_150")) for r in rows),
+                         Decimal("0"))
+        share_90_150 = float(amt_90_150 / total)
+    if prev_rows:
+        prev_total_90150 = sum((_dec(r.get("total_amt")) for r in prev_rows), Decimal("0"))
+        prev_amt_90_150 = sum((_dec(r.get("b_90_120")) + _dec(r.get("b_120_150")) for r in prev_rows),
+                              Decimal("0"))
+        if prev_total_90150 > 0:
+            prev_share_90_150 = float(prev_amt_90_150 / prev_total_90150)
+
+    rise_pp_min = params.get("share_90_150_rise_pp_min")
+    amt_min_90_150 = params.get("share_90_150_amt_min")
+    if rise_pp_min is not None and amt_min_90_150 is not None:
+        delta_pp = None
+        if share_90_150 is not None and prev_share_90_150 is not None:
+            delta_pp = (share_90_150 - prev_share_90_150) * 100.0
+        hit_90150 = (
+            delta_pp is not None and delta_pp >= float(rise_pp_min)
+            and amt_90_150 is not None and amt_90_150 >= _dec(amt_min_90_150)
+        )
+        if hit_90150:
+            action = _action_text(params, "red")
+            msg = _clip(
+                "【应收账款逾期】%s 90–150 天档占比 %s%%（上期 %s%%）环比 %+.2fpp ≥ %gpp，"
+                "该档余额 ¥%s（≥¥%s），账龄前池恶化｜建议：%s"
+                % (period, _fmt_pct(share_90_150), _fmt_pct(prev_share_90_150), delta_pp,
+                   float(rise_pp_min), _fmt_money(amt_90_150), _fmt_money(_dec(amt_min_90_150)),
+                   action or "见规则配置")
+            )
+            _push_agg("structure_90_150", "red",
+                      {"share_90_150": round(share_90_150, 6),
+                       "prev_period": prev_period,
+                       "prev_share_90_150": (round(prev_share_90_150, 6)
+                                             if prev_share_90_150 is not None else None),
+                       "delta_pp": round(delta_pp, 4),
+                       "amt_90_150": _money(amt_90_150),
+                       "rise_pp_min": float(rise_pp_min),
+                       "amt_min": float(amt_min_90_150)},
+                      msg, action)
+        else:
+            notes.append(
+                "90–150 档：本期 %s%% / 上期 %s%%（Δ=%spp，门槛 ≥%gpp）｜余额 ¥%s（门槛 ≥¥%s）"
+                "⇒ 结构恶化判据未触发"
+                % (_fmt_pct(share_90_150), _fmt_pct(prev_share_90_150),
+                   ("%+.2f" % delta_pp) if delta_pp is not None else "-",
+                   float(rise_pp_min),
+                   _fmt_money(amt_90_150) if amt_90_150 is not None else "-",
+                   _fmt_money(_dec(amt_min_90_150))))
+        metrics["share_90_150"] = (round(share_90_150, 6) if share_90_150 is not None else None)
+        metrics["prev_share_90_150"] = (round(prev_share_90_150, 6)
+                                        if prev_share_90_150 is not None else None)
+        metrics["amt_90_150"] = (_money(amt_90_150) if amt_90_150 is not None else None)
+    else:
+        share_150 = None
+        prev_share_150 = None
+        if total > 0:
+            share_150 = float(sum((_dec(r.get("b_150p")) for r in rows), Decimal("0")) / total)
+        if prev_rows:
+            prev_total = sum((_dec(r.get("total_amt")) for r in prev_rows), Decimal("0"))
+            if prev_total > 0:
+                prev_share_150 = float(
+                    sum((_dec(r.get("b_150p")) for r in prev_rows), Decimal("0")) / prev_total)
+        structure_on = bool(red_cfg.get("structure_worsen"))
+        if structure_on:
+            if share_150 is not None and prev_share_150 is not None and share_150 > prev_share_150:
+                action = _action_text(params, "red")
+                msg = _clip(
+                    "【应收账款逾期】%s 150+ 档占比 %s%%（上期 %s%%）环比上升 %+.2fpp，账龄结构恶化"
+                    "｜建议：%s" % (period, _fmt_pct(share_150), _fmt_pct(prev_share_150),
+                                   (share_150 - prev_share_150) * 100.0, action or "见规则配置")
+                )
+                _push_agg("structure_150p", "red",
+                          {"share_150p": round(share_150, 6),
+                           "prev_period": prev_period,
+                           "prev_share_150p": (round(prev_share_150, 6)
+                                               if prev_share_150 is not None else None),
+                           "delta_pp": round((share_150 - prev_share_150) * 100.0, 4)},
+                          msg, action)
+            else:
+                notes.append(
+                    "150+ 占比：本期 %s%% / 上期 %s%% ⇒ 结构恶化判据未触发"
+                    % (_fmt_pct(share_150), _fmt_pct(prev_share_150)))
+        metrics["share_150p"] = (round(share_150, 6) if share_150 is not None else None)
+        metrics["prev_share_150p"] = (round(prev_share_150, 6)
+                                      if prev_share_150 is not None else None)
+
+    # ── 逐户：assignee 派生失败 ⇒ 不生成（进 skipped[]），禁落空 assignee ──
+    for c in customers:
+        if not c["level"]:
+            continue
+        r = c["row"]
+        who = resolve_assignee(db, entity_id, r.get("unit_code"), r.get("unit_name"), params, period)
+        if not who["name"]:
+            out["skipped"].append({
+                "scope": "customer",
+                "unit_code": r.get("unit_code"),
+                "unit_name": r.get("unit_name"),
+                "level": c["level"],
+                "overdue_amt": _money(c["overdue_amt"]),
+                "oldest_days": c["oldest_days"],
+                "reason": who["reason"],
+            })
+            continue
+        action = _action_text(params, c["level"])
+        unit_total = _dec(r.get("total_amt"))
+        pct = (float(c["overdue_amt"] / unit_total) if unit_total > 0 else None)
+        msg = _clip(
+            "【应收账款逾期】%s（%s）%s 账龄最老 %d 天，逾期合计 ¥%s（占该户 %s%%）｜建议：%s｜责任人：%s"
+            % (
+                r.get("unit_name"), r.get("unit_code"), period, c["oldest_days"],
+                _fmt_money(c["overdue_amt"]), _fmt_pct(pct), action or "见规则配置", who["name"],
+            )
+        )
+        out["per_customer"].append({
+            "unit_code": r.get("unit_code"),
+            "unit_name": r.get("unit_name"),
+            "overdue_amt": _money(c["overdue_amt"]),
+            "total_amt": _money(unit_total),
+            "oldest_days": c["oldest_days"],
+            "pct_of_unit": (round(pct, 6) if pct is not None else None),
+            "level": c["level"],
+            "alert_message": msg,
+            "suggestion": action,
+            "assignee": who["name"],
+            "assignee_source": who["source"],
+            "emp_id": who["emp_id"],
+        })
+    return out
+
+
+# ============================================================
+# 落库（只 INSERT；同 message 幂等跳过）
+# ============================================================
+def hits(result: Dict[str, Any]) -> List[Dict[str, Any]]:
+    """可落库命中项：只收 assignee 非空的（§6.2 禁落空 assignee 行）。"""
+    all_hits = list(result.get("aggregates") or []) + list(result.get("per_customer") or [])
+    return [h for h in all_hits if str(h.get("assignee") or "").strip()]
+
+
+# ============================================================
+# 去重键（生成侧语义键）：period|scope|ident
+# ============================================================
+_AGG_KEY_FAMILY = {
+    "share_90plus": "share_90plus",
+    "count_mom": "count_mom",
+    "structure_90_150": "structure",   # 族名，不跟随判据代次
+    "structure_150p": "structure",
+}
+
+_CUST_MSG_RE = re.compile(r"（([A-Za-z0-9_\-]+)）(\d{4}-\d{2})")
+_PERIOD_RE = re.compile(r"(\d{4}-\d{2})")
+
+
+def _agg_family(key: str) -> str:
+    if key in _AGG_KEY_FAMILY:
+        return _AGG_KEY_FAMILY[key]
+    return str(key).split("_")[0] or "unknown"
+
+
+def _item_dedupe_key(item: Dict[str, Any], period: str) -> Optional[str]:
+    """生成侧幂等键；不解析文案。
+
+    - 单户：`<period>|cust|<unit_code>`
+    - 聚合：`<period>|agg|<族名>`（structure_90_150 / structure_150p 同归 structure）
+    - 无法判定 → None（调用方退回文案逐字比较，方向安全）
+    """
+    p = str(item.get("period") or period or "").strip()
+    if not p:
+        return None
+    unit_code = str(item.get("unit_code") or "").strip()
+    if unit_code:
+        return "%s|cust|%s" % (p, unit_code)
+    key = str(item.get("key") or "").strip()
+    if key:
+        return "%s|agg|%s" % (p, _agg_family(key))
+    return None
+
+
+def _key_from_message(msg: str) -> Optional[str]:
+    """历史兼容：从旧文案只读反解 dedupe_key（仅用于 dedupe_key IS NULL 的行）。
+
+    反解失败返回 None ⇒ 视为未命中 ⇒ 照常插入（fail-open，多插一条而非吞真实告警）。
+    """
+    if not msg:
+        return None
+    m = _CUST_MSG_RE.search(msg)
+    if m:
+        return "%s|cust|%s" % (m.group(2), m.group(1))
+    pm = _PERIOD_RE.search(msg)
+    if not pm:
+        return None
+    period = pm.group(1)
+    if "90 天以上" in msg and "占应收总额" in msg:
+        return "%s|agg|share_90plus" % period
+    if "逾期户数" in msg and "环比" in msg:
+        return "%s|agg|count_mom" % period
+    if "账龄" in msg and ("结构" in msg or "90–150" in msg or "150 天以上" in msg):
+        return "%s|agg|structure" % period
+    return None
+
+
+def _legacy_keys(db: Session, entity_id: int, kpi_id: int) -> List[str]:
+    """取历史行（dedupe_key IS NULL）反解出的键集合（边界层，量级=历史行数）。"""
+    rows = db.query(KPIAlert.alert_message).filter(
+        KPIAlert.entity_id == entity_id,
+        KPIAlert.kpi_id == kpi_id,
+        KPIAlert.alert_type == ALERT_TYPE,
+        KPIAlert.dedupe_key.is_(None),
+    ).all()
+    keys = []
+    for (m,) in rows:
+        k = _key_from_message(m or "")
+        if k:
+            keys.append(k)
+    return keys
+
+
+def count_would_insert(db: Session, entity_id: int, kpi_id: Optional[int],
+                       result: Dict[str, Any]) -> int:
+    """dry-run 用：按与 write_alerts 完全相同的判重口径，算出真正会插入的条数。"""
+    if kpi_id is None:
+        return 0
+    period = str(result.get("period") or "")
+    existing = {
+        k for (k,) in db.query(KPIAlert.dedupe_key).filter(
+            KPIAlert.entity_id == entity_id,
+            KPIAlert.kpi_id == kpi_id,
+            KPIAlert.alert_type == ALERT_TYPE,
+            KPIAlert.dedupe_key.isnot(None),
+        ).all() if k
+    }
+    existing.update(_legacy_keys(db, entity_id, kpi_id))
+    seen = set()
+    n = 0
+    for h in hits(result):
+        msg = h.get("alert_message") or ""
+        if not msg:
+            continue
+        dk = _item_dedupe_key(h, period)
+        if dk is None:
+            # 无键 ⇒ 退回文案逐字（与旧行为一致）
+            n += 1
+            continue
+        if dk in existing or dk in seen:
+            continue
+        seen.add(dk)
+        n += 1
+    return n
+
+
+def write_alerts(db: Session, entity_id: int, kpi_id: int, result: Dict[str, Any]) -> Dict[str, int]:
+    """把命中集合写成 `kpi_alerts(alert_type='overdue')`；历史行零触碰。
+
+    判重三级：① dedupe_key 命中 ② 历史 NULL 行反解键命中 ③ 无命中 → INSERT。
+    并发兜底：唯一索引 + savepoint 捕获 IntegrityError 计跳过。
+    """
+    inserted = 0
+    skipped = 0
+    period = str(result.get("period") or "")
+    existing = {
+        k for (k,) in db.query(KPIAlert.dedupe_key).filter(
+            KPIAlert.entity_id == entity_id,
+            KPIAlert.kpi_id == kpi_id,
+            KPIAlert.alert_type == ALERT_TYPE,
+            KPIAlert.dedupe_key.isnot(None),
+        ).all() if k
+    }
+    existing.update(_legacy_keys(db, entity_id, kpi_id))
+    for h in hits(result):
+        msg = h.get("alert_message") or ""
+        if not msg:
+            continue
+        dk = _item_dedupe_key(h, period)
+        if dk is not None:
+            if dk in existing:
+                skipped += 1
+                continue
+        else:
+            # 无键（理论上不会发生）：退回逐字判重（旧行为）
+            dup = db.query(KPIAlert.id).filter(
+                KPIAlert.entity_id == entity_id,
+                KPIAlert.kpi_id == kpi_id,
+                KPIAlert.alert_type == ALERT_TYPE,
+                KPIAlert.alert_message == msg,
+            ).first()
+            if dup:
+                skipped += 1
+                continue
+        row = KPIAlert(
+            entity_id=entity_id,
+            kpi_id=kpi_id,
+            kpi_value_id=None,
+            alert_level=h.get("level") or "yellow",
+            alert_message=msg,
+            alert_type=ALERT_TYPE,
+            dedupe_key=dk,
+            suggestion=h.get("suggestion") or None,
+            status="pending",
+            assignee=h.get("assignee") or None,
+        )
+        sp = db.begin_nested()
+        try:
+            db.add(row)
+            db.flush()
+            sp.commit()
+            inserted += 1
+            if dk:
+                existing.add(dk)
+        except IntegrityError:
+            sp.rollback()
+            skipped += 1
+    if inserted:
+        db.commit()
+    return {"inserted": inserted, "skipped_duplicate": skipped}

backend/app/utils/ar_caliber.py

@@ -0,0 +1,71 @@
+"""应收账龄口径单一事实源（`hanke_ar_aging`）
+
+财务已定死的口径纪律（ERRATA-02 §五 / DQ 台账 DQ-AR-03）：
+
+1. **三口径分列，禁跨口径合计**：`全户口径` / `借方应收户口径` / `贷方户单列(净额)`。
+   字面值逐字匹配——`贷方户单列(净额)` 写成 `贷方户单列` 会漏 42 行。
+2. **对外「超期金额」只取 `借方应收户口径`**；贷方户单列不进分母，任何 `caliber=all` 请求 400。
+3. **六档互斥**：0-30 / 31-60 / 61-90 / 90-120 / 120-150 / 150+，逐行六档和 ≡ `total_amt`。
+4. `opening_amt`（期初）**只展示**，禁作账龄基数、禁进任何分子分母。
+5. 负值保留：禁 `ABS()` / `COALESCE(...,0)` 抹平（DQ-AR-04 杨创 `b_0_30=-20`）。
+"""
+from typing import Optional, Tuple
+
+# 表名（DDL：dama/_reports/prodwrite-cma-hanke-ar-aging-20261020/mirror/03_ddl.sql）
+AR_AGING_TABLE = "hanke_ar_aging"
+
+# assignee 派生依赖的只读维度/事实表（§6.2）。三者均为**只读 SELECT**，不写不建。
+ZXT_MASTERBILL_TABLE = "zxt_masterbill"   # 智享通单据流（unit_id / emp_id / bill_date / abst）
+EMPLOYEE_DICT_TABLE = "employee_dict"     # 员工字典（id → u_Name）；当前库内**未落库**
+UNITS_DICT_TABLE = "units_dict"           # 往来单位字典（unit_code → unit_id）；当前库内**未落库**
+
+# 三口径字面值（逐字匹配，勿改写）
+CALIBER_FULL = "全户口径"
+CALIBER_DEBIT = "借方应收户口径"
+CALIBER_CREDIT_SEPARATE = "贷方户单列(净额)"
+CALIBERS: Tuple[str, ...] = (CALIBER_FULL, CALIBER_DEBIT, CALIBER_CREDIT_SEPARATE)
+
+# 对外超期口径/看板默认口径
+DEFAULT_CALIBER = CALIBER_DEBIT
+
+# 跨口径合计请求字面值 —— 一律 400
+CALIBER_ALL = "all"
+
+# 六档列（顺序即展示顺序）：(列名, 档位标签, 档位下界天数)
+BUCKETS: Tuple[Tuple[str, str, int], ...] = (
+    ("b_0_30", "0-30", 0),
+    ("b_31_60", "31-60", 31),
+    ("b_61_90", "61-90", 61),
+    ("b_90_120", "90-120", 91),
+    ("b_120_150", "120-150", 121),
+    ("b_150p", "150+", 151),
+)
+
+BUCKET_COLUMNS = tuple(b[0] for b in BUCKETS)
+# 逾期档位（>30 天）列；最老档下界天数即「逾期 N 天以上」的 N
+OVERDUE_BUCKETS: Tuple[Tuple[str, int], ...] = tuple((b[0], b[2]) for b in BUCKETS if b[2] > 0)
+# 90 天以上（含 90-120 档）
+NINETY_PLUS_COLUMNS = ("b_90_120", "b_120_150", "b_150p")
+
+
+class CaliberError(ValueError):
+    """口径入参非法（未知值 / 跨口径合计）。"""
+
+
+def validate_caliber(caliber: Optional[str], *, allow_none: bool = False) -> str:
+    """校验口径字面值；`all` 与未知值一律拒绝（禁跨口径合计）。"""
+    if caliber is None or caliber == "":
+        if allow_none:
+            return DEFAULT_CALIBER
+        raise CaliberError("caliber 不能为空")
+    if caliber == CALIBER_ALL:
+        raise CaliberError("禁止跨口径合计（caliber=all 不提供）：三口径必须分列查询")
+    if caliber not in CALIBERS:
+        raise CaliberError(
+            "未知口径 %r；仅接受字面值：%s" % (caliber, " / ".join(CALIBERS))
+        )
+    return caliber
+
+
+def is_credit_separate(caliber: str) -> bool:
+    return caliber == CALIBER_CREDIT_SEPARATE

backend/scripts/ddl_kpi_alerts_dedupe.py

@@ -0,0 +1,113 @@
+#!/usr/bin/env python3
+"""kpi_alerts 去重键结构变更（幂等执行器，A2 方案）
+
+用法：
+  ddl_kpi_alerts_dedupe.py --check     # 只读：打印列/索引是否存在（默认）
+  ddl_kpi_alerts_dedupe.py --apply     # 幂等执行：缺列则加列，缺索引则建索引
+  ddl_kpi_alerts_dedupe.py --rollback  # 回退：删索引 + 删列（数据无需恢复：本改造无 UPDATE/DELETE）
+
+特性：
+  * information_schema 判存在后再执行 ⇒ 重复跑安全（第二次全 NOOP）
+  * 只加可空列 + 唯一索引；历史行 dedupe_key 全 NULL ⇒ 建索引必然无冲突
+  * root 本机 socket 连接（与 migrate_map_connections.py 同口径），不读 .env
+"""
+from __future__ import annotations
+
+import argparse
+import sys
+
+from sqlalchemy import create_engine, text
+
+URL = "mysql+pymysql://root@localhost/cma?unix_socket=/var/run/mysqld/mysqld.sock&charset=utf8mb4"
+TABLE = "kpi_alerts"
+COLUMN = "dedupe_key"
+INDEX = "uk_kpi_alerts_dedupe"
+
+ADD_COLUMN_SQL = (
+    "ALTER TABLE kpi_alerts "
+    "ADD COLUMN dedupe_key VARCHAR(160) NULL "
+    "COMMENT '生成侧幂等键 period|scope|ident，仅 alert_type=overdue 填充；历史行为 NULL' "
+    "AFTER alert_type"
+)
+ADD_INDEX_SQL = (
+    "CREATE UNIQUE INDEX uk_kpi_alerts_dedupe "
+    "ON kpi_alerts (entity_id, kpi_id, alert_type, dedupe_key)"
+)
+DROP_INDEX_SQL = "DROP INDEX uk_kpi_alerts_dedupe ON kpi_alerts"
+DROP_COLUMN_SQL = "ALTER TABLE kpi_alerts DROP COLUMN dedupe_key"
+
+
+def col_exists(eng) -> bool:
+    with eng.connect() as c:
+        n = c.execute(text(
+            "SELECT COUNT(*) FROM information_schema.columns "
+            "WHERE table_schema=DATABASE() AND table_name=:t AND column_name=:c"
+        ), {"t": TABLE, "c": COLUMN}).scalar()
+    return int(n or 0) > 0
+
+
+def idx_exists(eng) -> bool:
+    with eng.connect() as c:
+        n = c.execute(text(
+            "SELECT COUNT(*) FROM information_schema.statistics "
+            "WHERE table_schema=DATABASE() AND table_name=:t AND index_name=:i"
+        ), {"t": TABLE, "i": INDEX}).scalar()
+    return int(n or 0) > 0
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser()
+    g = ap.add_mutually_exclusive_group()
+    g.add_argument("--check", action="store_true", help="只读检查（默认）")
+    g.add_argument("--apply", action="store_true", help="幂等执行")
+    g.add_argument("--rollback", action="store_true", help="回退结构")
+    a = ap.parse_args()
+    if not (a.check or a.apply or a.rollback):
+        a.check = True
+
+    eng = create_engine(URL, future=True)
+    col, idx = col_exists(eng), idx_exists(eng)
+    print("[state] %s.%s=%s  index %s=%s" % (TABLE, COLUMN, col, INDEX, idx))
+
+    if a.check:
+        print("[check] %s" % ("结构就绪（无操作）" if (col and idx) else
+                              ("缺列" if not col else "") + ("缺索引" if not idx else "")))
+        return 0
+
+    if a.rollback:
+        with eng.begin() as c:
+            if idx:
+                c.execute(text(DROP_INDEX_SQL))
+                print("[rollback] DROP INDEX %s ✅" % INDEX)
+            if col:
+                c.execute(text(DROP_COLUMN_SQL))
+                print("[rollback] DROP COLUMN %s ✅" % COLUMN)
+        print("[state] 回退后 col=%s idx=%s" % (col_exists(eng), idx_exists(eng)))
+        return 0
+
+    # --apply
+    done = []
+    if not col:
+        with eng.begin() as c:
+            c.execute(text(ADD_COLUMN_SQL))
+        done.append("ADD COLUMN %s" % COLUMN)
+        print("[apply] ADD COLUMN %s ✅" % COLUMN)
+    else:
+        print("[apply] column %s 已存在 → NOOP" % COLUMN)
+    if not idx:
+        with eng.begin() as c:
+            c.execute(text(ADD_INDEX_SQL))
+        done.append("CREATE INDEX %s" % INDEX)
+        print("[apply] CREATE UNIQUE INDEX %s ✅" % INDEX)
+    else:
+        print("[apply] index %s 已存在 → NOOP" % INDEX)
+
+    col2, idx2 = col_exists(eng), idx_exists(eng)
+    print("[state] 执行后 col=%s idx=%s" % (col2, idx2))
+    payload = {"changed": done, "column_exists": col2, "index_exists": idx2}
+    print("[result] %s" % payload)
+    return 0 if (col2 and idx2) else 5
+
+
+if __name__ == "__main__":
+    sys.exit(main())

backend/scripts/migrate_map_connections.py

@@ -0,0 +1,524 @@
+#!/usr/bin/env python3
+"""canvas_data.connections → map_connections 迁移（P1 四层因果结构化）
+
+任务：cma-map-connections-table-p1 / project-20261021-cma-map-connections-table-p1-fullstack
+
+设计要点
+--------
+1. **只搬不删**：canvas_data 原样保留（双写期，前端渲染仍读 JSON）。本脚本对
+   strategic_maps 只做 SELECT，任何情况下不 UPDATE。
+2. **幂等**：目标表唯一键 (map_id, from_node, to_node)。重跑 `--apply` 只补缺，
+   已存在的行 SKIP；`--idem` 为只读等价复跑（断言全部已在位），零写。
+3. **双格式兼容**（生产实况）：
+      `dim:中文目标名`  → 按 (dimension_key, 目标名) 解析 map_objectives.id
+      `dim-序号`        → 按 dimensions[dim].objectives 顺序解析（缺 dimensions 时退 sort_order）
+   原始串原样落 from_node/to_node，解析结果落 *_layer / *_objective_id。
+4. **四态**：--check（默认拒绝无标志调用）/ --apply / --idem / --rollback
+   写库通道：生产 apply/rollback = root over unix socket（cma_api 无新表 DML 授权）；
+   --check 走环境变量只读账号（cma_reader/cma_api 的 SELECT）。
+
+用法
+----
+    python3 migrate_map_connections.py --check   [--json-out PATH]
+    python3 migrate_map_connections.py --show-create
+    python3 migrate_map_connections.py --apply   [--backup-dir DIR] [--evidence PATH]
+    python3 migrate_map_connections.py --idem
+    python3 migrate_map_connections.py --rollback <apply-evidence.json>
+
+退出码：0 成功 ｜ 2 用法/缺标志 ｜ 3 表不存在（先走 db_migrate.py --apply）｜ 4 画布 JSON 不可解析（fail-closed）｜ 5 自检不一致
+"""
+import argparse
+import hashlib
+import json
+import os
+import re
+import sys
+from datetime import datetime
+
+HERE = os.path.dirname(os.path.abspath(__file__))
+sys.path.insert(0, os.path.dirname(HERE))
+
+LAYERS = ("learning", "process", "customer", "finance")
+LAYER_ORDER = {k: i for i, k in enumerate(LAYERS)}
+NODE_COLON = re.compile(r"^([a-z_]+):(.+)$")
+NODE_INDEX = re.compile(r"^([a-z_]+)-(\d+)$")
+
+
+# ── 连接（凭据只从 env 读，不回显）────────────────────────────────
+def _mask(url: str) -> str:
+    return re.sub(r"://([^:/@]+):[^@]*@", r"://\1:***@", url)
+
+
+def build_engine(db_url: str | None, mode: str):
+    from sqlalchemy import create_engine
+
+    if db_url:
+        url = db_url
+    elif mode == "check" or mode == "show-create":
+        user = os.getenv("CMA_DB_USER", "cma_reader")
+        pwd = os.getenv("CMA_DB_PASS", "")
+        host = os.getenv("CMA_DB_HOST", "127.0.0.1")
+        port = os.getenv("CMA_DB_PORT", "3306")
+        name = os.getenv("CMA_DB_NAME", "cma")
+        if not pwd:
+            print("[fail] 缺少 CMA_DB_PASS（只读预检需要 SELECT 账号凭据）", file=sys.stderr)
+            sys.exit(2)
+        url = f"mysql+pymysql://{user}:{pwd}@{host}:{port}/{name}?charset=utf8mb4"
+    else:
+        name = os.getenv("CMA_DB_NAME", "cma")
+        url = f"mysql+pymysql://root@localhost/{name}?unix_socket=/var/run/mysqld/mysqld.sock&charset=utf8mb4"
+    return create_engine(url, future=True), url
+
+
+def load_dotenv():
+    try:
+        from dotenv import load_dotenv as _ld
+        _ld(os.path.join(os.path.dirname(HERE), ".env"))
+    except Exception:
+        pass
+
+
+# ── JSON 规范化 ────────────────────────────────────────────────
+def as_json(v):
+    """DB JSON 列可能是 dict/list 或 str（SQLite/部分驱动）。"""
+    if isinstance(v, (bytes, bytearray)):
+        v = v.decode("utf-8")
+    if isinstance(v, str):
+        return json.loads(v)
+    return v
+
+
+def canvas_md5(canvas) -> str:
+    raw = json.dumps(canvas, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
+    return hashlib.md5(raw.encode("utf-8")).hexdigest()
+
+
+def norm_name(s: str) -> str:
+    return re.sub(r"\s+", " ", (s or "").strip())
+
+
+# ── 节点解析 ───────────────────────────────────────────────────
+def parse_node(node):
+    """→ (layer|None, ref|None, err|None)
+
+    ref = {"kind": "name", "name": ...} | {"kind": "index", "index": n}
+    """
+    if not isinstance(node, str) or not node.strip():
+        return None, None, "node_empty"
+    s = node.strip()
+    m = NODE_COLON.match(s)
+    if m:
+        layer = m.group(1)
+        name = m.group(2).strip()
+        if layer not in LAYERS:
+            return None, None, f"unknown_layer:{layer}"
+        if not name:
+            return None, None, "node_name_empty"
+        return layer, {"kind": "name", "name": name}, None
+    m = NODE_INDEX.match(s)
+    if m:
+        layer = m.group(1)
+        if layer not in LAYERS:
+            return None, None, f"unknown_layer:{layer}"
+        return layer, {"kind": "index", "index": int(m.group(2))}, None
+    return None, None, "node_format_unrecognized"
+
+
+class ObjectiveIndex:
+    """按 map 建索引：名字 → objective_id；层 + 序号 → objective_id。"""
+
+    def __init__(self, dims_by_map, objs_by_map):
+        self.by_name = {}
+        self.by_order = {}
+        self.dims = dims_by_map
+        for map_id, objs in objs_by_map.items():
+            name_idx, order_idx = {}, {}
+            for o in objs:
+                name_idx[(o["dimension_key"], norm_name(o["name"]))] = o["id"]
+                order_idx.setdefault(o["dimension_key"], []).append(o)
+            for k in order_idx:
+                order_idx[k].sort(key=lambda o: (o.get("sort_order") or 0, o["id"]))
+            self.by_name[map_id] = name_idx
+            self.by_order[map_id] = order_idx
+
+    def resolve(self, map_id, layer, ref):
+        """→ (objective_id|None, note|None)"""
+        if ref["kind"] == "name":
+            hit = self.by_name.get(map_id, {}).get((layer, norm_name(ref["name"])))
+            if hit:
+                return hit, None
+            # 仅大小写/空白差异的近似命中：报告但不静默
+            want = norm_name(ref["name"]).replace(" ", "")
+            for (lk, ln), oid in self.by_name.get(map_id, {}).items():
+                if lk == layer and ln.replace(" ", "") == want:
+                    return oid, "normalized_match"
+            return None, "objective_unresolved"
+        # index：优先按 dimensions[dim].objectives 顺序（前端 key 就是该数组下标）
+        idx = ref["index"]
+        dims = self.dims.get(map_id) or []
+        for d in dims if isinstance(dims, list) else []:
+            if d.get("key") == layer:
+                objs = d.get("objectives") or []
+                if 0 <= idx < len(objs):
+                    name = norm_name(objs[idx].get("name"))
+                    hit = self.by_name.get(map_id, {}).get((layer, name))
+                    if hit:
+                        return hit, None
+                    return None, "objective_unresolved"
+                return None, "index_out_of_range"
+        # dimensions 缺失 → 退 sort_order
+        cand = self.by_order.get(map_id, {}).get(layer) or []
+        for o in cand:
+            if (o.get("sort_order") or 0) == idx:
+                return o["id"], "resolved_by_sort_order"
+        if 0 <= idx < len(cand):
+            return cand[idx]["id"], "resolved_by_position"
+        return None, "index_out_of_range"
+
+
+# ── 计划构建（纯读）────────────────────────────────────────────
+def load_state(session):
+    from sqlalchemy import inspect
+    from app.models import StrategicMap, MapObjective, MapConnection
+
+    insp = inspect(session.get_bind())
+    has_table = insp.has_table(MapConnection.__tablename__)
+
+    maps = []
+    for m in session.query(StrategicMap).order_by(StrategicMap.id).all():
+        maps.append({
+            "id": m.id, "entity_id": m.entity_id, "title": m.title, "status": m.status,
+            "canvas": as_json(m.canvas_data), "dimensions": as_json(m.dimensions),
+        })
+
+    objs = {}
+    for o in session.query(MapObjective).order_by(MapObjective.map_id, MapObjective.sort_order, MapObjective.id).all():
+        objs.setdefault(o.map_id, []).append({
+            "id": o.id, "dimension_key": o.dimension_key, "name": o.name, "sort_order": o.sort_order,
+        })
+
+    existing = {}
+    if has_table:
+        for c in session.query(MapConnection).all():
+            existing[(c.map_id, c.from_node, c.to_node)] = c.id
+    return maps, objs, existing, has_table
+
+
+def build_plan(maps, objs):
+    oi = ObjectiveIndex({m["id"]: m["dimensions"] for m in maps}, objs)
+    plan, per_map, anomalies = [], [], []
+
+    def note(map_id, idx, kind, node, detail=""):
+        anomalies.append({"map_id": map_id, "canvas_index": idx, "kind": kind,
+                          "node": node, "detail": detail})
+
+    for m in maps:
+        canvas = m["canvas"]
+        conns = []
+        if canvas is None:
+            canvas = {}
+        if not isinstance(canvas, dict):
+            note(m["id"], None, "canvas_not_object", None, f"type={type(canvas).__name__}")
+            per_map.append({"map_id": m["id"], "entity_id": m["entity_id"], "title": m["title"],
+                            "canvas_connections": None, "plan_rows": 0, "anomalies": 1})
+            continue
+        conns = canvas.get("connections") or []
+        if not isinstance(conns, list):
+            note(m["id"], None, "connections_not_list", None, f"type={type(conns).__name__}")
+            conns = []
+
+        rows, seen = [], {}
+        for i, c in enumerate(conns):
+            if not isinstance(c, dict):
+                note(m["id"], i, "connection_not_object", None, f"type={type(c).__name__}")
+                continue
+            frm, to = c.get("from"), c.get("to")
+            if not isinstance(frm, str) or not isinstance(to, str) or not frm or not to:
+                note(m["id"], i, "endpoint_missing", f"{frm!r}->{to!r}", "from/to 缺失或非字符串")
+                continue
+            key = (frm, to)
+            if key in seen:
+                note(m["id"], i, "duplicate_edge", f"{frm} -> {to}", f"与 index={seen[key]} 重复（唯一键会合并）")
+                continue
+            seen[key] = i
+
+            fl, fref, ferr = parse_node(frm)
+            tl, tref, terr = parse_node(to)
+            if ferr:
+                note(m["id"], i, ferr, frm, "from 解析失败")
+            if terr:
+                note(m["id"], i, terr, to, "to 解析失败")
+            if frm == to:
+                note(m["id"], i, "self_loop", frm, "自连")
+            if fl and tl and LAYER_ORDER[tl] < LAYER_ORDER[fl]:
+                note(m["id"], i, "reverse_direction", f"{frm} -> {to}",
+                     f"{fl}→{tl} 与 learning→process→customer→finance 相反（告警不拦截）")
+
+            f_obj = f_note = t_obj = t_note = None
+            if fref:
+                f_obj, f_note = oi.resolve(m["id"], fl, fref)
+                if f_note and f_note != "normalized_match":
+                    note(m["id"], i, f_note, frm, "from 目标未解析")
+            if tref:
+                t_obj, t_note = oi.resolve(m["id"], tl, tref)
+                if t_note and t_note != "normalized_match":
+                    note(m["id"], i, t_note, to, "to 目标未解析")
+
+            rows.append({
+                "entity_id": m["entity_id"] if m["entity_id"] is not None else 1,
+                "map_id": m["id"], "from_node": frm, "to_node": to,
+                "from_layer": fl, "to_layer": tl,
+                "from_objective_id": f_obj, "to_objective_id": t_obj,
+                "effect": c.get("effect") or "positive",
+                "label": c.get("label") or None,
+                "sort_order": i,
+            })
+
+        plan.extend(rows)
+        per_map.append({
+            "map_id": m["id"], "entity_id": m["entity_id"], "title": m["title"],
+            "status": m["status"], "canvas_connections": len(conns), "plan_rows": len(rows),
+            "anomalies": sum(1 for a in anomalies if a["map_id"] == m["id"]),
+        })
+    return plan, per_map, anomalies
+
+
+def summarize(plan, per_map, anomalies, existing=None):
+    by_pair, by_layer = {}, {}
+    for r in plan:
+        pair = f"{r['from_layer'] or '?'}->{r['to_layer'] or '?'}"
+        by_pair[pair] = by_pair.get(pair, 0) + 1
+        for lay in (r["from_layer"], r["to_layer"]):
+            if lay:
+                by_layer[lay] = by_layer.get(lay, 0) + 1
+    existing = existing or {}
+    to_insert = sum(1 for r in plan if (r["map_id"], r["from_node"], r["to_node"]) not in existing)
+    return {
+        "canvas_connections": sum(p["canvas_connections"] or 0 for p in per_map),
+        "plan_rows": len(plan),
+        "already_in_table": len([r for r in plan if (r["map_id"], r["from_node"], r["to_node"]) in existing]),
+        "to_insert": to_insert,
+        "unresolved_layer": sum(1 for r in plan if not r["from_layer"] or not r["to_layer"]),
+        "unresolved_objective": sum(1 for r in plan if not r["from_objective_id"] or not r["to_objective_id"]),
+        "anomalies": len(anomalies),
+        "layer_pair_distribution": by_pair,
+        "layer_node_refs": by_layer,
+    }
+
+
+# ── 三种模式的实现 ─────────────────────────────────────────────
+def cmd_check(engine, args):
+    from sqlalchemy.orm import Session
+    with Session(engine) as s:
+        maps, objs, existing, has_table = load_state(s)
+    plan, per_map, anomalies = build_plan(maps, objs)
+    report = {
+        "generated_at": datetime.now().isoformat(timespec="seconds"),
+        "mode": "check", "read_only": True,
+        "db": _mask(engine.url.render_as_string(hide_password=True).replace("***", "***")),
+        "engine_dialect": engine.dialect.name,
+        "table_exists": has_table,
+        "totals": summarize(plan, per_map, anomalies, existing),
+        "maps": per_map,
+        "anomalies": anomalies,
+        "sample_rows": plan[:5],
+    }
+    if args.json_out:
+        os.makedirs(os.path.dirname(os.path.abspath(args.json_out)), exist_ok=True)
+        with open(args.json_out, "w", encoding="utf-8") as f:
+            json.dump(report, f, ensure_ascii=False, indent=2)
+    print(f"[check] 库={report['db']}  方言={report['engine_dialect']}  目标表存在={has_table}")
+    t = report["totals"]
+    print(f"[check] 画布连线={t['canvas_connections']}  计划行={t['plan_rows']}  "
+          f"已在表={t['already_in_table']}  待插入={t['to_insert']}")
+    print(f"[check] 层未解析={t['unresolved_layer']}  目标未解析={t['unresolved_objective']}  异常条目={t['anomalies']}")
+    print(f"[check] 层对分布={t['layer_pair_distribution']}")
+    for p in per_map:
+        print(f"[check]   map {p['map_id']} (entity={p['entity_id']}, {p['status']}): "
+              f"canvas={p['canvas_connections']} → plan={p['plan_rows']}  异常={p['anomalies']}")
+    for a in anomalies[:50]:
+        print(f"[check]   ✗ map={a['map_id']} idx={a['canvas_index']} {a['kind']} {a['node']} {a['detail']}")
+    if not has_table:
+        print("[check] ⚠ 目标表不存在：建表须先走 db_migrate.py --apply（受批面）")
+    return 0
+
+
+def cmd_show_create():
+    from sqlalchemy.schema import CreateTable
+    from sqlalchemy.dialects import mysql
+    from app.models import MapConnection
+    print(CreateTable(MapConnection.__table__).compile(dialect=mysql.dialect()))
+    return 0
+
+
+def cmd_apply(engine, args):
+    from sqlalchemy.orm import Session
+    from sqlalchemy.exc import IntegrityError
+    from app.models import MapConnection
+
+    with Session(engine) as s:
+        maps, objs, existing, has_table = load_state(s)
+    if not has_table:
+        print("[fail] map_connections 表不存在 → 先执行（受批）db_migrate.py --apply", file=sys.stderr)
+        return 3
+    plan, per_map, anomalies = build_plan(maps, objs)
+    hard = [a for a in anomalies if a["kind"] in ("canvas_not_object", "connections_not_list", "connection_not_object")]
+    if hard:
+        print(f"[fail] 画布结构不可解析 {len(hard)} 处 → fail-closed，未写任何行："
+              f"{hard[:3]}", file=sys.stderr)
+        return 4
+
+    # 回滚点：canvas_data 全量快照 + md5（只读 SELECT 得到）
+    ts = datetime.now().strftime("%Y%m%d-%H%M%S")
+    backup_dir = os.path.abspath(args.backup_dir or f"/srv/hermes/backups/map_connections_p1/{ts}")
+    os.makedirs(backup_dir, exist_ok=True)
+    backup_path = os.path.join(backup_dir, "canvas_data_snapshot.json")
+    with open(backup_path, "w", encoding="utf-8") as f:
+        json.dump([{"map_id": m["id"], "entity_id": m["entity_id"], "title": m["title"],
+                    "canvas_data": m["canvas"], "md5": canvas_md5(m["canvas"])} for m in maps],
+                  f, ensure_ascii=False, indent=2)
+    pre_md5 = {m["id"]: canvas_md5(m["canvas"]) for m in maps}
+
+    inserted, skipped, errors = [], 0, []
+    with Session(engine) as s:
+        try:
+            for r in plan:
+                key = (r["map_id"], r["from_node"], r["to_node"])
+                if key in existing:
+                    skipped += 1
+                    continue
+                obj = MapConnection(**r)
+                s.add(obj)
+                try:
+                    s.flush()
+                except IntegrityError:
+                    s.rollback()
+                    skipped += 1
+                    continue
+                inserted.append({"id": obj.id, "map_id": r["map_id"],
+                                 "from_node": r["from_node"], "to_node": r["to_node"]})
+            s.commit()
+        except Exception as exc:
+            s.rollback()
+            print(f"[fail] 写入失败已回滚：{type(exc).__name__}: {exc}", file=sys.stderr)
+            return 1
+
+    # 写后自检：行数 + canvas 未变
+    with Session(engine) as s:
+        maps2, _, _, _ = load_state(s)
+        table_rows = s.query(MapConnection).filter(
+            MapConnection.map_id.in_([m["id"] for m in maps])).count() if maps else 0
+    post_md5 = {m["id"]: canvas_md5(m["canvas"]) for m in maps2}
+    canvas_unchanged = pre_md5 == post_md5
+    expected = len({(r["map_id"], r["from_node"], r["to_node"]) for r in plan})
+
+    evidence = {
+        "action": "apply", "status": "ok",
+        "started_at": ts, "finished_at": datetime.now().isoformat(timespec="seconds"),
+        "db_engine": engine.dialect.name,
+        "backup_path": backup_path, "backup_pre_md5": pre_md5, "post_md5": post_md5,
+        "canvas_unchanged": canvas_unchanged,
+        "plan_rows": len(plan), "inserted": len(inserted), "skipped_existing": skipped,
+        "table_rows_for_scope": table_rows, "expected_distinct_edges": expected,
+        "anomalies": anomalies, "maps": per_map, "inserted_ids": inserted,
+    }
+    ok = canvas_unchanged and table_rows == expected and not errors
+    evidence["verdict"] = "PASS" if ok else "FAIL"
+    ev_path = os.path.abspath(args.evidence or os.path.join(backup_dir, "apply_evidence.json"))
+    os.makedirs(os.path.dirname(ev_path), exist_ok=True)
+    with open(ev_path, "w", encoding="utf-8") as f:
+        json.dump(evidence, f, ensure_ascii=False, indent=2)
+
+    print(f"[apply] 计划 {len(plan)} 行；新插入 {len(inserted)}，已存在跳过 {skipped}")
+    print(f"[apply] 表内本范围行数 {table_rows} / 期望 {expected}；canvas 未改={canvas_unchanged}")
+    print(f"[apply] 备份 {backup_path}")
+    print(f"[apply] 证据 {ev_path}")
+    print(f"[apply] verdict={evidence['verdict']}")
+    return 0 if ok else 5
+
+
+def cmd_idem(engine, args):
+    from sqlalchemy.orm import Session
+    with Session(engine) as s:
+        maps, objs, existing, has_table = load_state(s)
+        if not has_table:
+            print("[fail] 表不存在", file=sys.stderr)
+            return 3
+        plan, per_map, anomalies = build_plan(maps, objs)
+        mismatches = []
+        for r in plan:
+            key = (r["map_id"], r["from_node"], r["to_node"])
+            cid = existing.get(key)
+            if not cid:
+                mismatches.append({"kind": "missing", "key": list(key)})
+                continue
+            row = s.get(__import__("app.models", fromlist=["MapConnection"]).MapConnection, cid)
+            for f in ("from_layer", "to_layer", "from_objective_id", "to_objective_id"):
+                if (getattr(row, f) or None) != (r[f] or None):
+                    mismatches.append({"kind": "value_diff", "key": list(key), "field": f,
+                                       "db": getattr(row, f), "plan": r[f]})
+    print(f"[idem] 计划 {len(plan)} 行，已在位 {len(plan) - len([m for m in mismatches if m['kind'] == 'missing'])}；"
+          f"差异 {len(mismatches)}")
+    for m in mismatches[:20]:
+        print(f"[idem]   ✗ {m}")
+    print(f"[idem] verdict={'PASS' if not mismatches else 'FAIL'}")
+    return 0 if not mismatches else 5
+
+
+def cmd_rollback(engine, args):
+    from sqlalchemy.orm import Session
+    from app.models import MapConnection
+    path = args.rollback
+    if not path or not os.path.exists(path):
+        print(f"[fail] 回滚证据不存在：{path}", file=sys.stderr)
+        return 2
+    with open(path, encoding="utf-8") as f:
+        ev = json.load(f)
+    ids = [x["id"] for x in ev.get("inserted_ids", [])]
+    if not ids:
+        print("[rollback] 证据里没有 inserted_ids，无需回滚")
+        return 0
+    with Session(engine) as s:
+        n = s.query(MapConnection).filter(MapConnection.id.in_(ids)).delete(synchronize_session=False)
+        s.commit()
+        left = s.query(MapConnection).filter(MapConnection.id.in_(ids)).count()
+    print(f"[rollback] 删除 {n} 行，残留 {left}（证据 {path}）；canvas_data 本脚本从未写入，无需还原")
+    return 0 if left == 0 else 5
+
+
+def main():
+    ap = argparse.ArgumentParser(description="canvas_data.connections → map_connections（四态幂等迁移）")
+    g = ap.add_mutually_exclusive_group()
+    g.add_argument("--check", action="store_true", help="只读预检（dry-run）：行数/异常/待插入")
+    g.add_argument("--apply", action="store_true", help="唯一写：幂等插入（需表已存在）")
+    g.add_argument("--idem", action="store_true", help="只读幂等复跑：断言全部已在位")
+    g.add_argument("--rollback", metavar="EVIDENCE_JSON", help="按 apply 证据删除本批插入行")
+    g.add_argument("--show-create", action="store_true", help="打印模型编译出的 CREATE TABLE")
+    ap.add_argument("--db-url", help="覆盖连接串（演练副本/SQLite 用）")
+    ap.add_argument("--json-out", help="--check 的机器可读报告路径")
+    ap.add_argument("--backup-dir", help="--apply 回滚点目录")
+    ap.add_argument("--evidence", help="--apply 证据 JSON 路径")
+    args = ap.parse_args()
+
+    load_dotenv()
+
+    if args.show_create:
+        return cmd_show_create()
+    if not any([args.check, args.apply, args.idem, args.rollback]):
+        print("拒绝无标志调用：请显式给 --check / --apply / --idem / --rollback", file=sys.stderr)
+        return 2
+
+    mode = "check" if args.check else "apply" if args.apply else "idem" if args.idem else "rollback"
+    engine, url = build_engine(args.db_url, mode)
+    print(f"[{mode}] 通道 {_mask(url)}")
+    if args.check:
+        return cmd_check(engine, args)
+    if args.apply:
+        return cmd_apply(engine, args)
+    if args.idem:
+        return cmd_idem(engine, args)
+    return cmd_rollback(engine, args)
+
+
+if __name__ == "__main__":
+    sys.exit(main())

backend/scripts/migrate_strategy_charter_20261021.py

@@ -0,0 +1,196 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""战略章程表迁移（strategy_charter）— 幂等建表 + 回滚，零示例数据
+
+任务：`project-20261021-cma-strategy-charter-p0-fullstack`（P0 战略基座）
+
+用法：
+  python3 migrate_strategy_charter_20261021.py --check    # 只读预检（默认）：表/列/唯一键是否齐
+  python3 migrate_strategy_charter_20261021.py --apply    # 建表（CREATE TABLE IF NOT EXISTS）+ 回读断言
+  python3 migrate_strategy_charter_20261021.py --down     # 回滚（先 mysqldump 备份，再 DROP；有数据需 --force）
+
+连接（与 db_migrate.py / treasury_migrate.py 同口径）：
+  env CMA_MIGRATE_DB_USER / CMA_MIGRATE_DB_PASS / CMA_MIGRATE_DB_HOST / CMA_MIGRATE_DB_PORT 优先（TCP）；
+  否则本机 unix socket + root（/var/run/mysqld/mysqld.sock）。
+  **显式拒绝**应用运行账号 cma_api（无 DDL/GRANT 权限）。
+
+纪律：
+  · 迁移脚本**零 INSERT** —— 不写任何示例/默认章程文案（任总红线：真实数据非 Demo）；
+  · 幂等：二跑 created=0；列级差异只报告，不改既有列；
+  · 授权面不在本脚本：由部署链 [4b-2] grant_surface_sync 按白名单授予 cma_api INSERT,UPDATE。
+"""
+import argparse
+import json
+import os
+import subprocess
+import sys
+from datetime import datetime
+
+TABLE = "strategy_charter"
+NEW_COLUMNS = [
+    "id", "entity_id", "mission", "vision", "vision_year", "vision_measurable",
+    "values_text", "value_items", "updated_by", "updated_by_name", "created_at", "updated_at",
+]
+UNIQUE_KEY = "uq_strategy_charter_entity"
+
+DDL = """
+CREATE TABLE IF NOT EXISTS `strategy_charter` (
+  `id` INT NOT NULL AUTO_INCREMENT,
+  `entity_id` INT NOT NULL COMMENT '企业ID（账套，多租户隔离）',
+  `mission` TEXT NOT NULL COMMENT '使命：我们为何存在（做什么、为谁做）',
+  `vision` TEXT NOT NULL COMMENT '愿景：要成为什么',
+  `vision_year` VARCHAR(20) NOT NULL COMMENT '愿景达成年份 YYYY（不可空）',
+  `vision_measurable` TEXT NOT NULL COMMENT '可验证状态：可核对规模/业务口径描述（不可空）',
+  `values_text` TEXT NOT NULL COMMENT '价值观总述（不可交易的底线）',
+  `value_items` JSON NOT NULL COMMENT '价值观条目数组 [{name, meaning}]',
+  `updated_by` INT NULL COMMENT '最后修改人 user.id',
+  `updated_by_name` VARCHAR(50) NULL COMMENT '最后修改人姓名（快照）',
+  `created_at` DATETIME DEFAULT CURRENT_TIMESTAMP,
+  `updated_at` DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+  PRIMARY KEY (`id`),
+  UNIQUE KEY `uq_strategy_charter_entity` (`entity_id`)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='战略章程（使命/愿景/价值观，一账套一行）'
+"""
+
+DB_NAME = os.environ.get("CMA_MIGRATE_DB_NAME", "cma")
+APP_ACCOUNT = os.environ.get("CMA_APP_ACCOUNT", "cma_api")
+
+
+def connect(args):
+    import pymysql
+    if os.environ.get("CMA_MIGRATE_DB_USER") or args.user:
+        user = args.user or os.environ.get("CMA_MIGRATE_DB_USER")
+        if user == APP_ACCOUNT:
+            raise SystemExit("[FAIL] 迁移身份不得为应用运行账号 cma_api（无 DDL 权限）")
+        return pymysql.connect(
+            host=args.host or os.environ.get("CMA_MIGRATE_DB_HOST", "127.0.0.1"),
+            port=int(args.port or os.environ.get("CMA_MIGRATE_DB_PORT", 3306)),
+            user=user,
+            password=args.password if args.password is not None else os.environ.get("CMA_MIGRATE_DB_PASS", ""),
+            database=DB_NAME, charset="utf8mb4", autocommit=True,
+        )
+    socket = os.environ.get("CMA_MIGRATE_DB_SOCKET", "/var/run/mysqld/mysqld.sock")
+    if not os.path.exists(socket):
+        raise SystemExit(f"[FAIL] 无迁移账号环境变量且 socket 不存在: {socket}")
+    return pymysql.connect(unix_socket=socket, user="root", database=DB_NAME,
+                           charset="utf8mb4", autocommit=True)
+
+
+def table_columns(cur):
+    cur.execute(
+        "SELECT COLUMN_NAME, IS_NULLABLE, DATA_TYPE FROM information_schema.COLUMNS "
+        "WHERE TABLE_SCHEMA=%s AND TABLE_NAME=%s ORDER BY ORDINAL_POSITION",
+        (DB_NAME, TABLE))
+    return {r[0]: {"nullable": r[1], "type": r[2]} for r in cur.fetchall()}
+
+
+def unique_keys(cur):
+    cur.execute(
+        "SELECT INDEX_NAME, COLUMN_NAME FROM information_schema.STATISTICS "
+        "WHERE TABLE_SCHEMA=%s AND TABLE_NAME=%s AND NON_UNIQUE=0",
+        (DB_NAME, TABLE))
+    keys = {}
+    for idx, col in cur.fetchall():
+        keys.setdefault(idx, []).append(col)
+    return keys
+
+
+def rel(*parts):
+    return os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), *parts)
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    g = ap.add_mutually_exclusive_group()
+    g.add_argument("--check", action="store_true")
+    g.add_argument("--apply", action="store_true")
+    g.add_argument("--down", action="store_true")
+    ap.add_argument("--force", action="store_true", help="--down 时表内有数据仍删（先 dump）")
+    ap.add_argument("--host")
+    ap.add_argument("--port", type=int)
+    ap.add_argument("--user")
+    ap.add_argument("--password", default=None)
+    args = ap.parse_args()
+    mode = "apply" if args.apply else ("down" if args.down else "check")
+
+    conn = connect(args)
+    out: dict = {"script": "migrate_strategy_charter_20261021.py", "mode": mode, "db": DB_NAME,
+           "table": TABLE, "started_at": datetime.now().isoformat(timespec="seconds")}
+    try:
+        with conn.cursor() as cur:
+            cur.execute("SHOW TABLES LIKE %s", (TABLE,))
+            exists = cur.fetchone() is not None
+            out["table_exists_before"] = exists
+
+            if mode == "check":
+                cols = table_columns(cur) if exists else {}
+                keys = unique_keys(cur) if exists else {}
+                missing = [c for c in NEW_COLUMNS if c not in cols]
+                out.update({
+                    "columns": sorted(cols.keys()),
+                    "missing_columns": missing,
+                    "unique_keys": keys,
+                    "unique_ok": keys.get(UNIQUE_KEY) == ["entity_id"],
+                    "ready": exists and not missing and keys.get(UNIQUE_KEY) == ["entity_id"],
+                    "note": "只读预检，零写入",
+                })
+                print(json.dumps(out, ensure_ascii=False, indent=2))
+                return 0
+
+            if mode == "apply":
+                cur.execute(DDL)
+                cols = table_columns(cur)
+                keys = unique_keys(cur)
+                missing = [c for c in NEW_COLUMNS if c not in cols]
+                cur.execute(f"SELECT COUNT(*) FROM `{TABLE}`")
+                rows = cur.fetchone()[0]
+                ok = (not missing) and keys.get(UNIQUE_KEY) == ["entity_id"]
+                out.update({
+                    "created": 0 if exists else 1,
+                    "columns": sorted(cols.keys()),
+                    "missing_columns": missing,
+                    "unique_keys": keys,
+                    "rows": rows,
+                    "verified": ok,
+                    "note": "零示例数据：本脚本不含任何 INSERT",
+                })
+                print(json.dumps(out, ensure_ascii=False, indent=2))
+                return 0 if ok else 1
+
+            # --down
+            if not exists:
+                out.update({"action": "NOOP", "note": "表不存在，无需回滚"})
+                print(json.dumps(out, ensure_ascii=False, indent=2))
+                return 0
+            cur.execute(f"SELECT COUNT(*) FROM `{TABLE}`")
+            rows = cur.fetchone()[0]
+            out["rows"] = rows
+            if rows > 0 and not args.force:
+                out.update({"action": "REFUSED", "reason": "表内有数据；先确认再 --force"})
+                print(json.dumps(out, ensure_ascii=False, indent=2))
+                return 3
+
+            backup_dir = rel("backups", "strategy-charter-rollback")
+            os.makedirs(backup_dir, exist_ok=True)
+            dump = os.path.join(backup_dir, f"{TABLE}-{datetime.now():%Y%m%d-%H%M%S}.sql")
+            cmd = ["mysqldump", "-u", "root", "--single-transaction", "--no-tablespaces",
+                   DB_NAME, TABLE]
+            with open(dump, "wb") as fh:
+                rc = subprocess.call(cmd, stdout=fh, stderr=subprocess.PIPE)
+            if rc != 0 or os.path.getsize(dump) == 0:
+                out.update({"action": "REFUSED", "reason": f"mysqldump 失败 rc={rc}（fail-closed）"})
+                print(json.dumps(out, ensure_ascii=False, indent=2))
+                return 4
+            out["dump"] = dump
+
+            cur.execute(f"DROP TABLE `{TABLE}`")
+            cur.execute("SHOW TABLES LIKE %s", (TABLE,))
+            out.update({"action": "DROPPED", "table_exists_after": cur.fetchone() is not None})
+            print(json.dumps(out, ensure_ascii=False, indent=2))
+            return 0
+    finally:
+        conn.close()
+
+
+if __name__ == "__main__":
+    sys.exit(main())

backend/scripts/seed_ar_overdue_rule.py

@@ -0,0 +1,153 @@
+#!/usr/bin/env python3
+"""播种应收账龄逾期预警规则（`alert_rules.rule_type='ar_overdue'`）— 阶段B 专用
+
+用法：
+    # 只打印将写入的行（默认 dry-run，阶段A 唯一允许的形式）
+    python3 backend/scripts/seed_ar_overdue_rule.py --entity-id 1
+
+    # 真正写库（阶段B，须另铸门/执行件授权）
+    python3 backend/scripts/seed_ar_overdue_rule.py --entity-id 1 --apply
+
+    # 把参数 JSON 打到 stdout/文件，供阶段A `POST /run-ar-overdue` 的 dry-run body 复演
+    python3 backend/scripts/seed_ar_overdue_rule.py --entity-id 1 --emit-params /tmp/ar_params.json
+
+阈值与责任人**只存在于本脚本**（业务数字唯一定义处）；运行期一律读 DB 行，
+代码侧不含任何阈值常量（任务硬约束 §3.4）。
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import sys
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from dotenv import load_dotenv  # noqa: E402
+
+load_dotenv(os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), ".env"))
+
+from app.database import get_session_local  # noqa: E402
+from app.models import KPIDefinition  # noqa: E402
+from app.utils import ar_caliber  # noqa: E402
+
+RULE_TYPE = "ar_overdue"
+DEFAULT_KPI_CODE = "F_AR_002"  # 逾期应收账款率（财务口径）；库内暂缺时须由财务定绑定
+
+# ── 业务数字唯一定义处（阶段B G5 2026-10-21 裁定：停用 150+ 占比，改 90–150 前池）──
+DEFAULT_PARAMS = {
+    "kpi_code": DEFAULT_KPI_CODE,
+    "caliber": ar_caliber.CALIBER_DEBIT,
+    "share_90plus_gt": None,
+    "share_150p_gt": None,
+    "share_90_150_rise_pp_min": 5.0,
+    "share_90_150_amt_min": 100000,
+    "overdue_amt_display_only": True,
+    "single_overdue_amt_gt": 50000,
+    "assignee_by_unit": {},
+    "assignee_source": {"aggregate_assignee": "", "escalate": "任富海"},
+}
+
+
+def resolve_kpi(db, entity_id: int, kpi_code: str, kpi_id: int | None):
+    if kpi_id:
+        return db.query(KPIDefinition).filter(
+            KPIDefinition.id == kpi_id, KPIDefinition.entity_id == entity_id).first()
+    return db.query(KPIDefinition).filter(
+        KPIDefinition.kpi_code == kpi_code, KPIDefinition.entity_id == entity_id).first()
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(description="播种 ar_overdue 预警规则")
+    ap.add_argument("--entity-id", type=int, required=True)
+    ap.add_argument("--kpi-code", default=DEFAULT_KPI_CODE)
+    ap.add_argument("--kpi-id", type=int, default=None, help="显式绑定 KPI（优先于 kpi-code）")
+    ap.add_argument("--assignee", default=None, help="覆盖 assignee_default（财务指定责任人）")
+    ap.add_argument("--apply", action="store_true", help="真正写库（阶段B 才允许）")
+    ap.add_argument("--emit-params", default=None, help="把最终 params JSON 写入指定文件")
+    ap.add_argument("--params-file", default=None,
+                    help="整份 params JSON（与 DB 行逐字一致；给了就覆盖 DEFAULT_PARAMS）")
+    args = ap.parse_args()
+
+    if args.params_file:
+        with open(args.params_file, encoding="utf-8") as f:
+            raw = json.load(f)
+        params = raw.get("params", raw) if isinstance(raw, dict) else raw
+        if not isinstance(params, dict):
+            print("[BLOCK] --params-file 必须是 JSON 对象或 {params:{...}}", file=sys.stderr)
+            return 3
+        params = json.loads(json.dumps(params, ensure_ascii=False))  # deep copy
+    else:
+        params = json.loads(json.dumps(DEFAULT_PARAMS))  # deep copy
+    if args.assignee is not None:
+        params["assignee_default"] = args.assignee
+    if args.emit_params:
+        with open(args.emit_params, "w", encoding="utf-8") as f:
+            json.dump({"params": params}, f, ensure_ascii=False, indent=2)
+        print("params 已写出: %s" % args.emit_params)
+
+    db = get_session_local()()
+    try:
+        kpi = resolve_kpi(db, args.entity_id, args.kpi_code, args.kpi_id)
+        if not kpi:
+            print("[BLOCK] entity=%s 内找不到 KPI code=%r（kpi_id=%s）——无法绑定，未写入任何行"
+                  % (args.entity_id, args.kpi_code, args.kpi_id), file=sys.stderr)
+            print("[INFO] 该账套现存 AR 相关 KPI：", file=sys.stderr)
+            rows = db.query(KPIDefinition).filter(
+                KPIDefinition.entity_id == args.entity_id,
+                KPIDefinition.kpi_code.like("%AR%"),
+            ).order_by(KPIDefinition.id).all()
+            for r in rows:
+                print("        id=%s code=%s name=%s" % (r.id, r.kpi_code, r.kpi_name), file=sys.stderr)
+            print("[INFO] 处理：--kpi-id 显式指定，或财务先在 kpi_definitions 建 F_AR_002。",
+                  file=sys.stderr)
+            return 2
+
+        from app.api.alert_rules import AlertRule
+        exist = db.query(AlertRule).filter(
+            AlertRule.entity_id == args.entity_id,
+            AlertRule.rule_type == RULE_TYPE,
+        ).all()
+        print("== 将写入的行（%s）==" % ("APPLY" if args.apply else "DRY-RUN"))
+        print(json.dumps({
+            "entity_id": args.entity_id,
+            "kpi_id": kpi.id,
+            "kpi_code": kpi.kpi_code,
+            "kpi_name": kpi.kpi_name,
+            "rule_type": RULE_TYPE,
+            "trigger_on": "actual",
+            "enabled": 1,
+            "params": params,
+        }, ensure_ascii=False, indent=2))
+        src = params.get("assignee_source") or {}
+        if not (params.get("assignee_default") or params.get("assignee_by_unit")
+                or src.get("mode") or src.get("aggregate_assignee")):
+            print("[WARN] 责任人均未配置 → 运行期 fail-closed 400（阶段B 前须由财务给值）")
+
+        if exist:
+            print("[SKIP] entity=%s 已有 %d 条 ar_overdue 规则（id=%s），不重复播种"
+                  % (args.entity_id, len(exist), [r.id for r in exist]))
+            return 0
+        if not args.apply:
+            print("[DRY-RUN] 未写库。阶段B 授权后加 --apply。")
+            return 0
+
+        rule = AlertRule(
+            entity_id=args.entity_id,
+            kpi_id=kpi.id,
+            rule_type=RULE_TYPE,
+            trigger_on="actual",
+            enabled=1,
+            params=params,
+        )
+        db.add(rule)
+        db.commit()
+        db.refresh(rule)
+        print("[APPLIED] 已写入 alert_rules.id=%s" % rule.id)
+        return 0
+    finally:
+        db.close()
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

backend/tests/test_ar_aging_overdue_20261020.py

@@ -0,0 +1,510 @@
+# -*- coding: utf-8 -*-
+"""应收账龄看板 + 逾期预警（T1/T2 · 2026-10-20 R2）单测
+
+任务：`project-20261020-ar-overdue-alert-and-aging-board-fullstack`（阶段A）
+
+口径与边界（全部为硬断言）：
+  · 六档互斥：逐行 `b_0_30+…+b_150p == total_amt`；
+  · 逾期额 `o_i = total_amt − b_0_30`，负档位**不** ABS/COALESCE（DQ-AR-04 杨创 -20）；
+  · 三口径分列、禁跨口径合计（`caliber=all` → 400）；
+  · 缺期（2026-10）⇒ `has_data=false` 且金额 null（禁补 0）；
+  · §6 裁决：`red.share_90plus_gt=null` ⇒ 90 天占比红线停用（不回落 0.25）；
+    新增「单户 >30 天且 >5 万」逐户红线 + 「150+ 占比环比上升」结构恶化红线；
+  · §6.2 assignee：手工点名 → 近 3 个月主业务员（employee_dict / 同库摘要降级）
+    → 部门负责人（渠道 118 / 商务 114 / 直属酒窖 117）；两路皆失败 ⇒ skipped[]，不落空行；
+  · 阈值/责任人全部走 `alert_rules.params`（本套件在本地 SQLite 夹具里改参数复跑，
+    证明「改参数不改代码」）；
+  · 写通道只 INSERT，历史行零 UPDATE/DELETE，同 `alert_message` 幂等跳过。
+
+本套件**不连任何生产库**：`tests/conftest.py` 已把引擎替换为内存 SQLite；
+`hanke_ar_aging` 非 ORM 模型，故用原始 DDL 在夹具里建表。
+"""
+from __future__ import annotations
+
+import json
+
+import pytest
+from sqlalchemy import text
+from sqlalchemy.orm import Session
+
+from app.api.alert_rules import AlertRule
+from app.models import KPIAlert
+from app.services import ar_overdue_alert
+from app.utils import ar_caliber
+from tests.conftest import create_test_kpi, create_test_user, get_token_for_user, auth_header
+
+ENT = 1
+DDL = """
+CREATE TABLE hanke_ar_aging (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    entity_id INTEGER NOT NULL,
+    period VARCHAR(7) NOT NULL,
+    caliber VARCHAR(24) NOT NULL,
+    unit_code VARCHAR(20) NOT NULL,
+    unit_name VARCHAR(100) NOT NULL,
+    opening_amt NUMERIC(18,2) NOT NULL DEFAULT 0,
+    total_amt NUMERIC(18,2) NOT NULL DEFAULT 0,
+    b_0_30 NUMERIC(18,2) NOT NULL DEFAULT 0,
+    b_31_60 NUMERIC(18,2) NOT NULL DEFAULT 0,
+    b_61_90 NUMERIC(18,2) NOT NULL DEFAULT 0,
+    b_90_120 NUMERIC(18,2) NOT NULL DEFAULT 0,
+    b_120_150 NUMERIC(18,2) NOT NULL DEFAULT 0,
+    b_150p NUMERIC(18,2) NOT NULL DEFAULT 0,
+    source_file VARCHAR(160) NOT NULL DEFAULT 'fixture.csv',
+    source_sha256 VARCHAR(64) NOT NULL DEFAULT 'fixture',
+    loaded_at DATETIME DEFAULT CURRENT_TIMESTAMP
+)
+"""
+
+# §6.1 参数最终形状（含 aggregate_assignee，仅测试用显式给值以验证聚合可落库）
+BASE_PARAMS = {
+    "kpi_code": "F_AR_002",
+    "caliber": ar_caliber.CALIBER_DEBIT,
+    "red": {"share_90plus_gt": None, "single_overdue_amt_gt": 50000,
+            "single_overdue_min_days": 30, "count_mom_pct_gte": 0.20,
+            "structure_worsen": True},
+    "yellow": {"share_90plus_min": 0.15, "share_90plus_max": 0.25,
+               "single_overdue_amt_min": 10000, "single_overdue_amt_max": 50000},
+    "assignee_source": {
+        "mode": "salesman_then_dept",
+        "lookback_months": 3,
+        "dept_fallback": {"渠道": 118, "商务": 114, "直属酒窖": 117},
+        "dept_by_unit": {"U3": "商务"},
+        "aggregate_assignee": "财务应收会计",
+    },
+    "assignee_by_unit": {},
+    "exclude_unit_codes": [],
+    "action_texts": {"red": "催收+停供评估", "yellow": "对账+催收"},
+}
+
+ZXT_DDL = """
+CREATE TABLE zxt_masterbill (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    entity_id INTEGER,
+    unit_id INTEGER,
+    emp_id INTEGER,
+    bill_type INTEGER DEFAULT 1,
+    bill_date DATE,
+    sum_money NUMERIC(18,2) DEFAULT 0,
+    abst VARCHAR(500)
+)
+"""
+
+
+def _reset_table(db: Session) -> None:
+    db.execute(text("DROP TABLE IF EXISTS hanke_ar_aging"))
+    db.execute(text(DDL))
+    db.commit()
+
+
+def _add(db: Session, period: str, code: str, name: str, total: float, buckets: list,
+         opening: float = 0.0, caliber: str = ar_caliber.CALIBER_DEBIT, entity_id: int = ENT) -> None:
+    assert abs(sum(buckets) - total) < 0.005, "夹具必须满足六档和 = 合计"
+    db.execute(text(
+        "INSERT INTO hanke_ar_aging (entity_id, period, caliber, unit_code, unit_name, opening_amt,"
+        " total_amt, b_0_30, b_31_60, b_61_90, b_90_120, b_120_150, b_150p)"
+        " VALUES (:e,:p,:c,:code,:name,:op,:tot,:b0,:b1,:b2,:b3,:b4,:b5)"
+    ), {"e": entity_id, "p": period, "c": caliber, "code": code, "name": name, "op": opening,
+        "tot": total, "b0": buckets[0], "b1": buckets[1], "b2": buckets[2],
+        "b3": buckets[3], "b4": buckets[4], "b5": buckets[5]})
+    db.commit()
+
+
+@pytest.fixture()
+def aging(db: Session):
+    """两期夹具（2026-08 / 2026-09），含负档位行 + 未逾期户 + 贷方口径干扰行。
+
+    2026-09：合计 173,159；90+ = 63,179（36.48%）；逾期户 = U1/U3/U4 共 3 户。
+    2026-08：合计 163,159；90+ = 43,179（26.46%）；逾期户 3 户 ⇒ 环比 0%。
+    150+ 占比：2026-08 = 14.21% → 2026-09 = 16.27%（上升 ⇒ 结构恶化红）。
+    """
+    _reset_table(db)
+    _add(db, "2026-09", "U1", "甲公司", 100000, [20000, 10000, 10000, 20000, 15000, 25000], opening=5000)
+    _add(db, "2026-09", "U2", "乙公司", 40000, [40000, 0, 0, 0, 0, 0])
+    _add(db, "2026-09", "U3", "丙公司", 30000, [0, 30000, 0, 0, 0, 0])
+    # 负档位（DQ-AR-04 同形）：total=3159, b_0_30=-20, b_150p=3179 ⇒ o_i=3179（不得 ABS 抹成 3139/3159）
+    _add(db, "2026-09", "U4", "杨创", 3159, [-20, 0, 0, 0, 0, 3179])
+    # 干扰行：若查询漏 caliber 过滤 → 会污染借方口径合计
+    _add(db, "2026-09", "C1", "贷方户", -34010, [0, 0, 0, 0, 0, -34010],
+         caliber=ar_caliber.CALIBER_CREDIT_SEPARATE)
+    _add(db, "2026-08", "U1", "甲公司", 90000, [30000, 10000, 10000, 10000, 10000, 20000])
+    _add(db, "2026-08", "U2", "乙公司", 50000, [50000, 0, 0, 0, 0, 0])
+    _add(db, "2026-08", "U3", "丙公司", 20000, [0, 20000, 0, 0, 0, 0])
+    _add(db, "2026-08", "U4", "杨创", 3159, [-20, 0, 0, 0, 0, 3179])
+    return db
+
+
+def _staff(db: Session, *, with_employee_dict: bool = True, with_units_dict: bool = True,
+           with_zxt: bool = True) -> None:
+    """维度夹具：units_dict(U1/U3/U4) + employee_dict(14/32/114) + zxt_masterbill。"""
+    for t in ("units_dict", "employee_dict", "zxt_masterbill"):
+        db.execute(text("DROP TABLE IF EXISTS %s" % t))
+    if with_units_dict:
+        db.execute(text("CREATE TABLE units_dict (unit_code VARCHAR(20), unit_id INTEGER)"))
+        for code, uid in (("U1", 101), ("U3", 103), ("U4", 104)):
+            db.execute(text("INSERT INTO units_dict VALUES (:c,:u)"), {"c": code, "u": uid})
+    if with_employee_dict:
+        db.execute(text("CREATE TABLE employee_dict (id INTEGER, u_Name VARCHAR(64))"))
+        for i, n in ((14, "李世昌"), (32, "兀晨博"), (114, "王经理"), (118, "渠道负责人"),
+                     (117, "酒窖负责人")):
+            db.execute(text("INSERT INTO employee_dict VALUES (:i,:n)"), {"i": i, "n": n})
+    if with_zxt:
+        db.execute(text(ZXT_DDL))
+        rows = [
+            (ENT, 101, 14, "2026-08-10", "向【甲公司】出售商品：李世昌"),
+            (ENT, 101, 14, "2026-08-11", "向【甲公司】出售商品：李世昌"),
+            (ENT, 101, 14, "2026-09-05", "向【甲公司】出售商品：李世昌"),
+            (ENT, 101, 32, "2026-09-20", "向【甲公司】出售商品：兀晨博"),
+        ]
+        for e, u, m, d, a in rows:
+            db.execute(text("INSERT INTO zxt_masterbill (entity_id, unit_id, emp_id, bill_date, abst)"
+                            " VALUES (:e,:u,:m,:d,:a)"),
+                       {"e": e, "u": u, "m": m, "d": d, "a": a})
+    db.commit()
+
+
+@pytest.fixture()
+def staff(aging: Session):
+    _staff(aging)
+    return aging
+
+
+# ─────────────────────── 数据口径 ───────────────────────
+
+def test_six_buckets_equal_total_per_row(aging: Session):
+    rows = aging.execute(text(
+        "SELECT period, unit_code, total_amt, b_0_30+b_31_60+b_61_90+b_90_120+b_120_150+b_150p AS s"
+        " FROM hanke_ar_aging WHERE entity_id=1 AND caliber=:c"
+    ), {"c": ar_caliber.CALIBER_DEBIT}).mappings().all()
+    assert len(rows) == 8
+    for r in rows:
+        assert abs(float(r["s"]) - float(r["total_amt"])) < 0.005, r
+
+
+def test_caliber_literals_are_exact():
+    assert ar_caliber.CALIBERS == ("全户口径", "借方应收户口径", "贷方户单列(净额)")
+    with pytest.raises(ar_caliber.CaliberError):
+        ar_caliber.validate_caliber("贷方户单列")  # 少「(净额)」⇒ 会漏 42 行
+    with pytest.raises(ar_caliber.CaliberError):
+        ar_caliber.validate_caliber("all")
+    assert ar_caliber.validate_caliber(None, allow_none=True) == ar_caliber.CALIBER_DEBIT
+
+
+# ─────────────────────── 计算与分级（§6 阈值） ───────────────────────
+
+def test_share_90plus_red_disabled_and_structure_worsen(staff: Session):
+    """§6.1：90 天占比红线置 null ⇒ 不生成该聚合项；150+ 环比上升 ⇒ 结构恶化红。"""
+    res = ar_overdue_alert.compute(staff, ENT, "2026-09", BASE_PARAMS)
+    assert res["skipped_reason"] is None
+    m = res["metrics"]
+    assert m["total_amt"] == pytest.approx(173159.0, abs=0.01)      # 贷方口径行未被混入
+    assert m["overdue_90plus"] == pytest.approx(63179.0, abs=0.01)
+    assert m["share_90plus"] == pytest.approx(63179 / 173159, rel=1e-6)
+    assert m["share_150p"] == pytest.approx(28179 / 173159, abs=1e-6)   # metrics 保留 6 位
+    assert m["prev_share_150p"] == pytest.approx(23179 / 163159, abs=1e-6)
+
+    keys = [a["key"] for a in res["aggregates"]]
+    assert "share_90plus" not in keys, "红线停用后 36.48% 不应再出聚合项（无区分度）"
+    assert "structure_150p" in keys
+    agg = [a for a in res["aggregates"] if a["key"] == "structure_150p"][0]
+    assert agg["level"] == "red"
+    assert agg["assignee"] == "财务应收会计"
+    for token in ("2026-09", "150+ 档占比", "16.27%", "14.21%", "结构恶化", "催收+停供评估"):
+        assert token in agg["alert_message"], token
+    assert len(agg["alert_message"]) <= 500
+    assert any("90 天占比红线停用" in n for n in res["notes"])
+
+
+def test_single_overdue_red_and_yellow_households(staff: Session):
+    res = ar_overdue_alert.compute(staff, ENT, "2026-09", BASE_PARAMS)
+    by_code = {c["unit_code"]: c for c in res["per_customer"]}
+    # 单户 >30 天且 >5 万 ⇒ 红
+    assert by_code["U1"]["level"] == "red"
+    assert by_code["U1"]["overdue_amt"] == pytest.approx(80000.0)
+    assert by_code["U1"]["oldest_days"] == 151
+    # 单户 1–5 万 ⇒ 黄
+    assert by_code["U3"]["level"] == "yellow" and by_code["U3"]["oldest_days"] == 31
+    # U4 逾期 3,179 < 1 万 ⇒ 不生成；U2 全额 0-30 ⇒ 不生成
+    assert "U4" not in by_code and "U2" not in by_code
+    for c in res["per_customer"]:
+        assert c["assignee"], "禁落空 assignee"
+        assert c["alert_message"].count("｜建议：") == 1
+        assert "｜责任人：" in c["alert_message"]
+
+
+def test_single_customer_message_wording_hkjj120_20261021(aging: Session):
+    """D2（2026-10-21）：单户文案须把「最老账龄」与「逾期合计」分开写清。
+
+    样例 = 生产 HKJJ120 逐档实数（借方应收户口径，只读复核）：
+      total 352,494.00 = 52,374 + 34,549 + 35,799 + 80,191 + 84,466 + 65,115
+      31 天以上逾期合计 = 300,120.00（占该户 85.14%），150+ 档 = 65,115.00。
+    旧文案「期末逾期 151 天以上 ¥300,120.00」会被读成「最老档就是 30 万」，
+    故本用例硬断言旧措辞不再出现、新措辞逐字出现，且判据字段（金额/天数/占比）零变化。
+    """
+    _reset_table(aging)
+    _add(aging, "2026-09", "HKJJ120", "西安酣君品创贸易有限责任公司", 352494.00,
+         [52374.00, 34549.00, 35799.00, 80191.00, 84466.00, 65115.00])
+    p = json.loads(json.dumps(BASE_PARAMS))
+    p["assignee_by_unit"] = {"HKJJ120": "李世昌"}
+    res = ar_overdue_alert.compute(aging, ENT, "2026-09", p)
+    c = {x["unit_code"]: x for x in res["per_customer"]}["HKJJ120"]
+    # 判据字段（禁随文案改动）
+    assert c["level"] == "red"
+    assert c["oldest_days"] == 151
+    assert c["overdue_amt"] == pytest.approx(300120.00)
+    assert c["total_amt"] == pytest.approx(352494.00)
+    assert c["pct_of_unit"] == pytest.approx(0.851419, abs=1e-6)
+    # 文案口径
+    assert "账龄最老 151 天，逾期合计 ¥300,120.00（占该户 85.14%）" in c["alert_message"]
+    assert "期末逾期" not in c["alert_message"]
+    assert c["alert_message"].count("｜建议：") == 1
+    assert c["alert_message"].endswith("｜责任人：李世昌")
+    assert len(c["alert_message"]) <= 500
+
+
+def test_assignee_chain_salesman_then_dept(staff: Session):
+    res = ar_overdue_alert.compute(staff, ENT, "2026-09", BASE_PARAMS)
+    by_code = {c["unit_code"]: c for c in res["per_customer"]}
+    # U1：units_dict → unit_id=101 → 主业务员 emp14 单据最多（3 vs 1）→ 李世昌
+    assert by_code["U1"]["assignee"] == "李世昌"
+    assert "salesman" in by_code["U1"]["assignee_source"]
+    assert by_code["U1"]["emp_id"] == 14
+    # U3：近 3 个月无单据 → 部门负责人 商务 114 → 王经理
+    assert by_code["U3"]["assignee"] == "王经理"
+    assert by_code["U3"]["assignee_source"].startswith("dept:商务")
+
+
+def test_assignee_manual_map_has_priority(staff: Session):
+    p = json.loads(json.dumps(BASE_PARAMS))
+    p["assignee_by_unit"] = {"U1": "财务点名人甲"}
+    res = ar_overdue_alert.compute(staff, ENT, "2026-09", p)
+    by_code = {c["unit_code"]: c for c in res["per_customer"]}
+    assert by_code["U1"]["assignee"] == "财务点名人甲"
+    assert by_code["U1"]["assignee_source"] == "manual:assignee_by_unit"
+
+
+def test_assignee_abst_fallback_when_employee_dict_absent(aging: Session):
+    """employee_dict 缺表 ⇒ 同库单据摘要尾段制单人姓名降级（仍非空、可审计）。"""
+    _staff(aging, with_employee_dict=False)
+    res = ar_overdue_alert.compute(aging, ENT, "2026-09", BASE_PARAMS)
+    by_code = {c["unit_code"]: c for c in res["per_customer"]}
+    assert by_code["U1"]["assignee"] == "李世昌"
+    assert "abst_operator" in by_code["U1"]["assignee_source"]
+
+
+def test_assignee_unresolved_goes_to_skipped(aging: Session):
+    """维度表全缺 + 部门无法归类 ⇒ 该户不生成，skipped[] 写原因（禁落空 assignee）。"""
+    _staff(aging, with_employee_dict=False, with_units_dict=False, with_zxt=False)
+    p = json.loads(json.dumps(BASE_PARAMS))
+    p["assignee_source"].pop("dept_by_unit")
+    res = ar_overdue_alert.compute(aging, ENT, "2026-09", p)
+    assert res["per_customer"] == []
+    skipped = {s["unit_code"]: s for s in res["skipped"] if s["scope"] == "customer"}
+    assert set(skipped) == {"U1", "U3"}
+    assert "unit_link_unresolved" in skipped["U1"]["reason"]
+    assert "dept_unclassified" in skipped["U1"]["reason"]
+    assert [h for h in ar_overdue_alert.hits(res) if h.get("unit_code")] == []
+
+
+def test_assignee_missing_is_fail_closed(aging: Session):
+    p = {"red": {}, "yellow": {}, "assignee_default": "", "assignee_by_unit": {}}
+    with pytest.raises(ar_overdue_alert.AssigneeMissing):
+        ar_overdue_alert.ensure_assignee_configured(p)
+    # 配了逐户派生模式即视为已配置（派生失败按户 skipped，不整体 400）
+    p["assignee_source"] = {"mode": "salesman_then_dept"}
+    ar_overdue_alert.ensure_assignee_configured(p)
+
+
+def test_level_boundaries_are_param_driven(aging: Session):
+    """改 DB 参数复跑 ⇒ 命中集合变化，代码零改动（§6.3 第 11 项：single_overdue_amt_gt）。"""
+    _staff(aging)
+    p = json.loads(json.dumps(BASE_PARAMS))
+    assert {c["unit_code"] for c in ar_overdue_alert.compute(aging, ENT, "2026-09", p)["per_customer"]} == {"U1", "U3"}
+    p["red"]["single_overdue_amt_gt"] = 300000          # 5 万 → 30 万 ⇒ U1(8 万) 掉出红线且高于黄线上限
+    res = ar_overdue_alert.compute(aging, ENT, "2026-09", p)
+    assert {c["unit_code"] for c in res["per_customer"]} == {"U3"}
+    p["red"]["single_overdue_amt_gt"] = 50000
+    p["yellow"]["single_overdue_amt_min"] = 100000      # 黄线下限 10 万 ⇒ U3 消失
+    assert {c["unit_code"] for c in ar_overdue_alert.compute(aging, ENT, "2026-09", p)["per_customer"]} == {"U1"}
+
+
+def test_missing_period_is_skipped_not_zero(aging: Session):
+    res = ar_overdue_alert.compute(aging, ENT, "2026-10", BASE_PARAMS)
+    assert res["skipped_reason"] == "no_data_for_period:2026-10"
+    assert res["aggregates"] == [] and res["per_customer"] == []
+
+
+def test_default_period_is_max_and_mom_computed(staff: Session):
+    res = ar_overdue_alert.compute(staff, ENT, None, BASE_PARAMS)
+    assert res["period"] == "2026-09"
+    assert res["metrics"]["prev_period"] == "2026-08"
+    assert res["metrics"]["prev_overdue_customers"] == 3
+    assert res["metrics"]["overdue_customers"] == 3
+    assert res["metrics"]["count_mom_pct"] == pytest.approx(0.0)
+    assert "count_mom" not in {a["key"] for a in res["aggregates"]}
+
+
+def test_no_previous_period_no_mom(aging: Session):
+    res = ar_overdue_alert.compute(aging, ENT, "2026-08", BASE_PARAMS)
+    assert res["metrics"]["prev_period"] is None
+    assert res["metrics"]["count_mom_pct"] is None
+    assert any("不计算环比" in n for n in res["notes"])
+
+
+def test_exclude_unit_codes(aging: Session):
+    _staff(aging)
+    p = json.loads(json.dumps(BASE_PARAMS))
+    p["exclude_unit_codes"] = ["U1"]
+    res = ar_overdue_alert.compute(aging, ENT, "2026-09", p)
+    assert "U1" not in {c["unit_code"] for c in res["per_customer"]}
+    assert res["metrics"]["total_amt"] < 173159
+
+
+# ─────────────────────── 落库（只 INSERT） ───────────────────────
+
+def test_write_alerts_idempotent_and_history_intact(staff: Session):
+    kpi = create_test_kpi(staff, kpi_code="F_AR_002", kpi_name="逾期应收账款率")
+    hist = KPIAlert(entity_id=ENT, kpi_id=kpi.id, alert_level="red",
+                    alert_message="历史行-不得被改写", alert_type="actual", status="pending",
+                    assignee="old-owner")
+    staff.add(hist)
+    staff.commit()
+    hist_id = hist.id
+
+    res = ar_overdue_alert.compute(staff, ENT, "2026-09", BASE_PARAMS)
+    first = ar_overdue_alert.write_alerts(staff, ENT, kpi.id, res)
+    assert first["inserted"] == len(ar_overdue_alert.hits(res)) > 0
+    second = ar_overdue_alert.write_alerts(staff, ENT, kpi.id, res)
+    assert second == {"inserted": 0, "skipped_duplicate": first["inserted"]}
+
+    rows = staff.query(KPIAlert).filter(KPIAlert.alert_type == "overdue").all()
+    assert len(rows) == first["inserted"]
+    assert all(r.assignee and r.status == "pending" and r.kpi_value_id is None for r in rows)
+    assert all(len(r.alert_message) <= 500 for r in rows)
+    old = staff.get(KPIAlert, hist_id)
+    assert old.alert_message == "历史行-不得被改写" and old.assignee == "old-owner"
+
+
+# ─────────────────────── HTTP 面（TestClient + 内存库） ───────────────────────
+
+@pytest.fixture()
+def fin_headers(client, db):
+    create_test_user(db, username="fin_probe", role="finance")
+    return auth_header(get_token_for_user(client, username="fin_probe"))
+
+
+def test_api_meta_trend_overview(staff, client, fin_headers):
+    r = client.get("/api/cma/ar-aging/meta", headers=fin_headers)
+    assert r.status_code == 200, r.text
+    body = r.json()
+    assert body["available"] is True
+    assert body["calibers"] == list(ar_caliber.CALIBERS)
+    assert body["latest_period"] == "2026-09"
+    assert body["row_count"] == 9                      # 8 借方 + 1 贷方（三口径不混算，仅计数）
+
+    r = client.get("/api/cma/ar-aging/trend", headers=fin_headers,
+                   params={"caliber": ar_caliber.CALIBER_DEBIT})
+    assert r.status_code == 200
+    pts = {p["period"]: p for p in r.json()["points"]}
+    assert pts["2026-09"]["has_data"] is True
+    assert pts["2026-10"]["has_data"] is False
+    assert pts["2026-10"]["total_amt"] is None and pts["2026-10"]["customers"] is None
+
+    r = client.get("/api/cma/ar-aging/overview", headers=fin_headers,
+                   params={"period": "2026-09", "caliber": ar_caliber.CALIBER_DEBIT})
+    ov = r.json()
+    assert ov["has_data"] is True
+    assert ov["buckets_sum_matches_total"] is True
+    assert ov["opening_amt"] == pytest.approx(5000.0)
+    assert ov["opening_amt_non_aging_base"] is True
+    assert ov["total_amt"] == pytest.approx(173159.0, abs=0.01)
+
+
+def test_api_caliber_all_rejected(aging, client, fin_headers):
+    for ep in ("overview", "trend", "customers"):
+        r = client.get(f"/api/cma/ar-aging/{ep}", headers=fin_headers, params={"caliber": "all"})
+        assert r.status_code == 400, (ep, r.text)
+        assert "禁止跨口径合计" in r.json()["detail"]
+
+
+def test_api_customers_sort_and_export(aging, client, fin_headers):
+    r = client.get("/api/cma/ar-aging/customers", headers=fin_headers,
+                   params={"period": "2026-09", "caliber": ar_caliber.CALIBER_DEBIT,
+                           "overdue_only": "true", "sort": "overdue_amt", "order": "desc"})
+    items = r.json()["items"]
+    assert [i["unit_code"] for i in items][:3] == ["U1", "U3", "U4"]
+    assert items[0]["overdue_amt"] == pytest.approx(80000.0)
+
+    r = client.get("/api/cma/ar-aging/customers/export", headers=fin_headers,
+                   params={"period": "2026-09", "caliber": ar_caliber.CALIBER_DEBIT})
+    assert r.status_code == 200
+    assert r.content[:3] == b"\xef\xbb\xbf"                     # UTF-8 BOM（Excel 直开）
+    lines = r.content.decode("utf-8-sig").splitlines()
+    assert lines[0].startswith("截止期,2026-09,口径,借方应收户口径,导出时刻,")
+    assert "U1" in lines[2] or "U1" in lines[3]
+
+
+def test_api_run_ar_overdue_rule_params_parameterized(staff, client, db, fin_headers):
+    """端到端：规则参数来自 DB 行 ⇒ 改 `single_overdue_amt_gt` 复跑命中变化，代码零改动。"""
+    kpi = create_test_kpi(db, kpi_code="F_AR_002", kpi_name="逾期应收账款率")
+    rule = AlertRule(entity_id=ENT, kpi_id=kpi.id, rule_type=ar_overdue_alert.RULE_TYPE,
+                     trigger_on="actual", enabled=1, params=BASE_PARAMS)
+    db.add(rule)
+    db.commit()
+
+    r = client.post("/api/cma/alert-rules/run-ar-overdue",
+                    headers=fin_headers, params={"period": "2026-09", "dry_run": "true"})
+    assert r.status_code == 200, r.text
+    j = r.json()
+    assert j["params_source"] == "alert_rules#%s" % rule.id
+    assert j["kpi_id"] == kpi.id
+    assert j["write_blockers"] == []
+    assert j["hit_count"] == 3                       # 结构恶化聚合红 + U1 红 + U3 黄
+    assert {c["unit_code"] for c in j["per_customer"]} == {"U1", "U3"}
+    assert all(c["assignee"] for c in j["per_customer"])
+    assert j["skipped"] == []
+    assert "structure_150p" in {a["key"] for a in j["aggregates"]}
+
+    params2 = json.loads(json.dumps(BASE_PARAMS))
+    params2["red"]["single_overdue_amt_gt"] = 300000   # 5 万 → 30 万
+    rule.params = params2
+    db.commit()
+    r = client.post("/api/cma/alert-rules/run-ar-overdue",
+                    headers=fin_headers, params={"period": "2026-09", "dry_run": "true"})
+    assert r.status_code == 200
+    j2 = r.json()
+    assert {c["unit_code"] for c in j2["per_customer"]} == {"U3"}
+    assert j2["hit_count"] == 2 and j2["hit_count"] < j["hit_count"]   # 聚合红 + U3 黄
+
+
+def test_api_run_ar_overdue_assignee_empty_400(aging, client, db, fin_headers):
+    kpi = create_test_kpi(db, kpi_code="F_AR_002x", kpi_name="逾期应收账款率")
+    p = {"red": {}, "yellow": {}, "assignee_default": "", "assignee_by_unit": {}}
+    db.add(AlertRule(entity_id=ENT, kpi_id=kpi.id, rule_type=ar_overdue_alert.RULE_TYPE,
+                     trigger_on="actual", enabled=1, params=p))
+    db.commit()
+    r = client.post("/api/cma/alert-rules/run-ar-overdue",
+                    headers=fin_headers, params={"period": "2026-09", "dry_run": "true"})
+    assert r.status_code == 400
+    assert "责任人" in r.json()["detail"]
+
+    # dry_run=false + 参数覆盖 → 400（防绕过 DB 配置）
+    r = client.post("/api/cma/alert-rules/run-ar-overdue", headers=fin_headers,
+                    params={"period": "2026-09", "dry_run": "false"},
+                    json={"params": BASE_PARAMS})
+    assert r.status_code == 400
+    assert "禁止参数/KPI 覆盖" in r.json()["detail"]
+
+
+def test_api_meta_degrades_when_table_missing(client, db, fin_headers, monkeypatch):
+    _reset_table(db)
+    db.execute(text("DROP TABLE hanke_ar_aging"))
+    db.commit()
+    import app.api.ar_aging as aa
+    monkeypatch.setattr(aa, "_TABLE", "hanke_ar_aging_missing_zzz")
+    r = client.get("/api/cma/ar-aging/meta", headers=fin_headers)
+    assert r.status_code == 200
+    body = r.json()
+    assert body["available"] is False and "不可用" in body["reason"]
+    r = client.get("/api/cma/ar-aging/trend", headers=fin_headers)
+    assert r.status_code == 200 and r.json()["available"] is False

backend/tests/test_conn_key_rename_stability.py

@@ -0,0 +1,415 @@
+"""P3 阶段A · 连线键解耦（读侧解析键）验收测试 —— 沙箱，不连生产。
+
+判据（对应 P3 单验收标准 2/3）：
+  1. 目标改名后 `from_key/to_key` 跟随**同一节点**（key 由 objective_id 反查，不随中文名变）；
+  2. 索引位移（删掉前一个目标）后 key 跟随节点到新位置；
+  3. 负测：拿掉 `objective_id` 锚点（等价于"未实现解析键"）且名字已改 ⇒ 必须解析失败
+     并被计入 `unresolved`，证明判据有区分度、不是空过。
+"""
+import copy
+
+import pytest
+from fastapi.testclient import TestClient
+from sqlalchemy.orm import Session
+
+from app.models import StrategicMap, MapObjective, MapConnection
+from tests.conftest import create_test_user, get_token_for_user, auth_header
+
+DIMS = [
+    {"key": "finance", "name": "财务层", "icon": "F", "color": "#F56C6C",
+     "objectives": [{"name": "营业收入"}, {"name": "营业利润"}]},
+    {"key": "customer", "name": "客户层", "icon": "C", "color": "#409EFF",
+     "objectives": [{"name": "客户留存"}]},
+    {"key": "process", "name": "流程层", "icon": "P", "color": "#67C23A",
+     "objectives": [{"name": "交付效率"}, {"name": "质量闭环"}]},
+    {"key": "learning", "name": "学习层", "icon": "L", "color": "#E6A23C",
+     "objectives": [{"name": "全员培训"}, {"name": "能力认证"}]},
+]
+
+
+def _seed(db: Session) -> StrategicMap:
+    m = StrategicMap(entity_id=1, title="p3-conn-key", status="draft",
+                     dimensions=copy.deepcopy(DIMS), canvas_data={"connections": []})
+    db.add(m)
+    db.commit()
+    db.refresh(m)
+    for d in DIMS:
+        for i, o in enumerate(d["objectives"]):
+            db.add(MapObjective(entity_id=1, map_id=m.id, dimension_key=d["key"],
+                                name=o["name"], sort_order=i))
+    db.commit()
+    return m
+
+
+def _save_dims(client, headers, db: Session, m: StrategicMap, dims) -> None:
+    """整表 PUT（legacy 通道，批处理/脚本真实会走这条）"""
+    db.refresh(m)
+    body = {"dimensions": dims,
+            "canvas_data": copy.deepcopy(m.canvas_data),
+            "base_updated_at": m.updated_at.isoformat() if m.updated_at else None}
+    r = client.put(f"/api/cma/maps/{m.id}", json=body, headers=headers)
+    assert r.status_code == 200, f"整表 PUT 失败: {r.status_code} {r.text[:300]}"
+
+
+def _conn(client, headers, map_id: int) -> dict:
+    r = client.get(f"/api/cma/maps/{map_id}/connections", headers=headers)
+    assert r.status_code == 200, r.text[:300]
+    return r.json()
+
+
+def _mk_conn(client, headers, map_id: int, frm: str, to: str) -> None:
+    r = client.post(f"/api/cma/maps/{map_id}/connections", headers=headers,
+                    json={"from": frm, "to": to})
+    assert r.status_code == 200, f"建连线失败: {r.status_code} {r.text[:300]}"
+
+
+def test_rename_keeps_render_key(client: TestClient, db: Session):
+    """UI 通道（带回 `_mo_id`）改名 → 连线仍挂同一节点（key 不变、unresolved=0）
+
+    口径 B（ASK-20261023-04，任总 2026-10-23 15:13「放行 4」）：只有带回 `_mo_id` 的
+    通道能精确保住锚点；不带 `_mo_id` 的老通道改名会断线但可见，
+    见 test_rename_without_moid_breaks_visibly。
+    """
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:全员培训", "process:交付效率")
+
+    base = _conn(client, h, m.id)
+    assert base["total"] == 1 and base["unresolved"] == 0, base
+    c0 = base["connections"][0]
+    assert c0["from_key"] == "learning-0" and c0["to_key"] == "process-0", c0
+    assert c0["from_key_source"] == "objective_id"
+
+    # 真实 UI 通道：先读（读侧 m_to_dict→_merge_map_objectives 回写 `_mo_id`），改名后整表回传
+    r = client.get("/api/cma/maps", headers=h)
+    assert r.status_code == 200, r.text[:300]
+    got_map = next(x for x in r.json()["data"] if x["id"] == m.id)
+    dims = got_map["dimensions"]
+    assert dims[3]["objectives"][0].get("_mo_id"), "读侧未回写 _mo_id，UI 通道无法保锚点"
+    dims[3]["objectives"][0]["name"] = "全员培训A"
+    _save_dims(client, h, db, m, dims)
+
+    after = _conn(client, h, m.id)
+    assert after["total"] == 1, f"连线数减少: {after}"
+    assert after["unresolved"] == 0, after["unresolved_items"]
+    a0 = after["connections"][0]
+    # 原始串仍是旧名（禁回填），但渲染键已由锚点解析到新节点
+    assert a0["from"] == "learning:全员培训", a0
+    assert a0["from_key"] == "learning-0", a0
+    assert a0["from_key_source"] == "objective_id" and a0["from_key_reason"] is None, a0
+
+
+def test_index_shift_follows_node(client: TestClient, db: Session):
+    """删除同层前一个目标 ⇒ 节点索引位移，key 跟随节点而非冻结在旧序号"""
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:能力认证", "process:交付效率")
+
+    base = _conn(client, h, m.id)["connections"][0]
+    assert base["from_key"] == "learning-1", base
+
+    dims = copy.deepcopy(m.dimensions)
+    dims[3]["objectives"] = dims[3]["objectives"][1:]  # 删掉「全员培训」
+    _save_dims(client, h, db, m, dims)
+
+    after = _conn(client, h, m.id)
+    assert after["unresolved"] == 0, after["unresolved_items"]
+    a0 = after["connections"][0]
+    assert a0["from_key"] == "learning-0", f"key 未跟随节点位移: {a0}"
+    assert a0["from_key_source"] == "objective_id", a0
+
+
+def test_negative_without_anchor_must_fail(client: TestClient, db: Session):
+    """负测：拿掉 objective_id 锚点 + 名字已改 ⇒ 必须解析失败并计入 unresolved
+
+    证明判据有区分度：若解析键能力缺失（=改造前行为），这条必然断线且被计数，
+    不会静默画到错节点上。
+    """
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:全员培训", "process:交付效率")
+
+    dims = copy.deepcopy(m.dimensions)
+    dims[3]["objectives"][0]["name"] = "全员培训A"
+    _save_dims(client, h, db, m, dims)
+
+    # 模拟"无锚点"（历史脏行 / 未回填）：抹掉 from_objective_id
+    row = db.query(MapConnection).filter(MapConnection.map_id == m.id).first()
+    row.from_objective_id = None
+    db.commit()
+
+    broken = _conn(client, h, m.id)
+    b0 = broken["connections"][0]
+    assert b0["from_key"] is None, f"无锚点 + 旧名不应解析成功: {b0}"
+    assert b0["from_key_reason"] == "objective_unresolved", b0
+    assert broken["unresolved"] == 1, broken["unresolved_items"]
+    assert broken["unresolved_items"][0]["side"] == "from"
+    assert broken["unresolved_items"][0]["reason"] == "objective_unresolved"
+    # 另一端不受影响
+    assert b0["to_key"] == "process-0" and b0["to_key_source"] == "objective_id", b0
+
+
+def test_fallback_name_source_still_works(client: TestClient, db: Session):
+    """兼容期：历史行无 objective_id 但名字未改 ⇒ 仍按 `层:名字` 解析成功（不误报未解析）"""
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:全员培训", "process:交付效率")
+
+    row = db.query(MapConnection).filter(MapConnection.map_id == m.id).first()
+    row.from_objective_id = None
+    db.commit()
+
+    got = _conn(client, h, m.id)
+    assert got["unresolved"] == 0, got["unresolved_items"]
+    c0 = got["connections"][0]
+    assert c0["from_key"] == "learning-0" and c0["from_key_source"] == "name", c0
+
+
+# ── project 2026-10-22 23:25 裁定的 3 项准入条件（阶段B 开跑前补测）─────────────
+# 条件2-a：连续两次保存（不改名，不带 _mo_id）⇒ map_objectives 行 id 集合不变（幂等）；
+# 条件2-b：同一次保存内「改名 + 同层调序」且不带 _mo_id ⇒ 两个目标的 id 不得串位；
+# 条件3  ：删除面 —— dims 少传整层 ⇒ 该层行才删；被删目标的连线端计入 unresolved，不得静默消失。
+# 另：hold 口径断言 total = green + yellow + red + hold（hold 不得二次计入 miss）。
+
+def _row_ids(db: Session, map_id: int) -> dict:
+    """{(层, 名字): 行 id}"""
+    rows = db.query(MapObjective).filter(MapObjective.map_id == map_id).all()
+    return {(r.dimension_key, r.name): r.id for r in rows}
+
+
+def test_save_twice_is_idempotent(client: TestClient, db: Session):
+    """连续两次保存不改名 ⇒ 行 id 集合逐字不变（否则锚点必成野指针）"""
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:全员培训", "process:交付效率")
+
+    first = _row_ids(db, m.id)
+    dims = copy.deepcopy(m.dimensions)
+    _save_dims(client, h, db, m, dims)          # 第 1 次保存（结构未变）
+    second = _row_ids(db, m.id)
+    _save_dims(client, h, db, m, dims)          # 第 2 次保存
+    third = _row_ids(db, m.id)
+
+    assert first == second, f"首次保存换了行 id: {first} -> {second}"
+    assert second == third, f"二次保存换了行 id: {second} -> {third}"
+    got = _conn(client, h, m.id)
+    assert got["unresolved"] == 0, got["unresolved_items"]
+    assert got["connections"][0]["from_key"] == "learning-0"
+
+
+def test_rename_plus_reorder_without_moid_no_crossbind(client: TestClient, db: Session):
+    """口径 B：同一次保存内「改名 + 同层调序」且不带 _mo_id ⇒ 不得串位
+
+    名字层先把未改名项钉住原行；改名项不再按位置兜底 ⇒ 当新行（旧行删除），
+    连线锚点指向旧行 ⇒ unresolved + stale_objective_id（断线可见，绝不静默错绑）。
+    """
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:能力认证", "process:交付效率")
+    before = _row_ids(db, m.id)
+    assert _conn(client, h, m.id)["connections"][0]["from_key"] == "learning-1"
+
+    # 调序（能力认证升到第 0 位）+ 改名（能力认证→能力认证A），整表 PUT，不带 _mo_id
+    dims = copy.deepcopy(m.dimensions)
+    dims[3]["objectives"] = [{"name": "能力认证A"}, {"name": "全员培训"}]
+    _save_dims(client, h, db, m, dims)
+
+    after = _row_ids(db, m.id)
+    assert after[("learning", "全员培训")] == before[("learning", "全员培训")], \
+        f"未改名的「全员培训」行 id 变了（串位）: {before} -> {after}"
+    assert after[("learning", "能力认证A")] != before[("learning", "能力认证")], \
+        f"口径 B 下老通道改名应新建行（不做位置兜底）: {before} -> {after}"
+
+    got = _conn(client, h, m.id)
+    assert got["total"] == 1, got
+    c0 = got["connections"][0]
+    assert got["unresolved"] == 1, got["unresolved_items"]
+    assert c0["from"] == "learning:能力认证", "原始串被改写（禁回填）"
+    assert c0["from_key"] is None and c0["from_key_reason"] == "stale_objective_id", c0
+    # 关键：锚点不得被改指到别的节点（旧行为会把线静默挂到「全员培训」上）
+    row = db.query(MapConnection).filter(MapConnection.map_id == m.id).first()
+    assert row.from_objective_id == before[("learning", "能力认证")], "连线锚点被静默改指（错绑）"
+
+
+def test_rename_without_moid_breaks_visibly(client: TestClient, db: Session):
+    """口径 B 单点验证：不带 `_mo_id` 的纯改名 ⇒ 新行 + 连线进 unresolved（断线可见）
+
+    同通道下「删A加B」与「A改名为B」数据上不可区分，位置兜底会让前者静默错绑；
+    放弃兜底的代价就是这里：改名会断线，但断线在接口里可见、可计数。
+    """
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:全员培训", "process:交付效率")
+    before = _row_ids(db, m.id)
+
+    dims = copy.deepcopy(m.dimensions)
+    assert not any("_mo_id" in o for o in dims[3]["objectives"]), "fixture 不应带 _mo_id（老通道）"
+    dims[3]["objectives"][0]["name"] = "全员培训A"
+    _save_dims(client, h, db, m, dims)
+
+    after = _row_ids(db, m.id)
+    assert after[("learning", "全员培训A")] != before[("learning", "全员培训")], \
+        f"老通道改名竟保住锚点（口径 B 下应为新行）: {before} -> {after}"
+    assert ("learning", "全员培训") not in after, f"旧行未删: {after}"
+
+    got = _conn(client, h, m.id)
+    assert got["total"] == 1, f"连线行被静默丢弃: {got}"
+    c0 = got["connections"][0]
+    assert c0["from"] == "learning:全员培训", "原始串被改写（禁回填）"
+    assert c0["from_key"] is None and c0["from_key_reason"] == "stale_objective_id", c0
+    assert got["unresolved"] == 1 and got["unresolved_items"][0]["reason"] == "stale_objective_id"
+    assert c0["to_key"] == "process-0", "另一端不应受影响"
+    row = db.query(MapConnection).filter(MapConnection.map_id == m.id).first()
+    assert row.from_objective_id == before[("learning", "全员培训")], "连线锚点被静默改指（错绑）"
+
+
+def test_delete_face_unresolved_not_silent(client: TestClient, db: Session):
+    """删除面：dims 少传整层 ⇒ 该层行删；被删目标的连线端计入 unresolved 且原始串保留"""
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    _mk_conn(client, h, m.id, "learning:全员培训", "process:交付效率")
+
+    dims = copy.deepcopy(m.dimensions)
+    dims = dims[:3]                              # 整层去掉 learning
+    _save_dims(client, h, db, m, dims)
+
+    left = db.query(MapObjective).filter(MapObjective.map_id == m.id,
+                                         MapObjective.dimension_key == "learning").all()
+    assert left == [], f"该层已不在 dims，行未删: {[r.id for r in left]}"
+
+    got = _conn(client, h, m.id)
+    assert got["total"] == 1, f"连线被静默丢弃: {got}"      # 行仍在，不消失
+    c0 = got["connections"][0]
+    assert c0["from"] == "learning:全员培训", "原始串被改写（禁回填）"
+    # 锚点指向已删行 ⇒ reason 是更精确的 `stale_objective_id`（区别于「无锚点」的 objective_unresolved）
+    assert c0["from_key"] is None and c0["from_key_reason"] == "stale_objective_id", c0
+    assert got["unresolved"] == 1, got["unresolved_items"]
+    assert got["unresolved_items"][0]["side"] == "from"
+    assert c0["to_key"] == "process-0", "另一端不应受影响"
+
+
+def test_review_hold_total_invariant(client: TestClient, db: Session):
+    """hold 口径：total = green + yellow + red + hold；hold 不得二次计入 miss（红/黄）"""
+    from app.models import KPIDefinition, KPIValue
+    from app.api.maps import DATA_HOLD_KPIS
+    from tests.conftest import create_test_kpi
+
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    hold_code = next(iter(DATA_HOLD_KPIS))
+    red_code = "P3_TOTAL_INVARIANT_RED"
+    db.query(KPIDefinition).filter(KPIDefinition.kpi_code == red_code).delete(synchronize_session=False)
+    db.commit()
+    red_kpi = create_test_kpi(db, kpi_code=red_code, kpi_name="P3 红项", dimension="customer",
+                              target_value=100.0, kpi_level="strategic", unit="%")
+    db.add(KPIValue(kpi_id=red_kpi.id, period="2026-09", actual_value=10.0, source_type="actual"))
+    db.commit()
+
+    dims = [{"key": "customer", "name": "客户层", "icon": "", "color": "#409EFF", "objectives": [
+        {"name": "待核项", "kpis": [hold_code]},
+        {"name": "红项", "kpis": [red_code]},
+        {"name": "无指标项"},
+    ]}]
+    m = StrategicMap(entity_id=1, title="p3-hold-invariant", status="active",
+                     dimensions=dims, canvas_data={"connections": []})
+    db.add(m)
+    db.commit()
+    db.refresh(m)
+
+    r = client.get(f"/api/cma/maps/{m.id}/review", params={"level": "strategic"}, headers=h)
+    assert r.status_code == 200, r.text
+    s = r.json()["summary"]
+    assert s["total"] == 3, s
+    assert s["hold"] == 1, s
+    assert s["red"] == 1, s
+    assert s["green"] == 1, s
+    assert s["total"] == s["green"] + s["yellow"] + s["red"] + s["hold"], \
+        f"分母不守恒（hold 被二次计入?）: {s}"
+
+    db.query(KPIValue).filter(KPIValue.kpi_id == red_kpi.id).delete(synchronize_session=False)
+    db.query(KPIDefinition).filter(KPIDefinition.kpi_code == red_code).delete(synchronize_session=False)
+    db.commit()
+
+
+
+
+def test_negative_greedy_matching_would_crossbind(client: TestClient, db: Session, monkeypatch):
+    """负测（判据有效性的反证）：把定位算法换回历史「逐项贪心（名→序位→空位）」，
+    则 rename+reorder 场景下两个目标的 id 必然交换 —— 证明
+    test_rename_plus_reorder_without_moid_no_crossbind 有区分度，不是空过。"""
+    from app.api import maps as maps_mod
+
+    def legacy_greedy(m, db_session=None):
+        sess = db_session or db
+        dims = maps_mod._dims_list(m.dimensions)
+        ent = m.entity_id if m.entity_id is not None else 1
+        rows = sess.query(MapObjective).filter(MapObjective.map_id == m.id).order_by(
+            MapObjective.sort_order, MapObjective.id).all()
+        used, keep = set(), []
+        for dim in dims:
+            dk = dim.get("key", "")
+            for i, obj in enumerate(dim.get("objectives") or []):
+                if not isinstance(obj, dict):
+                    continue
+                target = None
+                want = maps_mod._norm_obj_name(obj.get("name"))
+                for r in rows:                                   # ① 名字
+                    if r.id in used or r.dimension_key != dk:
+                        continue
+                    if maps_mod._norm_obj_name(r.name) == want:
+                        target = r
+                        break
+                if target is None:
+                    for r in rows:                               # ② 同层同序位
+                        if r.id in used or r.dimension_key != dk:
+                            continue
+                        if (r.sort_order or 0) == i:
+                            target = r
+                            break
+                if target is None:
+                    for r in rows:                               # ③ 同层空位
+                        if r.id in used or r.dimension_key != dk:
+                            continue
+                        target = r
+                        break
+                if target is None:
+                    target = MapObjective(map_id=m.id, entity_id=ent, dimension_key=dk or "finance")
+                    sess.add(target)
+                else:
+                    used.add(target.id)
+                target.entity_id = ent
+                target.dimension_key = dk
+                target.name = (obj.get("name") or "")[:200]
+                target.icon = obj.get("icon") or "target"
+                target.kpis = obj.get("kpis") or []
+                target.sort_order = i
+                keep.append(target)
+        sess.flush()
+        for r in rows:
+            if r.id not in used:
+                sess.delete(r)
+        sess.commit()
+
+    create_test_user(db)
+    h = auth_header(get_token_for_user(client))
+    m = _seed(db)
+    before = _row_ids(db, m.id)
+
+    monkeypatch.setattr(maps_mod, "_sync_map_objectives", legacy_greedy)
+    dims = copy.deepcopy(m.dimensions)
+    dims[3]["objectives"] = [{"name": "能力认证A"}, {"name": "全员培训"}]
+    _save_dims(client, h, db, m, dims)
+    after = _row_ids(db, m.id)
+
+    assert after[("learning", "全员培训")] != before[("learning", "全员培训")], \
+        f"旧算法竟未串位，则该负测无法证明新判据有效: {before} -> {after}"
+    assert after[("learning", "能力认证A")] == before[("learning", "全员培训")], \
+        f"预期旧算法把「能力认证A」绑到「全员培训」的旧行 id（串位）: {before} -> {after}"

backend/tests/test_dedupe_key.py

@@ -0,0 +1,162 @@
+"""A2 幂等键：kpi_alerts.dedupe_key（任务 project-20261021-ar-overdue-dedupe-key-fullstack）
+
+覆盖：
+  T1 历史旧文案行（dedupe_key=NULL）零改动 + 同语义口径重跑 → 全跳过
+  T2 重复跑不增行（幂等）
+  T3 负测：新单户 → 插入 1 条（证明不是「一律跳过」）
+  T4 负测：同户金额变化 → 不插（语义幂等，键不含金额）
+  T5 同批两条同键 → 1 插 1 跳（批内去重 + 唯一索引兜底）
+  T6 dry-run count_would_insert 与真实 inserted 一致
+  T7 键生成/反解单测（结构 + 聚合族归一化）
+"""
+import pytest
+from sqlalchemy import text
+
+from app.models import KPIAlert
+from app.services import ar_overdue_alert as A
+
+EID, KPI = 1, 517
+P = "2026-09"
+
+LEGACY = [
+    # 1 条聚合（旧文案，能被反解成 agg|share_90plus）
+    "【应收账款逾期】2026-09 借方应收户口径 90 天以上 ¥1,000.00，占应收总额 15.56%（红线 >25.00%）｜建议：见规则配置",
+    # 7 条单户（旧文案，反解成 cust|<code>）
+    "【应收账款逾期】王敏（HKJJ053）2026-09 账龄最老 151 天，逾期合计 ¥666,537.00（占该户 100.00%）｜建议：催收｜责任人：李世昌",
+    "【应收账款逾期】西安酣君品创（HKJJ120）2026-09 账龄最老 151 天，逾期合计 ¥300,120.00（占该户 85.14%）｜建议：催收｜责任人：李世昌",
+    "【应收账款逾期】吉林酣客（HKJJ156）2026-09 账龄最老 151 天，逾期合计 ¥75,640.00（占该户 100.00%）｜建议：催收｜责任人：李世昌",
+    "【应收账款逾期】荣升酒窖（HKJJ166）2026-09 账龄最老 151 天，逾期合计 ¥36,586.00（占该户 100.00%）｜建议：催收｜责任人：兀晨博",
+    "【应收账款逾期】宝鸡裕腾（HKJJ175）2026-09 账龄最老 91 天，逾期合计 ¥32,560.00（占该户 100.00%）｜建议：催收｜责任人：兀晨博",
+    "【应收账款逾期】酣聚酒窖（HKJJ122）2026-09 账龄最老 151 天，逾期合计 ¥32,538.54（占该户 100.00%）｜建议：催收｜责任人：兀晨博",
+    "【应收账款逾期】孙飞龙客户（007）2026-09 账龄最老 45 天，逾期合计 ¥10,000.00（占该户 50.00%）｜建议：催收｜责任人：孙飞龙",
+]
+
+
+def _seed_legacy(db):
+    for m in LEGACY:
+        db.add(KPIAlert(entity_id=EID, kpi_id=KPI, alert_level="red", alert_message=m,
+                        alert_type="overdue", suggestion="催收", status="pending",
+                        assignee="李世昌"))
+    db.commit()
+    return len(LEGACY)
+
+
+def _cust(code, amt="1.00", name="客户", assignee="李世昌"):
+    return {"unit_code": code, "unit_name": name, "overdue_amt": amt, "total_amt": "1.00",
+            "oldest_days": 151, "pct_of_unit": 1.0, "level": "red",
+            "alert_message": "【应收账款逾期】%s（%s）%s 账龄最老 151 天，逾期合计 ¥%s（占该户 100.00%%）"
+                             "｜建议：催收｜责任人：%s" % (name, code, P, amt, assignee),
+            "suggestion": "催收", "assignee": assignee, "assignee_source": "unit", "emp_id": "E1"}
+
+
+def _agg(key="share_90plus", msg=None):
+    return {"key": key, "level": "red",
+            "metrics": {"share_90plus": 0.1556},
+            "alert_message": msg or ("【应收账款逾期】%s 借方应收户口径 90 天以上 ¥1,000.00，"
+                                     "占应收总额 15.56%%（红线 >25.00%%）｜建议：见规则配置" % P),
+            "suggestion": "见规则配置", "assignee": "李世昌", "assignee_source": "aggregate"}
+
+
+def _result(customers, aggs=None):
+    return {"period": P, "aggregates": aggs or [], "per_customer": customers}
+
+
+def _total(db):
+    return db.execute(text("select count(*) from kpi_alerts")).scalar()
+
+
+# ── T7 键生成/反解（纯函数）────────────────────────────────────────────
+def test_t7_key_gen_and_parse():
+    assert A._item_dedupe_key({"unit_code": "HKJJ053"}, P) == "2026-09|cust|HKJJ053"
+    assert A._item_dedupe_key({"key": "share_90plus"}, P) == "2026-09|agg|share_90plus"
+    assert A._item_dedupe_key({"key": "count_mom"}, P) == "2026-09|agg|count_mom"
+    assert A._item_dedupe_key({"key": "structure_90_150"}, P) == "2026-09|agg|structure"
+    assert A._item_dedupe_key({"key": "structure_150p"}, P) == "2026-09|agg|structure"
+    assert A._item_dedupe_key({}, P) is None
+    # 反解（历史兼容）
+    assert A._key_from_message(LEGACY[0]) == "2026-09|agg|share_90plus"
+    for m in LEGACY[1:]:
+        code = m.split("（")[1].split("）")[0]
+        assert A._key_from_message(m) == "2026-09|cust|%s" % code
+    assert A._key_from_message("完全无关的文案") is None
+
+
+# ── T1/T2 历史行零改 + 重跑幂等 ────────────────────────────────────────
+def test_t1_t2_legacy_untouched_and_idempotent(db):
+    n0 = _seed_legacy(db)
+    ids_before = [r[0] for r in db.execute(text("select id from kpi_alerts order by id")).all()]
+    snap_before = [tuple(r) for r in db.execute(text(
+        "select id, alert_message, dedupe_key, status, coalesce(assignee,'') from kpi_alerts order by id")).all()]
+
+    # 同语义口径 8 条（1 聚合 + 7 单户），文案由新代码生成（措辞与历史略有差异也可）
+    res = _result([_cust(c) for c in ("HKJJ053", "HKJJ120", "HKJJ156", "HKJJ166", "HKJJ175", "HKJJ122", "007")],
+                  [_agg()])
+    w1 = A.write_alerts(db, EID, KPI, res)
+    assert w1 == {"inserted": 0, "skipped_duplicate": 8}, w1
+
+    # 再跑一次（T2）
+    w2 = A.write_alerts(db, EID, KPI, res)
+    assert w2 == {"inserted": 0, "skipped_duplicate": 8}, w2
+
+    # 历史行零改动（R1 红线）
+    snap_after = [tuple(r) for r in db.execute(text(
+        "select id, alert_message, dedupe_key, status, coalesce(assignee,'') from kpi_alerts order by id")).all()]
+    assert snap_after == snap_before
+    assert _total(db) == n0
+    assert [r[0] for r in db.execute(text("select id from kpi_alerts order by id")).all()] == ids_before
+
+
+# ── T3 新户插入（负测）────────────────────────────────────────────────
+def test_t3_new_customer_inserts(db):
+    _seed_legacy(db)
+    n0 = _total(db)
+    res = _result([_cust("HKJJ999", amt="88,888.00", name="新户")])
+    w = A.write_alerts(db, EID, KPI, res)
+    assert w == {"inserted": 1, "skipped_duplicate": 0}, w
+    assert _total(db) == n0 + 1
+    row = db.query(KPIAlert).filter(KPIAlert.dedupe_key == "2026-09|cust|HKJJ999").one()
+    assert row.alert_type == "overdue" and row.assignee == "李世昌"
+
+
+# ── T4 同户金额变化不插（语义幂等）────────────────────────────────────
+def test_t4_amount_change_does_not_insert(db):
+    _seed_legacy(db)
+    res = _result([_cust("HKJJ053", amt="700,000.00", name="王敏")])
+    w = A.write_alerts(db, EID, KPI, res)
+    assert w == {"inserted": 0, "skipped_duplicate": 1}, w
+
+
+# ── T5 同批两条同键 → 1 插 1 跳 ───────────────────────────────────────
+def test_t5_same_key_in_one_batch(db):
+    n0 = _total(db)
+    res = _result([_cust("HKJJ888", amt="1.00"), _cust("HKJJ888", amt="2.00")])
+    w = A.write_alerts(db, EID, KPI, res)
+    assert w == {"inserted": 1, "skipped_duplicate": 1}, w
+    assert _total(db) == n0 + 1
+
+
+# ── T6 dry-run 与实插一致 ─────────────────────────────────────────────
+def test_t6_dry_run_matches_insert(db):
+    _seed_legacy(db)
+    res = _result([_cust("HKJJ053"), _cust("HKJJ777", amt="3.00")], [_agg()])
+    n_pred = A.count_would_insert(db, EID, KPI, res)
+    w = A.write_alerts(db, EID, KPI, res)
+    assert n_pred == w["inserted"] == 1, (n_pred, w)
+    # 写后再算 → 0
+    assert A.count_would_insert(db, EID, KPI, res) == 0
+
+
+# ── T7b 唯一索引语义：手工插同键第二次必须失败（并发兜底）─────────────
+def test_t7b_unique_index_enforced(db):
+    # conftest 的 create_all 不含 DDL 索引 ⇒ 测试内补建（生产由 ddl 脚本建）
+    db.execute(text("CREATE UNIQUE INDEX IF NOT EXISTS uk_kpi_alerts_dedupe_test "
+                    "ON kpi_alerts (entity_id, kpi_id, alert_type, dedupe_key)"))
+    db.commit()
+    db.add(KPIAlert(entity_id=EID, kpi_id=KPI, alert_message="a", alert_type="overdue",
+                    dedupe_key="2026-09|cust|DUP"))
+    db.commit()
+    db.add(KPIAlert(entity_id=EID, kpi_id=KPI, alert_message="b", alert_type="overdue",
+                    dedupe_key="2026-09|cust|DUP"))
+    with pytest.raises(Exception):
+        db.commit()
+    db.rollback()

backend/tests/test_map_connections_p1.py

@@ -0,0 +1,166 @@
+"""P1 四层因果结构化：map_connections 表 + /connections 读/写/analysis 回归
+
+任务 project-20261021-cma-map-connections-table-p1-fullstack
+覆盖：双格式解析（层:中文名 / 层-序号）、双写、读表回落 canvas、分析与告警不拦截。
+"""
+import pytest
+from fastapi.testclient import TestClient
+from sqlalchemy.orm import Session
+
+from app.models import StrategicMap, MapObjective, MapConnection
+from app.api.maps import parse_node_ref, resolve_node
+from tests.conftest import create_test_user, get_token_for_user, auth_header
+
+DIMS = [
+    {"key": "finance", "name": "财务层", "icon": "💰", "color": "#F56C6C",
+     "objectives": [{"name": "营业收入"}, {"name": "营业利润"}]},
+    {"key": "customer", "name": "客户层", "icon": "👥", "color": "#409EFF",
+     "objectives": [{"name": "客户留存"}]},
+    {"key": "process", "name": "流程层", "icon": "⚙️", "color": "#67C23A",
+     "objectives": [{"name": "交付效率"}, {"name": "质量闭环"}]},
+    {"key": "learning", "name": "学习层", "icon": "📚", "color": "#E6A23C",
+     "objectives": [{"name": "全员培训"}, {"name": "能力认证"}]},
+]
+
+
+def _seed_map(db: Session, title="P1连线测试", canvas=None) -> StrategicMap:
+    m = StrategicMap(entity_id=1, title=title, status="draft", dimensions=DIMS,
+                     canvas_data=canvas or {"connections": []})
+    db.add(m)
+    db.commit()
+    db.refresh(m)
+    for d in DIMS:
+        for i, o in enumerate(d["objectives"]):
+            db.add(MapObjective(entity_id=1, map_id=m.id, dimension_key=d["key"],
+                                name=o["name"], sort_order=i))
+    db.commit()
+    return m
+
+
+@pytest.fixture(autouse=True)
+def _clean(db: Session):
+    db.query(MapConnection).delete()
+    db.query(MapObjective).filter(MapObjective.map_id >= 900000).delete()
+    db.query(StrategicMap).filter(StrategicMap.id >= 900000).delete()
+    db.commit()
+    yield
+
+
+class TestNodeParse:
+    def test_colon_format(self):
+        r = parse_node_ref("learning:全员培训")
+        assert (r["layer"], r["kind"], r["value"], r["error"]) == ("learning", "name", "全员培训", None)
+
+    def test_index_format(self):
+        r = parse_node_ref("process-1")
+        assert (r["layer"], r["kind"], r["value"], r["error"]) == ("process", "index", 1, None)
+
+    def test_unknown_formats(self):
+        assert parse_node_ref("nonsense")["error"] == "node_format_unrecognized"
+        assert parse_node_ref("hr:某目标")["error"] == "unknown_layer:hr"
+        assert parse_node_ref("")["error"] == "node_empty"
+
+    def test_resolve_both_formats(self, db: Session):
+        m = _seed_map(db)
+        # 冒号中文名
+        layer, oid, note = resolve_node(db, m, "learning:能力认证")
+        assert (layer, note) == ("learning", None) and oid is not None
+        # 序号（dimensions 数组下标）
+        layer, oid2, note = resolve_node(db, m, "process-1")
+        assert (layer, note) == ("process", None)
+        obj = db.query(MapObjective).filter(MapObjective.id == oid2).first()
+        assert obj.name == "质量闭环"
+
+
+class TestConnectionApi:
+    def test_post_dual_write_and_get_table(self, client: TestClient, db: Session):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        m = _seed_map(db, title="P1双写")
+        mid = m.id
+
+        resp = client.post(f"/api/cma/maps/{mid}/connections", headers=auth_header(token),
+                           json={"from": "learning:全员培训", "to": "process:交付效率"})
+        assert resp.status_code == 200, resp.text
+        body = resp.json()
+        assert body["db_synced"] is True
+        assert body["connections"][0]["from"] == "learning:全员培训"
+
+        # canvas_data 保留（前端渲染主路径）
+        db.expire_all()
+        m2 = db.query(StrategicMap).filter(StrategicMap.id == mid).first()
+        assert m2.canvas_data["connections"][0]["to"] == "process:交付效率"
+        # 表内落行且层/目标解析正确
+        row = db.query(MapConnection).filter(MapConnection.map_id == mid).one()
+        assert (row.from_layer, row.to_layer) == ("learning", "process")
+        assert row.from_objective_id and row.to_objective_id
+
+        got = client.get(f"/api/cma/maps/{mid}/connections", headers=auth_header(token))
+        assert got.status_code == 200
+        assert got.json()["source"] == "table"
+        assert got.json()["total"] == 1
+
+    def test_index_format_persisted(self, client: TestClient, db: Session):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        m = _seed_map(db, title="P1序号格式")
+        resp = client.post(f"/api/cma/maps/{m.id}/connections", headers=auth_header(token),
+                           json={"from": "learning-0", "to": "process-1"})
+        assert resp.status_code == 200, resp.text
+        row = db.query(MapConnection).filter(MapConnection.map_id == m.id).one()
+        assert (row.from_node, row.from_layer) == ("learning-0", "learning")
+        assert row.from_objective_id and row.to_objective_id
+
+    def test_get_falls_back_to_canvas(self, client: TestClient, db: Session):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        m = _seed_map(db, title="P1回落",
+                      canvas={"connections": [{"from": "learning:全员培训", "to": "process:质量闭环"}]})
+        got = client.get(f"/api/cma/maps/{m.id}/connections", headers=auth_header(token))
+        assert got.status_code == 200
+        assert got.json()["source"] == "canvas_data"
+        assert got.json()["total"] == 1
+
+    def test_delete_removes_both(self, client: TestClient, db: Session):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        m = _seed_map(db, title="P1删除")
+        client.post(f"/api/cma/maps/{m.id}/connections", headers=auth_header(token),
+                    json={"from": "learning:全员培训", "to": "process:交付效率"})
+        resp = client.request("DELETE", f"/api/cma/maps/{m.id}/connections", headers=auth_header(token),
+                              json={"from": "learning:全员培训", "to": "process:交付效率"})
+        assert resp.status_code == 200, resp.text
+        assert resp.json()["db_synced"] is True
+        assert db.query(MapConnection).filter(MapConnection.map_id == m.id).count() == 0
+        db.expire_all()
+        m2 = db.query(StrategicMap).filter(StrategicMap.id == m.id).first()
+        assert m2.canvas_data["connections"] == []
+
+    def test_analysis_stats_reverse_and_broken(self, client: TestClient, db: Session):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        m = _seed_map(db, title="P1分析")
+        pairs = [
+            ("learning:全员培训", "process:交付效率"),   # 正常
+            ("process:质量闭环", "customer:客户留存"),   # 正常
+            ("customer:客户留存", "finance:营业收入"),   # 正常
+            ("finance:营业利润", "customer:客户留存"),   # 反向（告警不拦截）
+            ("finance:营业收入", "finance:营业利润"),    # 同层
+            ("learning:不存在目标", "process:交付效率"), # 断链（照样可写）
+        ]
+        for f, t in pairs:
+            r = client.post(f"/api/cma/maps/{m.id}/connections", headers=auth_header(token),
+                            json={"from": f, "to": t})
+            assert r.status_code == 200, r.text
+
+        a = client.get(f"/api/cma/maps/{m.id}/connections/analysis", headers=auth_header(token))
+        assert a.status_code == 200, a.text
+        d = a.json()
+        assert d["source"] == "table" and d["total"] == 6
+        assert d["layer_stats"]["learning"]["as_from"] == 2
+        assert d["layer_stats"]["finance"]["as_to"] == 2  # customer→finance 1 + finance→finance 1
+        assert d["cross_layer"]["learning->process"] == 2  # 正常 1 + 断链 1（层可解析、目标不可解析）
+        assert d["same_layer_count"] == 1
+        assert len(d["reversed"]) == 1 and d["reversed"][0]["from_layer"] == "finance"
+        assert any(b["side"] == "from" for b in d["broken"])
+        assert any("反向连线" in w for w in d["warnings"])

backend/tests/test_maps_review_hold_20261022.py

@@ -0,0 +1,137 @@
+"""P2 回归：GET /api/cma/maps/{id}/review?level=strategic 不再 500
+
+任务 project-20261022-cma-map-canvas-connections-not-drawn-fullstack（缺陷2）
+根因：maps.py 的 total_hold 只在「目标含 DATA_HOLD_KPIS」分支被 += 累加、从未初始化，
+      首个待核目标即 UnboundLocalError → 500（线上 map65 栈顶实测）。
+修复：初始化 total_hold = 0，并在 summary.hold 透出计数（纯增量，消费方 green/yellow/red/health_score 不受影响）。
+
+覆盖：
+  ① 含待核 KPI 的地图 → 200，summary.hold=1，该目标 data_hold=True 且 actual 不外泄
+  ② 普通地图 → 200，summary.hold=0
+  ③ dimensions 缺失（None）/ 空数组 → 200，不抛（total=0，health_score=0）
+"""
+import pytest
+from fastapi.testclient import TestClient
+from sqlalchemy.orm import Session
+
+from app.models import StrategicMap, MapObjective, KPIDefinition, KPIValue
+from app.api.maps import DATA_HOLD_KPIS
+from tests.conftest import create_test_user, get_token_for_user, auth_header, create_test_kpi
+
+HOLD_CODE = "C_CUST_CONCENTRATION"
+PLAIN_CODE = "TEST_REVIEW_PLAIN"
+
+
+def _seed_review_map(db: Session, dims, title="P2 review 回归", map_id=None) -> StrategicMap:
+    kwargs = dict(entity_id=1, title=title, status="active", dimensions=dims,
+                  canvas_data={"connections": []})
+    if map_id is not None:
+        kwargs["id"] = map_id
+    m = StrategicMap(**kwargs)
+    db.add(m)
+    db.commit()
+    db.refresh(m)
+    return m
+
+
+@pytest.fixture(autouse=True)
+def _clean(db: Session):
+    yield
+    db.rollback()
+    db.query(KPIValue).filter(KPIValue.kpi_id.in_(
+        [k.id for k in db.query(KPIDefinition).filter(
+            KPIDefinition.kpi_code.in_([HOLD_CODE, PLAIN_CODE])).all()]
+    )).delete(synchronize_session=False)
+    db.query(KPIDefinition).filter(KPIDefinition.kpi_code.in_([HOLD_CODE, PLAIN_CODE])).delete(
+        synchronize_session=False)
+    db.query(MapObjective).filter(MapObjective.map_id >= 900000).delete(synchronize_session=False)
+    db.query(StrategicMap).filter(StrategicMap.id >= 900000).delete(synchronize_session=False)
+    db.commit()
+
+
+def _hold_dims():
+    # 一个待核目标（含 C_CUST_CONCENTRATION）+ 一个普通目标
+    return [{
+        "key": "customer", "name": "客户层", "icon": "👥", "color": "#409EFF",
+        "objectives": [
+            {"name": "客户集中度", "kpis": [HOLD_CODE]},
+            {"name": "客户满意", "kpis": [PLAIN_CODE]},
+        ],
+    }]
+
+
+def _seed_kpis(db: Session):
+    hold = create_test_kpi(db, kpi_code=HOLD_CODE, kpi_name="客户集中度", dimension="customer",
+                           target_value=60.0, kpi_level="strategic", unit="%")
+    plain = create_test_kpi(db, kpi_code=PLAIN_CODE, kpi_name="客户满意", dimension="customer",
+                            target_value=90.0, kpi_level="strategic", unit="分")
+    return hold, plain
+
+
+class TestReviewNo500:
+    def test_data_hold_kpi_returns_200_and_counts_hold(self, client: TestClient, db: Session):
+        """含待核 KPI：修复前此处 UnboundLocalError → 500；修复后 200 且 summary.hold=1。"""
+        create_test_user(db)
+        token = get_token_for_user(client)
+        hold_kpi, plain_kpi = _seed_kpis(db)
+        db.add(KPIValue(kpi_id=hold_kpi.id, period="2026-09", actual_value=72.0, source_type="actual"))
+        db.add(KPIValue(kpi_id=plain_kpi.id, period="2026-09", actual_value=93.0, source_type="actual"))
+        db.commit()
+        m = _seed_review_map(db, _hold_dims(), map_id=900101)
+
+        resp = client.get(f"/api/cma/maps/{m.id}/review", params={"level": "strategic"},
+                          headers=auth_header(token))
+        assert resp.status_code == 200, resp.text
+        body = resp.json()
+        assert body["map_id"] == m.id
+        s = body["summary"]
+        assert s["hold"] == 1, s
+        assert s["total"] == 2, s
+        assert s["green"] == 1, s               # 待核目标不进任何档
+        assert isinstance(s["health_score"], (int, float))
+
+        objs = {o["name"]: o for o in body["dimensions"][0]["objectives"]}
+        held = objs["客户集中度"]
+        assert held["data_hold"] is True and held["hold_reason"]
+        assert held["level"] == "pending"
+        # 数据前置：待核目标的实测值不得外泄
+        assert all(k["actual_value"] is None for k in held["kpis"])
+        assert objs["客户满意"]["data_hold"] is False
+
+    def test_plain_map_returns_200_hold_zero(self, client: TestClient, db: Session):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        _seed_kpis(db)
+        m = _seed_review_map(db, _hold_dims(), title="P2 普通地图", map_id=900102)
+        # 去掉待核目标，只留普通目标
+        m.dimensions = [{"key": "customer", "name": "客户层", "icon": "", "color": "#409EFF",
+                         "objectives": [{"name": "客户满意", "kpis": [PLAIN_CODE]}]}]
+        db.commit()
+
+        resp = client.get(f"/api/cma/maps/{m.id}/review", params={"level": "strategic"},
+                          headers=auth_header(token))
+        assert resp.status_code == 200, resp.text
+        assert resp.json()["summary"]["hold"] == 0
+
+    @pytest.mark.parametrize("dims", [None, []])
+    def test_empty_dimensions_returns_200(self, client: TestClient, db: Session, dims):
+        create_test_user(db)
+        token = get_token_for_user(client)
+        m = _seed_review_map(db, dims, title="P2 空地图", map_id=900103)
+        resp = client.get(f"/api/cma/maps/{m.id}/review", params={"level": "strategic"},
+                          headers=auth_header(token))
+        assert resp.status_code == 200, resp.text
+        s = resp.json()["summary"]
+        assert s["total"] == 0 and s["hold"] == 0 and s["health_score"] == 0
+
+    def test_level_none_still_200(self, client: TestClient, db: Session):
+        """不传 level（向后兼容路径）同样不得 500。"""
+        create_test_user(db)
+        token = get_token_for_user(client)
+        hold_kpi, _ = _seed_kpis(db)
+        db.add(KPIValue(kpi_id=hold_kpi.id, period="2026-09", actual_value=72.0, source_type="actual"))
+        db.commit()
+        m = _seed_review_map(db, _hold_dims(), title="P2 无 level", map_id=900104)
+        resp = client.get(f"/api/cma/maps/{m.id}/review", headers=auth_header(token))
+        assert resp.status_code == 200, resp.text
+        assert resp.json()["summary"]["hold"] == 1

backend/tests/test_strategy_charter_20261021.py

@@ -0,0 +1,249 @@
+# -*- coding: utf-8 -*-
+"""战略基座（strategy_charter + /api/cma/charter）验收用例 — 2026-10-21
+
+任务：`project-20261021-cma-strategy-charter-p0-fullstack`（P0 战略基座）
+
+覆盖（每条都是硬断言，不做“大概通过”）：
+  A. 禁止示例数据：库内零行时 GET 200 + filled=false + data=null，且表内 rowcount=0；
+  B. 正常写入：PUT 落一行，六字段逐字回读（含 value_items 数组）；GET 再读 filled=true；
+  C. 账套隔离：e1/e2 各写各的，互相 GET 不到对方；token=e1 显式 body.entity_id=2 → 403；
+     GET ?entity_id=<对方> → 403（平台统一语义）；
+  D. 不可空：缺 vision_year / vision_measurable / mission / vision / values_text / value_items → 400；
+     显式 null → 400；空串/空白 → 400；
+  E. 值域：vision_year 非 4 位年份 → 400（int 2030 归一为 "2030"）；
+     value_items 非数组 / 条目缺 name|meaning / 含非白名单键 → 400；
+  F. 字段白名单：非白名单字段（含 id/updated_by/entity_id 手写审计字段）→ 400；
+  G. 鉴权/角色：无 token → 401；business 角色 PUT → 403（GET 仍 200）。
+
+本套件**不连生产库**（conftest 已换内存 SQLite），也不插入任何“示例章程”。
+"""
+import hashlib
+
+import pytest
+
+from app.models import Entity, StrategyCharter, User, UserEntity
+from tests.conftest import (
+    TEST_USER_PASSWORD,
+    auth_header,
+    create_test_user,
+)
+
+CHARTER = "/api/cma/charter"
+
+VALID = {
+    "mission": "让每一家中国企业的经营决策都有据可依",
+    "vision": "成为西北最被信任的管理会计数字化伙伴",
+    "vision_year": "2030",
+    "vision_measurable": "2030 年服务 300 家付费企业、续约率 ≥85%（年报复核）",
+    "values_text": "客户成功｜数据说话｜长期主义",
+    "value_items": [
+        {"name": "客户成功", "meaning": "客户指标不改善的方案，内部再漂亮也不上"},
+        {"name": "数据说话", "meaning": "决策冲突时以可复核口径为准，不以职级定输赢"},
+    ],
+}
+
+
+def _ensure_entity(db, eid: int, name: str) -> Entity:
+    ent = db.query(Entity).filter(Entity.id == eid).first()
+    if not ent:
+        ent = Entity(id=eid, name=name, short_name=name, status="active")
+        db.add(ent)
+        db.commit()
+        db.refresh(ent)
+    return ent
+
+
+def _user_with_entities(db, username: str, role: str = "ceo", entity_ids=(1,)) -> User:
+    user = db.query(User).filter(User.username == username).first()
+    if not user:
+        user = create_test_user(
+            db, username=username, role=role,
+            password_hash=hashlib.sha256(TEST_USER_PASSWORD.encode()).hexdigest())
+    have = {ue.entity_id for ue in db.query(UserEntity).filter(UserEntity.user_id == user.id).all()}
+    for eid in entity_ids:
+        if eid not in have:
+            db.add(UserEntity(user_id=user.id, entity_id=eid))
+    db.commit()
+    return user
+
+
+def _login(client, username: str, entity_id: int) -> str:
+    resp = client.post("/api/cma/auth/login", json={
+        "username": username, "password": TEST_USER_PASSWORD, "entity_id": entity_id,
+    })
+    assert resp.status_code == 200, resp.text
+    return resp.json()["token"]
+
+
+@pytest.fixture
+def ctx(client, db):
+    """两账套 + 两个角色用户（ceo@e1+e2 / business@e1）"""
+    _ensure_entity(db, 1, "测试酣客")
+    _ensure_entity(db, 2, "测试博海")
+    _user_with_entities(db, "ceo1", "ceo", (1, 2))
+    _user_with_entities(db, "biz1", "business", (1,))
+    return {
+        "e1": auth_header(_login(client, "ceo1", 1)),
+        "e2": auth_header(_login(client, "ceo1", 2)),
+        "biz": auth_header(_login(client, "biz1", 1)),
+    }
+
+
+# ── A. 禁止示例数据 + 空态契约 ────────────────────────────────────────────
+
+def test_get_empty_state_no_seed_rows(client, db, ctx):
+    resp = client.get(CHARTER, headers=ctx["e1"])
+    assert resp.status_code == 200, resp.text
+    body = resp.json()
+    assert body["filled"] is False
+    assert body["data"] is None
+    assert body["entity_id"] == 1
+    # 派单红线：禁止示例/默认文案 —— 表内零行
+    assert db.query(StrategyCharter).count() == 0
+
+
+# ── B. 正常写入 + 回读 ───────────────────────────────────────────────────
+
+def test_put_then_get_roundtrip(client, db, ctx):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=VALID)
+    assert resp.status_code == 200, resp.text
+    data = resp.json()["data"]
+    for k, v in VALID.items():
+        assert data[k] == v, k
+    assert data["entity_id"] == 1
+    assert data["updated_by_name"] == "测试管理员"
+
+    row = db.query(StrategyCharter).filter(StrategyCharter.entity_id == 1).one()
+    assert row.vision_year == "2030"
+    assert row.vision_measurable == VALID["vision_measurable"]
+    assert row.value_items == VALID["value_items"]
+
+    got = client.get(CHARTER, headers=ctx["e1"]).json()
+    assert got["filled"] is True
+    assert got["data"]["mission"] == VALID["mission"]
+    assert got["data"]["vision_year"] == "2030"
+
+
+def test_put_is_upsert_single_row_per_entity(client, db, ctx):
+    assert client.put(CHARTER, headers=ctx["e1"], json=VALID).status_code == 200
+    changed = dict(VALID, vision="改过的愿景", vision_year="2035")
+    assert client.put(CHARTER, headers=ctx["e1"], json=changed).status_code == 200
+    rows = db.query(StrategyCharter).filter(StrategyCharter.entity_id == 1).all()
+    assert len(rows) == 1, "每账套必须单行（唯一键 uq_strategy_charter_entity）"
+    assert rows[0].vision == "改过的愿景"
+    assert rows[0].vision_year == "2035"
+
+
+def test_int_vision_year_normalized_to_string(client, db, ctx):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, vision_year=2030))
+    assert resp.status_code == 200, resp.text
+    assert resp.json()["data"]["vision_year"] == "2030"
+
+
+# ── C. 账套隔离 ─────────────────────────────────────────────────────────
+
+def test_entity_isolation_both_directions(client, db, ctx):
+    e2_payload = dict(VALID, mission="博海使命", vision="博海愿景")
+    assert client.put(CHARTER, headers=ctx["e1"], json=VALID).status_code == 200
+    assert client.put(CHARTER, headers=ctx["e2"], json=e2_payload).status_code == 200
+
+    d1 = client.get(CHARTER, headers=ctx["e1"]).json()["data"]
+    d2 = client.get(CHARTER, headers=ctx["e2"]).json()["data"]
+    assert d1["entity_id"] == 1 and d1["mission"] == VALID["mission"]
+    assert d2["entity_id"] == 2 and d2["mission"] == "博海使命"
+    assert d1["mission"] != d2["mission"]
+    assert db.query(StrategyCharter).count() == 2
+
+
+def test_put_with_foreign_entity_id_in_body_forbidden(client, db, ctx):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, entity_id=2))
+    assert resp.status_code == 403, resp.text
+    assert "无权写入" in resp.json()["detail"]
+    assert db.query(StrategyCharter).count() == 0, "403 必须零写入"
+
+
+def test_get_with_explicit_other_entity_query_forbidden(client, ctx):
+    resp = client.get(f"{CHARTER}?entity_id=2", headers=ctx["e1"])
+    assert resp.status_code == 403, resp.text
+
+
+def test_entity_without_charter_sees_empty_state(client, db, ctx):
+    """e1 有章程、e2 没有 ⇒ e2 读到 filled=false（不串 e1 数据）"""
+    assert client.put(CHARTER, headers=ctx["e1"], json=VALID).status_code == 200
+    got = client.get(CHARTER, headers=ctx["e2"]).json()
+    assert got["filled"] is False and got["data"] is None and got["entity_id"] == 2
+
+
+# ── D. 不可空约束 ───────────────────────────────────────────────────────
+
+@pytest.mark.parametrize("field", ["vision_year", "vision_measurable", "mission", "vision",
+                                   "values_text", "value_items"])
+def test_required_fields_cannot_be_missing(client, ctx, field):
+    payload = {k: v for k, v in VALID.items() if k != field}
+    resp = client.put(CHARTER, headers=ctx["e1"], json=payload)
+    assert resp.status_code == 400, resp.text
+    assert field in resp.json()["detail"]
+
+
+@pytest.mark.parametrize("field", ["vision_year", "vision_measurable", "mission", "vision",
+                                   "values_text", "value_items"])
+def test_required_fields_cannot_be_null(client, ctx, field):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, **{field: None}))
+    assert resp.status_code == 400, resp.text
+
+
+@pytest.mark.parametrize("field", ["mission", "vision", "vision_measurable", "values_text"])
+def test_blank_text_rejected(client, ctx, field):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, **{field: "   "}))
+    assert resp.status_code == 400, resp.text
+    assert field in resp.json()["detail"]
+
+
+def test_empty_value_items_rejected(client, ctx):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, value_items=[]))
+    assert resp.status_code == 400, resp.text
+    assert "value_items" in resp.json()["detail"]
+
+
+# ── E. 值域 ─────────────────────────────────────────────────────────────
+
+@pytest.mark.parametrize("year", ["", "203", "20300", "abcd", 2030.5, True, "  "])
+def test_vision_year_must_be_four_digits(client, ctx, year):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, vision_year=year))
+    assert resp.status_code == 400, (year, resp.text)
+
+
+@pytest.mark.parametrize("items", [
+    "客户成功",                                   # 非数组
+    [{"name": "客户成功"}],                        # 缺 meaning
+    [{"name": "", "meaning": "x"}],                # name 空
+    [{"name": "客户成功", "meaning": "x", "extra": 1}],  # 非白名单键
+    [1, 2],                                        # 条目非对象
+])
+def test_value_items_shape(client, ctx, items):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, value_items=items))
+    assert resp.status_code == 400, (items, resp.text)
+
+
+# ── F. 字段白名单 ───────────────────────────────────────────────────────
+
+@pytest.mark.parametrize("bad", [{"entity_id_x": 2}, {"id": 99}, {"updated_by": 7},
+                                 {"verification_status": "verified"}])
+def test_unknown_fields_rejected(client, ctx, bad):
+    resp = client.put(CHARTER, headers=ctx["e1"], json=dict(VALID, **bad))
+    assert resp.status_code == 400, resp.text
+    assert "非白名单字段" in resp.json()["detail"]
+
+
+# ── G. 鉴权 / 角色 ─────────────────────────────────────────────────────
+
+def test_no_token_401(client, ctx):
+    assert client.get(CHARTER).status_code == 401
+    assert client.put(CHARTER, json=VALID).status_code == 401
+
+
+def test_business_role_can_read_but_not_write(client, db, ctx):
+    assert client.get(CHARTER, headers=ctx["biz"]).status_code == 200
+    resp = client.put(CHARTER, headers=ctx["biz"], json=VALID)
+    assert resp.status_code == 403, resp.text
+    assert db.query(StrategyCharter).count() == 0

deploy.sh

@@ -56,10 +56,16 @@ assert_pnpm_version() {
   fi
   actual="$(pnpm -v 2>/dev/null | tail -1 | tr -d '[:space:]' || true)"
   if [ "$actual" != "$expect" ]; then
-    echo "❌ pnpm 版本断言失败：package.json pin=$expect，实读 pnpm -v=${actual:-<空>}"
-    echo "   多为 corepack lastKnownGood 漂移（/root/.cache/node/corepack/lastKnownGood.json）"
-    echo "   修法：HOME=/root corepack prepare pnpm@$expect --activate ｜ 中止（生产未被触碰）"
-    exit 1
+    echo "⚠️ pnpm 版本断言不匹配：package.json pin=$expect，实读 pnpm -v=${actual:-<空>}"
+    echo "   尝试自愈：corepack prepare pnpm@$expect --activate（integrity 依 packageManager 的 sha512 校验）"
+    corepack prepare "pnpm@$expect" --activate >/dev/null 2>&1 || true
+    actual="$(pnpm -v 2>/dev/null | tail -1 | tr -d '[:space:]' || true)"
+    if [ "$actual" != "$expect" ]; then
+      echo "❌ pnpm 版本断言失败（自愈未成功）：pin=$expect，实读=${actual:-<空>}"
+      echo "   多为 corepack lastKnownGood 漂移或离线/缓存缺失；中止（生产未被触碰）"
+      exit 1
+    fi
+    echo "    ✓ 自愈成功：pnpm $actual == package.json pin $expect"
   fi
   echo "    ✓ pnpm 版本断言通过：pnpm $actual == package.json pin $expect"
 }

frontend/components.d.ts

@@ -15,6 +15,7 @@ declare module 'vue' {
     BotSourceTag: typeof import('./src/components/BotSourceTag.vue')['default']
     BudgetCollectPanel: typeof import('./src/components/BudgetCollectPanel.vue')['default']
     BulletChart: typeof import('./src/components/charts/BulletChart.vue')['default']
+    CaliberBadge: typeof import('./src/components/CaliberBadge.vue')['default']
     CashClassifyPanel: typeof import('./src/components/CashClassifyPanel.vue')['default']
     ConnectionLines: typeof import('./src/components/strategy/ConnectionLines.vue')['default']
     ConnectionManagerDialog: typeof import('./src/components/map-canvas/ConnectionManagerDialog.vue')['default']

frontend/src/api/index.ts

@@ -546,4 +546,32 @@ export const treasuryApi = {
   listHits: (params?: any) => api.get('/treasury/control-hits', { params }),
 }
 
-export default api
+// ── 应收账龄看板（T2 2026-10-20：逐户·六档·三口径禁合计，缺期不补 0）──
+export const arAgingApi = {
+  meta: (params?: any) => api.get('/ar-aging/meta', { params }),
+  trend: (params?: any) => api.get('/ar-aging/trend', { params }),
+  overview: (params?: any) => api.get('/ar-aging/overview', { params }),
+  customers: (params?: any) => api.get('/ar-aging/customers', { params }),
+  exportUrl: (params: { period?: string; caliber?: string; overdue_only?: boolean }) => {
+    const q = new URLSearchParams()
+    if (params.period) q.set('period', params.period)
+    if (params.caliber) q.set('caliber', params.caliber)
+    if (params.overdue_only) q.set('overdue_only', 'true')
+    return `/api/cma/ar-aging/customers/export?${q.toString()}`
+  },
+}
+
+// ── 战略基座：战略章程（P0 2026-10-21）──
+export const charterApi = {
+  get: () => api.get('/charter'),
+  save: (payload: {
+    mission: string
+    vision: string
+    vision_year: string
+    vision_measurable: string
+    values_text: string
+    value_items: Array<{ name: string; meaning: string }>
+  }) => api.put('/charter', payload),
+}
+
+export default api

frontend/src/components/strategy-map/MapEditBar.vue

@@ -90,9 +90,32 @@
         回全景
       </el-button>
 
-      <el-button size="small" text @click="$emit('toggle-tree')">
+      <el-button size="small" text @click="$emit('toggle-tree')" :title="treeOpen ? '收起左侧目标树（画布全宽）' : '展开左侧目标树'">
         {{ treeOpen ? '收起目标树' : '展开目标树' }}
       </el-button>
+
+      <!-- C 方案：聚焦开关（全景 = 默认四层等宽） -->
+      <div class="eb-chips eb-focus" role="group" aria-label="聚焦层">
+        <button
+          type="button"
+          class="eb-chip"
+          :class="{ on: !focusLayer }"
+          title="四层全景（默认）"
+          @click="$emit('focus-layer', null)"
+        >全景</button>
+        <button
+          v-for="cfg in layerConfigs"
+          :key="'focus-' + cfg.key"
+          type="button"
+          class="eb-chip"
+          :class="{ on: focusLayer === cfg.key }"
+          :title="focusLayer === cfg.key ? '再次点击退出聚焦' : '聚焦' + cfg.label"
+          @click="$emit('focus-layer', focusLayer === cfg.key ? null : cfg.key)"
+        >
+          <span class="layer-dot" :style="{ background: cfg.color }"></span>
+          {{ cfg.short }}
+        </button>
+      </div>
     </div>
   </div>
 </template>
@@ -105,6 +128,8 @@ defineProps<{
   layerConfigs: { key: string; label: string; short: string; color: string }[]
   canUndo: boolean
   treeOpen: boolean
+  /** C 方案聚焦层：null = 四层全景 */
+  focusLayer?: string | null
 }>()
 
 const emit = defineEmits<{
@@ -117,6 +142,7 @@ const emit = defineEmits<{
   (e: 'fit'): void
   (e: 'toggle-tree'): void
   (e: 'completeness'): void
+  (e: 'focus-layer', key: string | null): void
 }>()
 
 const filterOptions: { value: 'all' | 'risk' | 'nodata'; label: string }[] = [

frontend/src/components/strategy-map/MapSummaryBar.vue

@@ -1,89 +1,150 @@
 <template>
   <section class="map-summary" aria-label="战略地图摘要">
-    <!-- ① 营业收入：分业务线全额售价（总额法）· 禁净额 / 禁合并 -->
-    <div class="sum-card sum-card--wide">
-      <div class="sum-head">
-        <span class="sum-label">营业收入</span>
-        <span v-if="revenue && revenue.caliber_label" class="sum-badge calm">{{ revenue.caliber_label }}</span>
+    <!-- C 方案（2026-10-23）：汇总条压成单行（≤44px），5 项指标同排；明细收进「明细」展开位 -->
+    <div class="ms-row">
+      <!-- ① 营业收入：分业务线全额售价（总额法）· 禁净额 / 禁合并 -->
+      <div class="ms-metric ms-metric--rev">
+        <span class="ms-label">营业收入</span>
+        <span class="ms-value mono" :class="{ muted: revenueTotal == null }">
+          <template v-if="revenueTotal != null">{{ fmtNum(revenueTotal) }}<i class="ms-unit">万元</i></template>
+          <template v-else>—<i class="ms-unit">无数据</i></template>
+        </span>
+        <span v-if="revenue && revenue.caliber_label" class="ms-badge calm">{{ revenue.caliber_label }}</span>
       </div>
-      <div v-if="revenueLines.length" class="rev-lines">
-        <div v-for="ln in revenueLines" :key="ln.key" class="rev-line">
-          <span class="rev-line-label">{{ ln.label }}</span>
-          <span class="rev-line-value mono" :class="{ muted: ln.amount == null }">
-            <template v-if="ln.amount != null">{{ fmtNum(ln.amount) }}<i class="sum-unit">万元</i></template>
-            <template v-else>—<i class="sum-unit">无数据</i></template>
-          </span>
+
+      <span class="ms-sep" aria-hidden="true"></span>
+
+      <!-- ② 毛利率 -->
+      <div class="ms-metric">
+        <span class="ms-label">毛利率</span>
+        <span class="ms-value mono" :class="{ muted: gmValue == null }">
+          <template v-if="gmValue != null">{{ gmValue.toFixed(2) }}<i class="ms-unit">%</i></template>
+          <template v-else>—<i class="ms-unit">无数据</i></template>
+        </span>
+        <span v-if="grossMargin && grossMargin.data_tag" class="ms-badge warn">{{ grossMargin.data_tag }}</span>
+      </div>
+
+      <span class="ms-sep" aria-hidden="true"></span>
+
+      <!-- ③ 战略目标达成率 -->
+      <div class="ms-metric">
+        <span class="ms-label">战略目标达成率</span>
+        <span class="ms-value mono" :class="achievement == null ? 'muted' : 'lv-' + achievementLevel">
+          <template v-if="achievement != null">{{ Math.round(achievement) }}<i class="ms-unit">%</i></template>
+          <template v-else>—<i class="ms-unit">无数据</i></template>
+        </span>
+      </div>
+
+      <span class="ms-sep" aria-hidden="true"></span>
+
+      <!-- ④ 风险目标数 -->
+      <div class="ms-metric">
+        <span class="ms-label">风险目标</span>
+        <span class="ms-value mono" :class="riskCount == null ? 'muted' : (riskCount > 0 ? 'lv-red' : 'lv-green')">
+          <template v-if="riskCount != null">{{ riskCount }}</template>
+          <template v-else>—<i class="ms-unit">无数据</i></template>
+        </span>
+      </div>
+
+      <span class="ms-sep" aria-hidden="true"></span>
+
+      <!-- ⑤ 目标·连线（当前画布规模） -->
+      <div class="ms-metric ms-metric--scale">
+        <span class="ms-label">目标·连线</span>
+        <span class="ms-value mono">
+          <template v-if="objectivesTotal != null">{{ objectivesTotal }}</template>
+          <template v-else>—</template>
+          <i class="ms-unit">个目标</i>
+          <span class="ms-div">/</span>
+          <template v-if="connectionsTotal != null">{{ connectionsTotal }}</template>
+          <template v-else>—</template>
+          <i class="ms-unit">条连线</i>
+        </span>
+      </div>
+
+      <button
+        type="button"
+        class="ms-detail-btn"
+        :aria-expanded="detailOpen ? 'true' : 'false'"
+        @click="detailOpen = !detailOpen"
+      >
+        <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" aria-hidden="true">
+          <path :d="detailOpen ? 'M6 15l6-6 6 6' : 'M6 9l6 6 6-6'" />
+        </svg>
+        {{ detailOpen ? '收起' : '明细' }}
+      </button>
+    </div>
+
+    <!-- 明细（默认收起；展开后保留分线金额 / 口径 / 管理指标全部信息，不丢信息） -->
+    <div v-if="detailOpen" class="ms-detail">
+      <div class="ms-card">
+        <div class="ms-detail-head">
+          <span class="ms-label">营业收入 · 分业务线</span>
+          <span v-if="revenue && revenue.caliber_label" class="ms-badge calm">{{ revenue.caliber_label }}</span>
         </div>
-        <div class="rev-line rev-line--total">
-          <span class="rev-line-label">{{ revenueTotalLabel }}</span>
-          <span class="rev-line-value mono">
-            <template v-if="revenueTotal != null">{{ fmtNum(revenueTotal) }}<i class="sum-unit">万元</i></template>
-            <template v-else>—<i class="sum-unit">无数据</i></template>
-          </span>
+        <div v-if="revenueLines.length" class="rev-lines">
+          <div v-for="ln in revenueLines" :key="ln.key" class="rev-line">
+            <span class="rev-line-label">{{ ln.label }}</span>
+            <span class="rev-line-value mono" :class="{ muted: ln.amount == null }">
+              <template v-if="ln.amount != null">{{ fmtNum(ln.amount) }}<i class="ms-unit">万元</i></template>
+              <template v-else>—<i class="ms-unit">无数据</i></template>
+            </span>
+          </div>
+          <div class="rev-line rev-line--total">
+            <span class="rev-line-label">{{ revenueTotalLabel }}</span>
+            <span class="rev-line-value mono">
+              <template v-if="revenueTotal != null">{{ fmtNum(revenueTotal) }}<i class="ms-unit">万元</i></template>
+              <template v-else>—<i class="ms-unit">无数据</i></template>
+            </span>
+          </div>
+        </div>
+        <div v-else class="ms-value muted">—<i class="ms-unit">无数据</i></div>
+        <div class="ms-sub">{{ revenuePeriod }}</div>
+      </div>
+
+      <div class="ms-card">
+        <div class="ms-detail-head"><span class="ms-label">毛利率口径</span></div>
+        <div class="ms-value mono" :class="{ muted: gmValue == null }">
+          <template v-if="gmValue != null">{{ gmValue.toFixed(2) }}<i class="ms-unit">%</i></template>
+          <template v-else>—<i class="ms-unit">无数据</i></template>
+        </div>
+        <div class="ms-sub">{{ gmSub }}</div>
+      </div>
+
+      <div class="ms-card">
+        <div class="ms-detail-head"><span class="ms-label">达成率 / 风险分布</span></div>
+        <div class="ms-sub">
+          <template v-if="achievement != null">{{ achievementCount }} 个目标有实测值 · 其余无数据</template>
+          <template v-else>所有目标均无可用 KPI 实测值</template>
+        </div>
+        <div class="ms-sub">
+          <template v-if="summary">达标 {{ summary.green }} · 关注 {{ summary.yellow }} · 风险 {{ summary.red }} · 共 {{ summary.total }} 个目标</template>
+          <template v-else>战略回顾数据未加载</template>
         </div>
       </div>
-      <div v-else class="sum-value muted">—<i class="sum-unit">无数据</i></div>
-      <div class="sum-sub">{{ revenuePeriod }}</div>
-    </div>
 
-    <!-- ② 毛利率：新式口径（折扣后净进价），历史缺票期带标识 -->
-    <div class="sum-card">
-      <div class="sum-head">
-        <span class="sum-label">毛利率</span>
-        <span v-if="grossMargin && grossMargin.data_tag" class="sum-badge warn">{{ grossMargin.data_tag }}</span>
+      <div class="ms-card">
+        <div class="ms-detail-head"><span class="ms-label">口径变更 / 管理指标</span></div>
+        <div class="ms-foot">
+          <span v-if="switchInfo && switchInfo.has_switch" class="ms-badge warn" :title="switchInfo.note || ''">
+            {{ switchInfo.label }}<template v-if="switchInfo.effective_date"> · 自 {{ switchInfo.effective_date }}</template>
+            <em v-if="switchInfo.comparability"> · {{ switchInfo.comparability }}</em>
+          </span>
+          <span v-for="m in management" :key="m.key" class="mgmt" :title="m.caliber_note || ''">
+            <span class="mgmt-label">{{ m.label }}</span>
+            <span class="mgmt-value mono">{{ m.value == null ? '—' : m.value.toFixed(2) + '%' }}</span>
+            <span class="mgmt-tag">{{ m.tag }}</span>
+            <span v-if="m.abnormal" class="mgmt-abnormal">异常·须核查</span>
+          </span>
+          <span v-if="!(switchInfo && switchInfo.has_switch) && management.length === 0" class="ms-sub">无口径变更与管理指标</span>
+        </div>
       </div>
-      <div class="sum-value mono" :class="{ muted: gmValue == null }">
-        <template v-if="gmValue != null">{{ gmValue.toFixed(2) }}<i class="sum-unit">%</i></template>
-        <template v-else>—<i class="sum-unit">无数据</i></template>
-      </div>
-      <div class="sum-sub">{{ gmSub }}</div>
-    </div>
-
-    <!-- ③ 战略目标达成率 -->
-    <div class="sum-card">
-      <div class="sum-head"><span class="sum-label">战略目标达成率</span></div>
-      <div class="sum-value mono" :class="achievement == null ? 'muted' : 'lv-' + achievementLevel">
-        <template v-if="achievement != null">{{ Math.round(achievement) }}<i class="sum-unit">%</i></template>
-        <template v-else>—<i class="sum-unit">无数据</i></template>
-      </div>
-      <div class="sum-sub">
-        <template v-if="achievement != null">{{ achievementCount }} 个目标有实测值 · 其余无数据</template>
-        <template v-else>所有目标均无可用 KPI 实测值</template>
-      </div>
-    </div>
-
-    <!-- ④ 风险目标数 -->
-    <div class="sum-card">
-      <div class="sum-head"><span class="sum-label">风险目标</span></div>
-      <div class="sum-value mono" :class="riskCount == null ? 'muted' : (riskCount > 0 ? 'lv-red' : 'lv-green')">
-        <template v-if="riskCount != null">{{ riskCount }}</template>
-        <template v-else>—<i class="sum-unit">无数据</i></template>
-      </div>
-      <div class="sum-sub">
-        <template v-if="summary">达标 {{ summary.green }} · 关注 {{ summary.yellow }} · 共 {{ summary.total }} 个目标</template>
-        <template v-else>战略回顾数据未加载</template>
-      </div>
-    </div>
-
-    <!-- ⑤ 口径变更标识位 + 管理指标（非会计科目，单列、不占主位） -->
-    <div class="sum-foot">
-      <span v-if="switchInfo && switchInfo.has_switch" class="sum-badge warn" :title="switchInfo.note || ''">
-        {{ switchInfo.label }}<template v-if="switchInfo.effective_date"> · 自 {{ switchInfo.effective_date }}</template>
-        <em v-if="switchInfo.comparability"> · {{ switchInfo.comparability }}</em>
-      </span>
-      <span v-for="m in management" :key="m.key" class="mgmt" :title="m.caliber_note || ''">
-        <span class="mgmt-label">{{ m.label }}</span>
-        <span class="mgmt-value mono">{{ m.value == null ? '—' : m.value.toFixed(2) + '%' }}</span>
-        <span class="mgmt-tag">{{ m.tag }}</span>
-        <span v-if="m.abnormal" class="mgmt-abnormal">异常·须核查</span>
-        <span class="mgmt-caliber">{{ fmtNum(m.numerator) }} ÷ {{ fmtNum(m.denominator) }}<template v-if="m.period"> · {{ m.period }}</template></span>
-      </span>
     </div>
   </section>
 </template>
 
 <script setup lang="ts">
-import { computed } from 'vue'
+import { computed, ref } from 'vue'
 
 type RevLine = { key: string; label: string; amount: number | null; unit?: string; missing_reason?: string | null }
 
@@ -122,8 +183,14 @@ const props = defineProps<{
   summary: { total: number; green: number; yellow: number; red: number; health_score?: number } | null
   switchInfo?: { has_switch?: boolean; label?: string; effective_date?: string | null; comparability?: string; note?: string } | null
   loading?: boolean
+  /** ⑤ 目标·连线：当前地图目标数（null ⇒ 显式「—」） */
+  objectivesTotal?: number | null
+  /** ⑤ 目标·连线：当前地图连线数（null ⇒ 显式「—」） */
+  connectionsTotal?: number | null
 }>()
 
+const detailOpen = ref(false)
+
 const revenueLines = computed<RevLine[]>(() => (props.revenue?.lines || []).filter(l => !!l && !!l.key))
 /** 合计 = 各线相加的复算校验位，不与分线并列成第二套口径 */
 const revenueTotal = computed<number | null>(() => {
@@ -163,92 +230,125 @@ function fmtNum(v: number | null | undefined): string {
 </script>
 
 <style scoped>
+/* C 方案：单行汇总条（行高 40px + 上下 1px 边框 ≤44px） */
 .map-summary {
-  display: grid;
-  grid-template-columns: repeat(4, minmax(0, 1fr));
-  gap: 10px;
   flex-shrink: 0;
-  margin-bottom: 10px;
-}
-.sum-card {
+  margin-bottom: 8px;
   background: var(--bh-bg-card, #fff);
   border: 1px solid var(--bh-border, #E5E7EB);
   border-radius: 8px;
-  padding: 10px 14px;
+  padding: 0 10px;
+}
+.ms-row {
+  display: flex;
+  align-items: center;
+  gap: 10px;
+  height: 40px;
   min-width: 0;
 }
-.sum-card--wide { grid-column: span 2; }
-.sum-head { display: flex; align-items: center; gap: 8px; min-width: 0; }
-.sum-label {
+.ms-metric {
+  display: flex;
+  align-items: baseline;
+  gap: 6px;
+  min-width: 0;
+  white-space: nowrap;
+}
+.ms-metric--rev { flex: 0 1 auto; overflow: hidden; }
+.ms-label {
   font-size: 12px;
   color: var(--bh-text-secondary, #64748B);
   letter-spacing: .02em;
-  white-space: nowrap;
+  flex-shrink: 0;
 }
-.sum-badge {
+.ms-value {
+  font-size: 17px;
+  font-weight: 700;
+  line-height: 1.1;
+  color: var(--bh-text-primary, #0F172A);
+  white-space: nowrap;
+  overflow: hidden;
+  text-overflow: ellipsis;
+}
+.ms-value.muted { font-size: 13px; color: var(--bh-text-muted, #94A3B8); font-weight: 500; }
+.ms-unit {
+  font-size: 10px;
+  font-weight: 500;
+  font-style: normal;
+  color: var(--bh-text-secondary, #64748B);
+  margin-left: 2px;
+}
+.ms-div { color: var(--bh-text-muted, #94A3B8); margin: 0 2px; }
+.ms-sep { width: 1px; height: 18px; background: var(--bh-border, #E5E7EB); flex-shrink: 0; }
+.ms-badge {
   font-size: 10px;
   line-height: 1.5;
-  padding: 1px 6px;
+  padding: 0 5px;
   border-radius: 4px;
+  border: 1px solid transparent;
   white-space: nowrap;
-  overflow: hidden;
-  text-overflow: ellipsis;
+  flex-shrink: 0;
 }
-.sum-badge.calm { color: #635BFF; background: rgba(99, 91, 255, .08); }
-.sum-badge.warn { color: #B45309; background: #FEF3C7; }
-.sum-badge.warn em { font-style: normal; }
-.sum-value {
-  margin-top: 4px;
-  font-size: 22px;
-  font-weight: 700;
-  line-height: 1.2;
-  color: var(--bh-text-primary, #0F172A);
-  display: flex;
-  align-items: baseline;
-  gap: 3px;
-}
-.sum-value.muted { font-size: 15px; color: var(--bh-text-muted, #94A3B8); font-weight: 500; }
-.sum-unit { font-size: 11px; font-weight: 500; font-style: normal; color: var(--bh-text-secondary, #64748B); margin-left: 2px; }
-.sum-sub {
-  margin-top: 4px;
-  font-size: 11px;
-  color: var(--bh-text-muted, #94A3B8);
-  white-space: nowrap;
-  overflow: hidden;
-  text-overflow: ellipsis;
-}
-.rev-lines { margin-top: 4px; display: flex; flex-direction: column; gap: 2px; }
-.rev-line { display: flex; align-items: baseline; justify-content: space-between; gap: 10px; }
-.rev-line-label { font-size: 12px; color: var(--bh-text-secondary, #64748B); }
-.rev-line-value { font-size: 18px; font-weight: 700; color: var(--bh-text-primary, #0F172A); }
-.rev-line-value.muted { font-size: 14px; font-weight: 500; color: var(--bh-text-muted, #94A3B8); }
-.rev-line--total { border-top: 1px dashed var(--bh-border, #E5E7EB); padding-top: 3px; margin-top: 2px; }
-.rev-line--total .rev-line-label,
-.rev-line--total .rev-line-value { font-size: 12px; font-weight: 500; color: var(--bh-text-secondary, #64748B); }
-.sum-foot {
-  grid-column: 1 / -1;
-  display: flex;
-  flex-wrap: wrap;
+.ms-badge.calm { color: #0369A1; background: #F0F9FF; border-color: #BAE6FD; }
+.ms-badge.warn { color: #B45309; background: #FFFBEB; border-color: #FDE68A; }
+.ms-badge.warn em { font-style: normal; }
+.ms-detail-btn {
+  margin-left: auto;
+  flex-shrink: 0;
+  display: inline-flex;
   align-items: center;
-  gap: 6px 16px;
-  padding: 6px 12px;
-  border: 1px dashed var(--bh-border, #E5E7EB);
-  border-radius: 6px;
+  gap: 3px;
+  height: 24px;
+  padding: 0 8px;
   font-size: 11px;
-  color: var(--bh-text-muted, #94A3B8);
+  color: var(--bh-text-secondary, #64748B);
+  background: transparent;
+  border: 1px solid var(--bh-border, #E5E7EB);
+  border-radius: 6px;
+  cursor: pointer;
 }
-.mgmt { display: inline-flex; align-items: baseline; gap: 6px; }
+.ms-detail-btn:hover { color: #0F172A; border-color: #CBD5E1; }
+.ms-detail-btn svg { width: 12px; height: 12px; }
+
+/* 明细展开区 */
+.ms-detail {
+  display: grid;
+  grid-template-columns: repeat(auto-fit, minmax(220px, 1fr));
+  gap: 10px;
+  padding: 0 0 10px;
+}
+.ms-card {
+  border: 1px solid var(--bh-border, #E5E7EB);
+  border-radius: 8px;
+  padding: 8px 12px;
+  min-width: 0;
+  background: var(--bh-bg-subtle, #FCFCFD);
+}
+.ms-detail-head {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+  margin-bottom: 4px;
+}
+.rev-lines { display: flex; flex-direction: column; gap: 2px; }
+.rev-line { display: flex; align-items: baseline; justify-content: space-between; gap: 8px; min-width: 0; }
+.rev-line-label { font-size: 11px; color: var(--bh-text-secondary, #64748B); }
+.rev-line-value { font-size: 12px; color: var(--bh-text-primary, #0F172A); }
+.rev-line-value.muted { color: var(--bh-text-muted, #94A3B8); }
+.rev-line--total { border-top: 1px dashed var(--bh-border, #E5E7EB); margin-top: 2px; padding-top: 2px; }
+.ms-sub { margin-top: 4px; font-size: 11px; color: var(--bh-text-muted, #94A3B8); line-height: 1.5; }
+.ms-foot { display: flex; flex-wrap: wrap; align-items: center; gap: 6px 12px; }
+.mgmt { display: inline-flex; align-items: baseline; gap: 4px; font-size: 11px; }
 .mgmt-label { color: var(--bh-text-secondary, #64748B); }
-.mgmt-value { font-size: 13px; font-weight: 700; color: var(--bh-text-primary, #0F172A); }
-.mgmt-tag { font-size: 10px; padding: 1px 5px; border-radius: 4px; background: #F1F5F9; color: #475569; }
-.mgmt-abnormal { color: #B91C1C; }
-.mgmt-caliber { color: var(--bh-text-muted, #94A3B8); }
+.mgmt-value { color: var(--bh-text-primary, #0F172A); font-weight: 600; }
+.mgmt-tag { font-size: 10px; padding: 0 4px; border-radius: 4px; background: #F1F5F9; color: #475569; }
+.mgmt-abnormal { font-size: 10px; color: #B91C1C; }
 .mono { font-family: 'IBM Plex Mono', Consolas, 'SFMono-Regular', monospace; font-variant-numeric: tabular-nums; }
 .lv-green { color: #00C896; }
 .lv-yellow { color: #F59E0B; }
 .lv-red { color: #EF4444; }
-@media (max-width: 1200px) {
-  .map-summary { grid-template-columns: repeat(2, minmax(0, 1fr)); }
-  .sum-card--wide { grid-column: span 2; }
+
+/* 窄屏：单行改为可横向滚动，保留全部 5 项、不换行不截断 */
+@media (max-width: 1280px) {
+  .ms-row { overflow-x: auto; }
 }
 </style>

frontend/src/components/strategy-map/StrategyLayer.vue

@@ -1,6 +1,17 @@
 <template>
   <div class="strategy-layer" :class="`layer-${layerKey}`" :style="{ borderColor: color }">
-    <div class="layer-header" :style="{ background: color }">
+    <div
+      class="layer-header"
+      :class="{ 'is-focusable': true, 'is-focused': focused }"
+      :style="{ background: color }"
+      :title="focused ? '再次点击退出聚焦，恢复四层等宽' : '点击聚焦该层（其余三层压成单行摘要）'"
+      role="button"
+      tabindex="0"
+      :aria-pressed="focused ? 'true' : 'false'"
+      @click="$emit('focus')"
+      @keydown.enter.prevent="$emit('focus')"
+      @keydown.space.prevent="$emit('focus')"
+    >
       <button
         type="button"
         class="layer-collapse-btn"
@@ -14,6 +25,12 @@
       </button>
       <span class="layer-icon">{{ icon }}</span>
       <span class="layer-name">{{ label }}</span>
+      <!-- 压缩态（C 方案聚焦模式）：保「目标数 + 达成率色条」，整层信息不全隐 -->
+      <span v-if="compact" class="layer-mini" :aria-label="'该层达成率 ' + ratioText">
+        <span class="mini-track"><i class="mini-fill" :style="{ width: ratioBarWidth }"></i></span>
+        <span class="mini-ratio mono">{{ ratioText }}</span>
+      </span>
+      <span v-if="focused" class="layer-focus-flag">聚焦中</span>
       <span class="layer-objective-count">{{ objectives.length }} 个目标</span>
       <el-dropdown trigger="click" @command="onLayerAction">
         <el-button class="layer-more-btn" size="small" text>
@@ -42,7 +59,7 @@
           <div class="layer-node-wrapper">
             <MapNode
               :key="`${layerKey}-${index}`"
-              :ref="el => setNodeRef(`${layerKey}-${index}`, el)"
+              :ref="el => setNodeRef(`${layerKey}-${index}`, el, element)"
               :objective="element"
               :node-key="`${layerKey}-${index}`"
               :index="index"
@@ -90,7 +107,7 @@
 </template>
 
 <script setup lang="ts">
-import { ref, reactive } from 'vue'
+import { ref, reactive, computed } from 'vue'
 import { MoreFilled } from '@element-plus/icons-vue'
 import draggable from 'vuedraggable'
 import MapNode from './MapNode.vue'
@@ -115,6 +132,12 @@ const props = defineProps<{
   dimmedKeys?: string[]
   /** 层折叠（仅隐藏层体，保留层头） */
   collapsed?: boolean
+  /** C 方案聚焦模式：其它层压成单行（层头 + 目标数 + 达成率色条） */
+  compact?: boolean
+  /** 该层达成率均值（无实测值传 null，界面显式「—」，不用示例数兜底） */
+  layerRatio?: number | null
+  /** 本层是否处于聚焦展开态 */
+  focused?: boolean
   getNodeLevel: (key: string) => string
   getNodeProgress: (key: string) => any
 }>()
@@ -138,14 +161,60 @@ const emit = defineEmits<{
   (e: 'edit-kr', payload: { dimKey: string; idx: number; obj: any; ki: number; kr: any }): void
   (e: 'delete-kr', payload: { dimKey: string; idx: number; obj: any; ki: number; kr: any }): void
   (e: 'add-kr', payload: { dimKey: string; idx: number; obj: any }): void
+  (e: 'focus'): void
 }>()
 
+/** 压缩态达成率文案：无实测值显式「—」，禁止示例数字兜底 */
+const ratioText = computed(() => {
+  const r = props.layerRatio
+  return r == null || Number.isNaN(Number(r)) ? '—' : `${Math.round(Number(r))}%`
+})
+/** 压缩态色条宽度（0–100%，无数据时 0 宽 + 灰色轨道） */
+const ratioBarWidth = computed(() => {
+  const r = Number(props.layerRatio)
+  if (props.layerRatio == null || Number.isNaN(r)) return '0%'
+  return `${Math.max(0, Math.min(100, Math.round(r)))}%`
+})
+
 const bodyRef = ref<HTMLElement | null>(null)
 
 // Node DOM refs for line calculation (reactive so parent can aggregate)
 const nodeRefs = reactive<Record<string, any>>({})
-function setNodeRef(key: string, el: any) {
-  if (el) nodeRefs[key] = el
+/**
+ * 双键注册：`${layerKey}-${index}`（渲染键，历史契约）+ `${layerKey}:${目标名}`（数据键）。
+ * canvas_data.connections / map_connections 里的节点标识形如 `learning:贡献率思维全员培训`，
+ * 只注册渲染键会导致 ConnectionLines 全部取不到 DOM → 一条线都不画（P2 缺陷）。
+ * 同层同名目标：先到先得（与数据侧 `dim:name` 本身的歧义语义保持一致）。
+ * 卸载（el=null）时按 key 反查清理自己注册的数据键，避免残留指向已卸载 DOM（线画到 0,0）。
+ */
+const dataKeyOwner: Record<string, string> = {}  // dataKey → 渲染键（谁注册的）
+const ownerDataKey: Record<string, string> = {}  // 渲染键 → dataKey
+function setNodeRef(key: string, el: any, element?: any) {
+  // 组件 ref 给的是实例（不是 DOM），ConnectionLines 直接调 getBoundingClientRect
+  // ⇒ 统一归一化为 DOM 元素（$el）；取不到 DOM 视同卸载，走清理分支。
+  const dom: HTMLElement | null = el
+    ? (el.nodeType === 1 ? el : (el.$el && el.$el.nodeType === 1 ? el.$el : null))
+    : null
+  if (!dom) {
+    const dk = ownerDataKey[key]
+    if (dk && dataKeyOwner[dk] === key) {
+      delete nodeRefs[dk]
+      delete dataKeyOwner[dk]
+    }
+    delete ownerDataKey[key]
+    delete nodeRefs[key]
+    return
+  }
+  nodeRefs[key] = dom
+  const nm = typeof element?.name === 'string' ? element.name : ''
+  if (nm) {
+    const dataKey = `${props.layerKey}:${nm}`
+    if (!nodeRefs[dataKey]) {
+      nodeRefs[dataKey] = dom
+      dataKeyOwner[dataKey] = key
+      ownerDataKey[key] = dataKey
+    }
+  }
 }
 
 // Expose nodeRefs to parent for connection line calculation
@@ -218,6 +287,54 @@ export default {
   letter-spacing: 0.3px;
 }
 
+/* C 方案：层头可点击聚焦 */
+.layer-header.is-focusable { cursor: pointer; }
+.layer-header.is-focusable:focus-visible {
+  outline: 2px solid #fff;
+  outline-offset: -3px;
+}
+.layer-header.is-focused { box-shadow: inset 0 -2px 0 rgba(255, 255, 255, 0.55); }
+
+.layer-focus-flag {
+  font-size: 10px;
+  font-weight: 500;
+  letter-spacing: .04em;
+  padding: 1px 6px;
+  border-radius: 4px;
+  background: rgba(255, 255, 255, 0.24);
+  flex-shrink: 0;
+}
+
+/* 压缩单行态：达成率色条 + 百分比（保信息不全隐） */
+.layer-mini {
+  display: inline-flex;
+  align-items: center;
+  gap: 6px;
+  flex-shrink: 0;
+  min-width: 96px;
+}
+.mini-track {
+  width: 56px;
+  height: 4px;
+  border-radius: 2px;
+  background: rgba(255, 255, 255, 0.28);
+  overflow: hidden;
+  display: inline-block;
+}
+.mini-fill {
+  display: block;
+  height: 100%;
+  border-radius: 2px;
+  background: #fff;
+  transition: width .2s ease;
+}
+.mini-ratio {
+  font-size: 11px;
+  font-weight: 600;
+  font-variant-numeric: tabular-nums;
+  opacity: .95;
+}
+
 .layer-icon {
   font-size: 18px;
   line-height: 1;

frontend/src/components/strategy/ConnectionLines.vue

@@ -136,12 +136,18 @@
 import { ref, computed, watch, onMounted, onUnmounted, nextTick } from 'vue'
 import { ElMessageBox } from 'element-plus'
 import { getChainForConnection, getChainForNode } from '../../utils/chainGraph'
+import { resolveEndpoints, relRectOf } from '../../utils/connKeys'
 
 const props = defineProps({
   /** 各层DOM元素引用，格式：{ financial: HTMLElement, customer: HTMLElement, ... } */
   layerRefs: { type: Object, required: true },
   /** 节点DOM元素引用，格式：{ 'financial-0': HTMLElement, ... } */
   nodeRefs: { type: Object, default: () => ({}) },
+  /**
+   * 服务端解析键（P3）：与 connections 逐下标对齐的 [{from_key,to_key}]。
+   * key = `${层}-${序号}`，由 objective_id 反查而来 ⇒ 目标改名后线仍指向同一节点。
+   */
+  connKeys: { type: Array, default: () => [] },
   /** 层顺序列表 */
   layerKeys: { type: Array, default: () => ['financial', 'customer', 'process', 'learning'] },
   /** 同层连线数据 [{from: 'financial-0', to: 'financial-1', effect, label}] */
@@ -328,26 +334,22 @@ function calcLayerConnections() {
 
   // 节点坐标缓存（相对 svg）：同一节点被多条连线引用时只 getBoundingClientRect 一次
   const rectCache = {}
+  const origin = { left: svgRect.left, top: svgRect.top }
   const getRect = (key) => {
-    if (rectCache[key]) return rectCache[key]
-    const el = props.nodeRefs[key]
-    if (!el) return null
-    const r = el.getBoundingClientRect()
-    rectCache[key] = {
-      left: r.left - svgRect.left,
-      top: r.top - svgRect.top,
-      right: r.right - svgRect.left,
-      bottom: r.bottom - svgRect.top,
-      width: r.width,
-      height: r.height,
-    }
+    if (!key) return null
+    if (key in rectCache) return rectCache[key]
+    // relRectOf：缺 ref / 已卸载 / 0×0（隐藏或未布局）一律 null —— 宁可不画，不画到 (0,0)
+    rectCache[key] = relRectOf(props.nodeRefs, key, origin) || null
     return rectCache[key]
   }
 
-  for (const conn of props.connections) {
-    const fr = getRect(conn.from)
-    const tr = getRect(conn.to)
-    if (!fr || !tr) continue
+  let skipped = 0
+  props.connections.forEach((conn, idx) => {
+    // P3：优先服务端解析键（objective_id 反查，改名不断线），缺失回落原标识串
+    const ep = resolveEndpoints(conn, idx, props.connKeys)
+    const fr = getRect(ep.from)
+    const tr = getRect(ep.to)
+    if (!fr || !tr) { skipped += 1; return }
 
     // 源节点右侧中点 → 目标节点左侧中点
     const x1 = fr.right
@@ -373,7 +375,11 @@ function calcLayerConnections() {
       effect: conn.effect || 'positive',
       label: conn.label || '',
     })
-  }
+  })
+
+  // 悬停/拖拽期节点可能尚未注册（预取未回、改名后 DOM 重建）⇒ 有跳过就做有界重算，
+  // 而不是留一条静默消失的线（P2 暴露的「零绘制」同类问题）。
+  if (skipped > 0 && depth < 6) scheduleRetry(depth + 1)
   layerConnections.value = result
 }
 

frontend/src/permission.ts

@@ -9,10 +9,10 @@ interface MenuItem {
 
 // ── 角色路由映射 ──
 export const ROLE_ROUTES: Record<string, string[]> = {
-  ceo: ['/my-dashboard', '/dashboard', '/ai-suggestions', '/kpis', '/maps', '/strategy-map', '/maps-review', '/maps/canvas', '/maps/review', '/alerts', '/notifications', '/org', '/users', '/permissions', '/budget', '/deviations', '/cost', '/cost-intelligence', '/predict', '/action-plans', '/knowledge', '/guide', '/customer', '/learning-dashboard', '/reports', '/alignment', '/dupont-analysis', '/analysis-confidence', '/expenses', '/cash-plan', '/receivables', '/growth-quality', '/product-matrix', '/tax-compliance', '/data-classification', '/cashflow-structure', '/treasury-dashboard', '/treasury-payment'],
-  finance: ['/my-dashboard', '/dashboard', '/ai-suggestions', '/kpis', '/maps', '/strategy-map', '/maps-review', '/maps/canvas', '/maps/review', '/alerts', '/data', '/budget', '/deviations', '/cost', '/cost-intelligence', '/predict', '/action-plans', '/knowledge', '/guide', '/customer', '/learning-dashboard', '/reports', '/alignment', '/dupont-analysis', '/analysis-confidence', '/expenses', '/cash-plan', '/receivables', '/growth-quality', '/product-matrix', '/tax-compliance', '/data-classification', '/cashflow-structure', '/treasury-dashboard', '/treasury-payment'],
-  business: ['/my-dashboard', '/dashboard', '/ai-suggestions', '/kpis', '/strategy-map', '/alerts', '/budget', '/deviations', '/action-plans', '/knowledge', '/guide', '/customer', '/expenses', '/cash-plan', '/growth-quality', '/product-matrix', '/tax-compliance', '/data-classification'],
-  it: ['/my-dashboard', '/dashboard', '/ai-suggestions', '/kpis', '/strategy-map', '/alerts', '/data', '/org', '/users', '/permissions', '/budget', '/deviations', '/cost', '/cost-intelligence', '/predict', '/action-plans', '/knowledge', '/guide', '/customer', '/learning-dashboard', '/analysis-confidence', '/expenses', '/cash-plan', '/product-matrix', '/tax-compliance', '/data-classification', '/treasury-payment'],
+  ceo: ['/my-dashboard', '/strategy-charter', '/dashboard', '/ai-suggestions', '/kpis', '/maps', '/strategy-map', '/maps-review', '/maps/canvas', '/maps/review', '/alerts', '/notifications', '/org', '/users', '/permissions', '/budget', '/deviations', '/cost', '/cost-intelligence', '/predict', '/action-plans', '/knowledge', '/guide', '/customer', '/learning-dashboard', '/reports', '/alignment', '/dupont-analysis', '/analysis-confidence', '/expenses', '/cash-plan', '/receivables', '/growth-quality', '/product-matrix', '/tax-compliance', '/data-classification', '/cashflow-structure', '/treasury-dashboard', '/treasury-payment'],
+  finance: ['/my-dashboard', '/strategy-charter', '/dashboard', '/ai-suggestions', '/kpis', '/maps', '/strategy-map', '/maps-review', '/maps/canvas', '/maps/review', '/alerts', '/data', '/budget', '/deviations', '/cost', '/cost-intelligence', '/predict', '/action-plans', '/knowledge', '/guide', '/customer', '/learning-dashboard', '/reports', '/alignment', '/dupont-analysis', '/analysis-confidence', '/expenses', '/cash-plan', '/receivables', '/growth-quality', '/product-matrix', '/tax-compliance', '/data-classification', '/cashflow-structure', '/treasury-dashboard', '/treasury-payment'],
+  business: ['/my-dashboard', '/strategy-charter', '/dashboard', '/ai-suggestions', '/kpis', '/strategy-map', '/alerts', '/budget', '/deviations', '/action-plans', '/knowledge', '/guide', '/customer', '/expenses', '/cash-plan', '/growth-quality', '/product-matrix', '/tax-compliance', '/data-classification'],
+  it: ['/my-dashboard', '/strategy-charter', '/dashboard', '/ai-suggestions', '/kpis', '/strategy-map', '/alerts', '/data', '/org', '/users', '/permissions', '/budget', '/deviations', '/cost', '/cost-intelligence', '/predict', '/action-plans', '/knowledge', '/guide', '/customer', '/learning-dashboard', '/analysis-confidence', '/expenses', '/cash-plan', '/product-matrix', '/tax-compliance', '/data-classification', '/treasury-payment'],
 }
 
 export const ROLE_ACTIONS: Record<string, string[]> = {
@@ -26,6 +26,7 @@ export const MENU_ITEMS: MenuItem[] = [
   // ── GROUP 1: 规划与目标（Plan）──
   { path: '/my-dashboard', label: '我的工作台', icon: 'Monitor', roles: ['ceo', 'finance', 'business', 'it'], group: '🔵 P 规划与目标' },
   { path: '/maps', label: '战略地图', icon: 'DataBoard', roles: ['ceo', 'finance'], group: '🔵 P 规划与目标' },
+  { path: '/strategy-charter', label: '战略章程', icon: 'Opportunity', roles: ['ceo', 'finance', 'business', 'it'], group: '🔵 P 规划与目标' },
   { path: '/strategy-map', label: '战略地图看板', icon: 'DataBoard', roles: ['ceo', 'finance', 'business', 'it'], group: '🔵 P 规划与目标' },
   { path: '/budget', label: '预算管理', icon: 'Coin', roles: ['ceo', 'finance', 'business', 'it'], group: '🔵 P 规划与目标' },
   { path: '/alignment', label: '战略执行看板', icon: 'Opportunity', roles: ['ceo', 'finance', 'business', 'it'], group: '🔵 P 规划与目标' },
@@ -39,6 +40,7 @@ export const MENU_ITEMS: MenuItem[] = [
   { path: '/receivables', label: '应收款催收', icon: 'Money', roles: ['ceo', 'finance', 'business'], group: '🟢 D 执行与控制' },
   { path: '/treasury-dashboard', label: '资金总览', icon: 'Money', roles: ['ceo','finance'], group: '🟢 D 执行与控制' },
   { path: '/treasury-payment', label: '资金内控', icon: 'Lock', roles: ['ceo','finance'], group: '🟢 D 执行与控制' },
+  { path: '/ar-aging', label: '应收账龄', icon: 'TrendCharts', roles: ['ceo', 'finance', 'business'], group: '🟡 C 监控与评价' },
   { path: '/growth-quality', label: '增长质量诊断', icon: 'TrendCharts', roles: ['ceo', 'finance', 'business', 'it'], group: '🟡 C 监控与评价' },
   { path: '/product-matrix', label: '波士顿产品矩阵', icon: 'PieChart', roles: ['ceo', 'finance', 'business', 'it'], group: '🟡 C 监控与评价' },
   { path: '/deviations', label: '差异分析', icon: 'DataAnalysis', roles: ['ceo', 'finance', 'business', 'it'], group: '🟢 D 执行与控制' },

frontend/src/router/index.ts

@@ -19,6 +19,8 @@ const routes = [
       { path: 'maps/canvas/:id', name: 'MapCanvas', component: () => import('@/views/MapCanvas.vue'), meta: { title: '战略地图画布', roles: ['ceo', 'finance'], zone: 'B', editable: true } },
       // /strategy-map: A区（只读活看板 —— 读库渲染，页内零写入口；2026-10-11 finance-20261011-strategy-map-frontend-fullstack）
       { path: 'strategy-map', name: 'StrategyMap', component: () => import('@/views/StrategyMap.vue'), meta: { title: '战略地图看板', roles: ['ceo', 'finance', 'business', 'it'], zone: 'A' } },
+      // /strategy-charter: P 组·战略基座入口（使命/愿景/价值观；页内含 MyDialog 写入口 ⇒ B区）
+      { path: 'strategy-charter', name: 'StrategyCharter', component: () => import('@/views/StrategyCharter.vue'), meta: { title: '战略章程', roles: ['ceo', 'finance', 'business', 'it'], zone: 'B', editable: true } },
       { path: 'maps/review/:id', name: 'MapReview', component: () => import('@/views/MapReview.vue'), meta: { title: '战略回顾会', roles: ['ceo', 'finance'], zone: 'B' } },
       { path: 'my-dashboard', name: 'MyDashboard', component: () => import('@/views/MyDashboard.vue'), meta: { title: '我的工作台', roles: ['ceo', 'finance', 'business', 'it'], zone: 'A' } },
       { path: 'knowledge', name: 'CMAKnowledge', component: () => import('@/views/CMAKnowledge.vue'), meta: { title: 'CMA知识库', roles: ['ceo', 'finance', 'business', 'it'], zone: 'B', editable: true } },
@@ -60,6 +62,7 @@ const routes = [
       { path: 'receivables', name: 'Receivables', component: () => import('@/views/Receivables.vue'), meta: { title: '应收款催收', roles: ['ceo', 'finance', 'business'], zone: 'B' } },
       { path: 'treasury-dashboard', name: 'TreasuryDashboard', component: () => import('@/views/TreasuryDashboard.vue'), meta: { title: '资金总览', roles: ['ceo','finance'], zone: 'B', editable: true } },
       { path: 'treasury-payment', name: 'TreasuryPayment', component: () => import('@/views/TreasuryPayment.vue'), meta: { title: '资金内控', roles: ['ceo','finance'], zone: 'B', editable: true } },
+      { path: 'ar-aging', name: 'ArAging', component: () => import('@/views/ArAging.vue'), meta: { title: '应收账龄', roles: ['ceo', 'finance', 'business'], zone: 'A' } },
       { path: 'growth-quality', name: 'GrowthQuality', component: () => import('@/views/GrowthQuality.vue'), meta: { title: '增长质量诊断', roles: ['ceo', 'finance', 'business', 'it'], zone: 'A' } },
       { path: 'product-matrix', name: 'ProductMatrix', component: () => import('@/views/ProductMatrix.vue'), meta: { title: '波士顿产品矩阵', roles: ['ceo', 'finance', 'business', 'it'], zone: 'A' } },
       // /tax-compliance: B区（页内职责=税务记录/发票/社保人工录入维护；见盘点表）

frontend/src/utils/connKeys.ts

@@ -0,0 +1,101 @@
+/**
+ * 连线端点解析（P3 cma-map-canvas-conn-key-rename-stable）
+ *
+ * 为什么需要它：
+ *   `canvas_data.connections[].from/to` 存的是**写入时刻**的节点标识串
+ *   （`层:中文名` 或 `层-序号`）。目标一改名，名字串指向不到任何节点 ⇒ 该线静默跳过。
+ *   服务端 `/api/cma/maps/{id}/connections` 现在为每条连接附 `from_key/to_key`
+ *   （= `${dim}-${idx}`，由 `from/to_objective_id` 在当前 dimensions 树反查得出），
+ *   渲染按 key 取节点 ⇒ 与中文名解耦；key 缺失（目标已删/历史脏 id）才回落到原值。
+ *
+ * 同时收口两类"悬停/拖拽期取不到节点"的兜底：
+ *   · 节点 ref 未注册（预取未返回/刚改名重建）⇒ 返回 null，交给调用方重试，绝不画到 (0,0)；
+ *   · 元素已卸载或隐藏（0×0 矩形）⇒ 视同缺失，同样不画。
+ */
+
+export type ConnectionKeyEntry = { from_key?: string | null; to_key?: string | null }
+
+export type ConnEndpoints = {
+  from: string | null
+  to: string | null
+  fromSource: 'key' | 'raw' | null
+  toSource: 'key' | 'raw' | null
+}
+
+/** 画布渲染键（`层-序号`）：只有这种键可以参与命中/吸附/合法目标判定 */
+export function isCanonicalNodeKey(key: unknown): boolean {
+  return typeof key === 'string' && /^[a-z_]+-\d+$/.test(key)
+}
+
+/**
+ * 单条连线两端取什么键去查 DOM。
+ * 优先服务端解析键（改名不断线），缺失回落数据里的原值（历史/降级路径）。
+ */
+export function resolveEndpoints(
+  conn: any,
+  idx: number,
+  resolved?: ConnectionKeyEntry[] | null,
+): ConnEndpoints {
+  const hit: ConnectionKeyEntry = (resolved && resolved[idx]) || {}
+  const pick = (side: 'from' | 'to') => {
+    const key = side === 'from' ? hit.from_key : hit.to_key
+    if (typeof key === 'string' && key) return { key, source: 'key' as const }
+    const raw = conn?.[side]
+    if (typeof raw === 'string' && raw) return { key: raw, source: 'raw' as const }
+    return { key: null, source: null }
+  }
+  const f = pick('from')
+  const t = pick('to')
+  return { from: f.key, to: t.key, fromSource: f.source, toSource: t.source }
+}
+
+export type RelRect = { left: number; top: number; right: number; bottom: number; width: number; height: number }
+
+/**
+ * 取节点矩形（相对容器原点）。
+ * 缺 ref / 已卸载 / 0×0（隐藏、未布局）一律 null —— 上层据此跳过该线并安排一次重算，
+ * 这条判据是「线不画到左上角 (0,0)」的唯一关卡。
+ */
+export function relRectOf(
+  nodeRefs: Record<string, any> | null | undefined,
+  key: string | null,
+  origin: { left: number; top: number },
+): RelRect | null {
+  if (!key || !nodeRefs) return null
+  const el: any = nodeRefs[key]
+  if (!el) return null
+  if (el.isConnected === false) return null
+  if (typeof el.getBoundingClientRect !== 'function') return null
+  const r = el.getBoundingClientRect()
+  if (!r || (!r.width && !r.height)) return null
+  return {
+    left: r.left - origin.left,
+    top: r.top - origin.top,
+    right: r.right - origin.left,
+    bottom: r.bottom - origin.top,
+    width: r.width,
+    height: r.height,
+  }
+}
+
+/**
+ * 解析键索引（服务端 /connections → 与本地 connections 数组逐下标对齐）。
+ * 按 `from→to` 配对消费（同名同向重复连线按出现顺序配对），配不上的留空 ⇒ 回落原值。
+ */
+export function buildKeyIndex(
+  localConnections: any[],
+  serverConnections: any[],
+): ConnectionKeyEntry[] {
+  const pool = new Map<string, any[]>()
+  for (const it of serverConnections || []) {
+    const pk = `${it?.from}\u0000${it?.to}`
+    const arr = pool.get(pk) || []
+    arr.push(it)
+    pool.set(pk, arr)
+  }
+  return (localConnections || []).map((c: any) => {
+    const arr = pool.get(`${c?.from}\u0000${c?.to}`) || []
+    const hit = arr.shift()
+    return { from_key: hit?.from_key ?? null, to_key: hit?.to_key ?? null }
+  })
+}

frontend/src/views/ArAging.vue

@@ -0,0 +1,478 @@
+<template>
+  <div class="ar-aging-page">
+    <!-- 页头：标题 + 口径后缀（强制显示）+ 数据截止期徽标 + 来源说明 -->
+    <header class="page-head">
+      <div class="head-main">
+        <h2 class="page-title">
+          应收账龄看板<span class="caliber-suffix">（{{ caliberLabel }}）</span>
+        </h2>
+        <div class="badge-row">
+          <el-tag size="small" effect="plain" :type="overview.has_data === false ? 'warning' : 'success'">
+            数据截止期 {{ displayPeriod || '无' }}
+          </el-tag>
+          <el-tag v-if="meta.loaded_at" size="small" effect="plain">入表 {{ meta.loaded_at }}</el-tag>
+          <el-tag size="small" effect="plain" type="primary">{{ entityLabel }}</el-tag>
+          <el-tag size="small" effect="plain" type="info">口径分列 · 禁跨口径合计</el-tag>
+        </div>
+        <p class="source-note">
+          来源：智享通「往来单位未收款账期统计」逐户导出；六档互斥（0-30/31-60/61-90/90-120/120-150/150+），
+          逐期六档和 = 期末应收合计；期初只展示，不作账龄基数。
+        </p>
+      </div>
+      <div class="head-controls">
+        <el-select v-model="caliber" size="small" style="width:190px" @change="onCaliberChange">
+          <el-option v-for="c in calibers" :key="c" :label="c" :value="c" />
+        </el-select>
+        <el-select
+          v-model="period" size="small" style="width:120px" :loading="loading"
+          placeholder="选择期间" @change="loadPeriodData"
+        >
+          <el-option v-for="p in periodOptions" :key="p" :label="p" :value="p" />
+        </el-select>
+        <el-button size="small" :loading="exporting" @click="doExport">
+          <svg class="ico" viewBox="0 0 24 24" aria-hidden="true">
+            <path d="M12 3v12m0 0l-4-4m4 4l4-4M4 21h16" fill="none" stroke="currentColor" stroke-width="1.8"
+                  stroke-linecap="round" stroke-linejoin="round" />
+          </svg>
+          导出逐户 CSV
+        </el-button>
+      </div>
+    </header>
+
+    <!-- 表不可用/无权限：降级可读，不留白、不补 0 -->
+    <el-alert
+      v-if="metaLoaded && meta.available === false"
+      type="warning" :closable="false" show-icon
+      title="该期无数据（表未就绪/无权限）"
+      :description="meta.reason || 'hanke_ar_aging 不可用'"
+    />
+
+    <template v-else>
+      <!-- 卡片区 -->
+      <el-row :gutter="16" v-loading="loading">
+        <el-col :xs="24" :sm="12" :md="6">
+          <el-card shadow="never" class="metric">
+            <div class="metric-label">应收总额（期末）</div>
+            <div class="metric-value">{{ money(overview.total_amt) }}</div>
+            <div class="metric-sub">
+              {{ overview.has_data ? `往来单位 ${overview.customers} 户` : '该期无数据' }}
+            </div>
+          </el-card>
+        </el-col>
+        <el-col :xs="24" :sm="12" :md="6">
+          <el-card shadow="never" class="metric">
+            <div class="metric-label">逾期金额（&gt;30 天）</div>
+            <div class="metric-value">{{ money(overview.overdue_amt) }}</div>
+            <div class="metric-sub">
+              环比
+              <span :class="momClass(overview.overdue_amt_mom_pct)">
+                {{ pctSigned(overview.overdue_amt_mom_pct) }}
+              </span>
+              <span class="dim">{{ overview.mom_note || (overview.prev_period ? `上期 ${overview.prev_period}` : '') }}</span>
+            </div>
+          </el-card>
+        </el-col>
+        <el-col :xs="24" :sm="12" :md="6">
+          <el-card shadow="never" class="metric">
+            <div class="metric-label">90 天以上占比</div>
+            <div class="metric-value" :style="{ color: shareColor }">{{ pct(overview.share_90plus) }}</div>
+            <div class="metric-sub">
+              <span v-if="thresholds.share_90plus_red_gt != null" class="th-red">
+                红线 &gt;{{ pct(thresholds.share_90plus_red_gt) }}
+              </span>
+              <span v-if="thresholds.share_90plus_yellow_min != null" class="th-yellow">
+                黄线 {{ pct(thresholds.share_90plus_yellow_min) }}–{{ pct(thresholds.share_90plus_yellow_max) }}
+              </span>
+              <span v-if="thresholds.share_90plus_red_gt == null" class="dim">阈值未配置（阶段B 播种后显示）</span>
+            </div>
+          </el-card>
+        </el-col>
+        <el-col :xs="24" :sm="12" :md="6">
+          <el-card shadow="never" class="metric">
+            <div class="metric-label">逾期户数（&gt;30 天）</div>
+            <div class="metric-value">{{ overview.overdue_customers ?? '该期无数据' }}</div>
+            <div class="metric-sub">
+              环比
+              <span :class="momClass(overview.overdue_customers_mom_pct)">
+                {{ pctSigned(overview.overdue_customers_mom_pct) }}
+              </span>
+            </div>
+          </el-card>
+        </el-col>
+      </el-row>
+
+      <!-- 图区 -->
+      <el-row :gutter="16" class="chart-row">
+        <el-col :xs="24" :lg="14">
+          <el-card shadow="never">
+            <template #header>
+              <div class="card-head">
+                <span>逐期趋势 · 应收总额 / 逾期户数</span>
+                <span class="dim small">{{ trendRange }}</span>
+              </div>
+            </template>
+            <div ref="trendRef" class="chart-box"></div>
+            <p class="chart-note">缺数据期以灰带标注「该期无数据」，曲线断点、不补 0。</p>
+          </el-card>
+        </el-col>
+        <el-col :xs="24" :lg="10">
+          <el-card shadow="never">
+            <template #header>
+              <div class="card-head">
+                <span>六档账龄堆叠</span>
+                <el-tag size="small" :type="bucketsBalanced ? 'success' : 'danger'" effect="plain">
+                  六档和=合计 {{ balancedCount }}/{{ dataPointCount }}
+                </el-tag>
+              </div>
+            </template>
+            <div ref="stackRef" class="chart-box"></div>
+            <p class="chart-note">tooltip 内断言六档和与合计一致；不一致会标红。</p>
+          </el-card>
+        </el-col>
+      </el-row>
+
+      <!-- 逐户明细 -->
+      <el-card shadow="never" class="table-card">
+        <template #header>
+          <div class="card-head">
+            <span>逐户逾期清单 · {{ displayPeriod || '—' }}（{{ caliberLabel }}）</span>
+            <div class="table-tools">
+              <el-switch v-model="overdueOnly" size="small" active-text="仅逾期" @change="loadCustomers" />
+              <span class="dim small">共 {{ customers.length }} 户</span>
+            </div>
+          </div>
+        </template>
+        <el-table
+          :data="customers" size="small" stripe border
+          :default-sort="{ prop: 'overdue_amt', order: 'descending' }"
+          @sort-change="onSortChange"
+          empty-text="该期无数据（不补 0）"
+        >
+          <el-table-column prop="unit_code" label="单位编码" width="170" sortable="custom" />
+          <el-table-column prop="unit_name" label="单位名称" min-width="130" sortable="custom" />
+          <el-table-column prop="total_amt" label="期末应收合计" width="130" align="right" sortable="custom">
+            <template #default="{ row }">{{ money(row.total_amt) }}</template>
+          </el-table-column>
+          <el-table-column prop="overdue_amt" label="逾期额(&gt;30天)" width="135" align="right" sortable="custom">
+            <template #default="{ row }">
+              <span :class="{ 'neg': row.overdue_amt > 0 }">{{ money(row.overdue_amt) }}</span>
+            </template>
+          </el-table-column>
+          <el-table-column prop="pct_of_unit" label="逾期占该户" width="110" align="right">
+            <template #default="{ row }">{{ pct(row.pct_of_unit) }}</template>
+          </el-table-column>
+          <el-table-column prop="oldest_days" label="最老逾期档" width="110" align="center" sortable="custom">
+            <template #default="{ row }">
+              <span v-if="row.oldest_days">{{ row.oldest_days }} 天以上</span>
+              <span v-else class="dim">未逾期</span>
+            </template>
+          </el-table-column>
+          <el-table-column v-for="b in buckets" :key="b.col" :prop="`buckets.${b.col}`"
+                           :label="b.label" width="108" align="right">
+            <template #default="{ row }">{{ money(row.buckets[b.col]) }}</template>
+          </el-table-column>
+        </el-table>
+      </el-card>
+    </template>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed, nextTick, onBeforeUnmount, onMounted, ref } from 'vue'
+import * as echarts from 'echarts'
+import { ElMessage } from 'element-plus'
+import api, { arAgingApi } from '@/api'
+import { sessionEntityName } from '@/utils/entity'
+
+const CALIBERS = ['借方应收户口径', '全户口径', '贷方户单列(净额)']
+const BUCKETS = [
+  { col: 'b_0_30', label: '0-30' },
+  { col: 'b_31_60', label: '31-60' },
+  { col: 'b_61_90', label: '61-90' },
+  { col: 'b_90_120', label: '90-120' },
+  { col: 'b_120_150', label: '120-150' },
+  { col: 'b_150p', label: '150+' },
+]
+
+const entityLabel = ref(sessionEntityName())
+const calibers = ref<string[]>(CALIBERS)
+const caliber = ref(CALIBERS[0])
+const period = ref('')
+const periodOptions = ref<string[]>([])
+const meta = ref<any>({ available: null })
+const metaLoaded = ref(false)
+const overview = ref<any>({})
+const thresholds = computed<any>(() => overview.value.thresholds || {})
+const trend = ref<any[]>([])
+const customers = ref<any[]>([])
+const overdueOnly = ref(false)
+const loading = ref(false)
+const exporting = ref(false)
+const sortState = ref<{ sort: string; order: string }>({ sort: 'overdue_amt', order: 'desc' })
+
+const caliberLabel = computed(() => caliber.value)
+const displayPeriod = computed(() => overview.value.period || period.value || '')
+const buckets = BUCKETS
+
+let trendChart: any = null
+let stackChart: any = null
+
+/** 金额格式化：null/undefined ⇒ 该期无数据（绝不显示 0.00） */
+function money(v: any): string {
+  if (v === null || v === undefined || v === '') return '该期无数据'
+  const n = Number(v)
+  if (!Number.isFinite(n)) return '该期无数据'
+  return `¥${n.toLocaleString('zh-CN', { minimumFractionDigits: 2, maximumFractionDigits: 2 })}`
+}
+
+function pct(v: any): string {
+  if (v === null || v === undefined || !Number.isFinite(Number(v))) return '—'
+  return `${(Number(v) * 100).toFixed(2)}%`
+}
+
+function pctSigned(v: any): string {
+  if (v === null || v === undefined || !Number.isFinite(Number(v))) return '—'
+  const n = Number(v) * 100
+  return `${n >= 0 ? '+' : ''}${n.toFixed(2)}%`
+}
+
+function momClass(v: any): string {
+  if (v === null || v === undefined) return 'dim'
+  return Number(v) > 0 ? 'up' : Number(v) < 0 ? 'down' : 'dim'
+}
+
+const shareColor = computed(() => {
+  const s = overview.value?.share_90plus
+  const th: any = thresholds.value
+  if (s === null || s === undefined) return 'inherit'
+  if (th.share_90plus_red_gt != null && Number(s) > Number(th.share_90plus_red_gt)) return '#f56c6c'
+  if (th.share_90plus_yellow_min != null && Number(s) >= Number(th.share_90plus_yellow_min)) return '#e6a23c'
+  return '#67c23a'
+})
+
+const dataPoints = computed(() => trend.value.filter(p => p.has_data))
+const dataPointCount = computed(() => dataPoints.value.length)
+const balancedCount = computed(() => dataPoints.value.filter(p => {
+  const sum = BUCKETS.reduce((acc, b) => acc + Number(p.buckets?.[b.col] ?? 0), 0)
+  return Math.abs(sum - Number(p.total_amt ?? 0)) < 0.01
+}).length)
+const bucketsBalanced = computed(() => dataPointCount.value > 0 && balancedCount.value === dataPointCount.value)
+const trendRange = computed(() => {
+  if (!trend.value.length) return ''
+  return `${trend.value[0].period} ~ ${trend.value[trend.value.length - 1].period}（${trend.value.length} 期）`
+})
+
+async function loadMeta() {
+  const r: any = await arAgingApi.meta()
+  meta.value = r || {}
+  metaLoaded.value = true
+  if (Array.isArray(r?.calibers) && r.calibers.length) calibers.value = r.calibers
+  const periods: string[] = Array.isArray(r?.periods_available) ? r.periods_available : []
+  // 期间候选 = 表内可用期 + 当前月（缺期也要能选中并显「该期无数据」）
+  const nowPeriod = new Date().toISOString().slice(0, 7)
+  const opts = [...periods]
+  if (!opts.includes(nowPeriod)) opts.push(nowPeriod)
+  periodOptions.value = opts.sort().reverse()
+  if (!period.value) period.value = r?.latest_period || opts[0] || ''
+}
+
+async function loadTrend() {
+  const r: any = await arAgingApi.trend({ caliber: caliber.value })
+  trend.value = Array.isArray(r?.points) ? r.points : []
+}
+
+async function loadPeriodData() {
+  loading.value = true
+  try {
+    const [ov, cu] = await Promise.all([
+      arAgingApi.overview({ period: period.value, caliber: caliber.value }),
+      arAgingApi.customers({
+        period: period.value, caliber: caliber.value,
+        overdue_only: overdueOnly.value, sort: sortState.value.sort, order: sortState.value.order,
+      }),
+    ])
+    overview.value = (ov as any) || {}
+    customers.value = Array.isArray((cu as any)?.items) ? (cu as any).items : []
+    if (!overview.value.period && period.value) overview.value.period = period.value
+    await nextTick()
+    renderCharts()
+  } catch (e: any) {
+    ElMessage.error(e?.message || '账龄数据加载失败')
+  } finally {
+    loading.value = false
+  }
+}
+
+async function loadCustomers() {
+  const r: any = await arAgingApi.customers({
+    period: period.value, caliber: caliber.value,
+    overdue_only: overdueOnly.value, sort: sortState.value.sort, order: sortState.value.order,
+  })
+  customers.value = Array.isArray(r?.items) ? r.items : []
+}
+
+function onSortChange({ prop, order }: any) {
+  if (!prop || !order) {
+    sortState.value = { sort: 'overdue_amt', order: 'desc' }
+  } else {
+    sortState.value = { sort: prop, order: order === 'ascending' ? 'asc' : 'desc' }
+  }
+  loadCustomers()
+}
+
+async function onCaliberChange() {
+  await Promise.all([loadTrend(), loadPeriodData()])
+}
+
+function renderCharts() {
+  const periods = trend.value.map(p => p.period)
+  const missing = trend.value.filter(p => !p.has_data).map(p => p.period)
+
+  if (trendRef.value) {
+    trendChart = trendChart || echarts.init(trendRef.value)
+    trendChart.setOption({
+      grid: { left: 60, right: 60, top: 40, bottom: 40 },
+      tooltip: {
+        trigger: 'axis',
+        formatter: (ps: any[]) => {
+          const p = ps[0]?.axisValue
+          const pt = trend.value.find(t => t.period === p)
+          if (!pt || !pt.has_data) return `${p}<br/>该期无数据`
+          return `${p}<br/>应收总额：${money(pt.total_amt)}<br/>逾期额(&gt;30天)：${money(pt.overdue_amt)}<br/>往来单位：${pt.customers} 户`
+        },
+      },
+      legend: { data: ['应收总额', '往来单位户数'], top: 0 },
+      xAxis: { type: 'category', data: periods, axisLabel: { fontSize: 11 } },
+      yAxis: [
+        { type: 'value', name: '金额(元)', axisLabel: { formatter: (v: number) => `${(v / 10000).toFixed(0)}万` } },
+        { type: 'value', name: '户数', minInterval: 1 },
+      ],
+      series: [
+        {
+          name: '应收总额', type: 'line', smooth: false, connectNulls: false,
+          data: trend.value.map(p => (p.has_data ? p.total_amt : null)),
+          itemStyle: { color: '#409EFF' }, areaStyle: { opacity: 0.08 },
+          markArea: missing.length ? {
+            silent: true,
+            itemStyle: { color: 'rgba(144,147,153,0.12)' },
+            label: { show: true, position: 'insideTop', color: '#909399', fontSize: 10, formatter: '该期无数据' },
+            data: missing.map(p => [{ xAxis: p }, { xAxis: p }]),
+          } : undefined,
+        },
+        {
+          name: '往来单位户数', type: 'bar', yAxisIndex: 1,
+          data: trend.value.map(p => (p.has_data ? p.customers : null)),
+          itemStyle: { color: '#c8d6e5' }, barMaxWidth: 18,
+        },
+      ],
+    }, true)
+  }
+
+  if (stackRef.value) {
+    stackChart = stackChart || echarts.init(stackRef.value)
+    stackChart.setOption({
+      grid: { left: 60, right: 24, top: 40, bottom: 40 },
+      tooltip: {
+        trigger: 'axis',
+        formatter: (ps: any[]) => {
+          const p = ps[0]?.axisValue
+          const pt = trend.value.find(t => t.period === p)
+          if (!pt || !pt.has_data) return `${p}<br/>该期无数据`
+          const sum = BUCKETS.reduce((acc, b) => acc + Number(pt.buckets?.[b.col] ?? 0), 0)
+          const diff = sum - Number(pt.total_amt ?? 0)
+          const ok = Math.abs(diff) < 0.01
+          const lines = BUCKETS.map((b, i) => `${b.label}：${money(ps[i]?.value)}`)
+          return [
+            `${p}（${caliber.value}）`,
+            ...lines,
+            `六档之和：${money(sum)}`,
+            `期末合计：${money(pt.total_amt)}`,
+            ok ? '校验：六档和 = 合计' : `校验异常：差额 ${diff.toFixed(2)}`,
+          ].join('<br/>')
+        },
+      },
+      legend: { data: BUCKETS.map(b => b.label), top: 0, itemWidth: 12, itemHeight: 8, textStyle: { fontSize: 11 } },
+      xAxis: { type: 'category', data: periods, axisLabel: { fontSize: 11 } },
+      yAxis: { type: 'value', axisLabel: { formatter: (v: number) => `${(v / 10000).toFixed(0)}万` } },
+      series: BUCKETS.map((b, i) => ({
+        name: b.label, type: 'bar', stack: 'aging', barMaxWidth: 26,
+        data: trend.value.map(p => (p.has_data ? p.buckets?.[b.col] : null)),
+        itemStyle: { color: ['#a3d0a3', '#e8c07d', '#e5a45f', '#e0765f', '#c74b4b', '#8f2f2f'][i] },
+      })),
+    }, true)
+  }
+}
+
+async function doExport() {
+  exporting.value = true
+  try {
+    const res: any = await api.get('/ar-aging/customers/export', {
+      params: {
+        period: period.value, caliber: caliber.value,
+        overdue_only: overdueOnly.value,
+      },
+      responseType: 'blob',
+    })
+    const url = window.URL.createObjectURL(res)
+    const a = document.createElement('a')
+    a.href = url
+    a.download = `应收账龄_${displayPeriod.value || 'noperiod'}_${caliber.value}.csv`
+    a.click()
+    window.URL.revokeObjectURL(url)
+    ElMessage.success('逐户清单已导出')
+  } catch (e: any) {
+    ElMessage.error('导出失败（需 finance/ceo/business 权限且该期有数据）')
+  } finally {
+    exporting.value = false
+  }
+}
+
+function onResize() {
+  trendChart?.resize()
+  stackChart?.resize()
+}
+
+const trendRef = ref<HTMLElement | null>(null)
+const stackRef = ref<HTMLElement | null>(null)
+
+onMounted(async () => {
+  window.addEventListener('resize', onResize)
+  await loadMeta()
+  await loadTrend()
+  await loadPeriodData()
+})
+
+onBeforeUnmount(() => {
+  window.removeEventListener('resize', onResize)
+  trendChart?.dispose()
+  stackChart?.dispose()
+})
+</script>
+
+<style scoped>
+.ar-aging-page { max-width: 1440px; margin: 0 auto; padding: 16px; }
+.page-head { display: flex; justify-content: space-between; align-items: flex-start; gap: 16px; margin-bottom: 16px; }
+.page-title { font-size: 20px; font-weight: 600; margin: 0 0 8px; }
+.caliber-suffix { color: #409eff; font-size: 16px; font-weight: 500; }
+.badge-row { display: flex; flex-wrap: wrap; gap: 6px; margin-bottom: 6px; }
+.source-note { font-size: 12px; color: #909399; margin: 0; line-height: 1.5; max-width: 720px; }
+.head-controls { display: flex; gap: 8px; align-items: center; flex-shrink: 0; }
+.metric { min-height: 108px; }
+.metric-label { font-size: 13px; color: #909399; }
+.metric-value { font-size: 26px; font-weight: 600; margin: 6px 0; word-break: break-all; }
+.metric-sub { font-size: 12px; color: #909399; }
+.chart-row { margin-top: 16px; }
+.chart-box { height: 320px; }
+.chart-note { font-size: 12px; color: #909399; margin: 6px 0 0; }
+.card-head { display: flex; justify-content: space-between; align-items: center; gap: 8px; }
+.table-card { margin-top: 16px; }
+.table-tools { display: flex; align-items: center; gap: 12px; }
+.small { font-size: 12px; }
+.dim { color: #909399; }
+.up { color: #f56c6c; font-weight: 600; }
+.down { color: #67c23a; font-weight: 600; }
+.neg { color: #f56c6c; }
+.th-red { color: #f56c6c; margin-right: 8px; }
+.th-yellow { color: #e6a23c; }
+.ico { width: 14px; height: 14px; margin-right: 4px; vertical-align: -2px; }
+</style>

frontend/src/views/MapCanvas.vue

@@ -34,6 +34,8 @@
       :achievement-count="achievementCount"
       :summary="reviewSummary"
       :loading="loading"
+      :objectives-total="totalObjectiveCount"
+      :connections-total="connections.length"
     />
 
     <MapEditBar
@@ -43,6 +45,7 @@
       :layer-configs="layerUiConfigs"
       :can-undo="undoStack.length > 0"
       :tree-open="treeOpen"
+      :focus-layer="focusLayer"
       @update:search="v => searchText = v"
       @update:filter="v => filterMode = v"
       @toggle-layer="toggleLayer"
@@ -52,6 +55,7 @@
       @fit="zoomFit"
       @toggle-tree="treeOpen = !treeOpen"
       @completeness="showCompleteness = true"
+      @focus-layer="toggleFocusLayer"
     />
 
     <!-- 口径标识（常驻，非 hover 临时提示）：A 口径 v1.1 来自会计科目、非发票同票列明 -->
@@ -80,7 +84,7 @@
       <el-button size="small" @click="cancelLink">取消</el-button>
     </div>
 
-    <!-- 仪表盘主视图：左目标树 / 中画布 / 右详情 -->
+    <!-- 仪表盘主视图：左目标树（默认收起，留 46px 图标条）/ 中画布 / 详情浮层 -->
     <div class="map-body-row">
       <MapTargetTree
         v-show="treeOpen"
@@ -98,6 +102,22 @@
         @drag-end="onTreeDragEnd"
       />
 
+      <!-- 目标树收起后的 46px 图标条（C 方案：默认收起，一键展开） -->
+      <div v-show="!treeOpen" class="tree-rail" aria-label="目标树已收起">
+        <button
+          type="button"
+          class="rail-btn"
+          title="展开目标树"
+          aria-label="展开目标树"
+          @click="treeOpen = true"
+        >
+          <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" aria-hidden="true">
+            <path d="M4 6h6M4 12h6M4 18h6" /><path d="M14 6h6M14 12h6M14 18h6" opacity=".45" />
+          </svg>
+        </button>
+        <span class="rail-count mono" :title="'当前可见目标 / 总目标'">{{ visibleObjectiveCount }}</span>
+      </div>
+
       <div class="map-canvas-col">
     <!-- 画布主体（田字格2×2布局，transform: scale 缩放） -->
     <div
@@ -108,7 +128,7 @@
       @drop.prevent="onTreeDrop"
     >
       <div class="canvas-zoom-wrap" :style="{ width: zoomWrapWidth, height: zoomWrapHeight }">
-      <div class="canvas-body" :class="{ 'grid-mode': gridMode }" ref="bodyRef" :style="bodyZoomStyle">
+      <div class="canvas-body" :class="{ 'grid-mode': gridMode, 'focus-active': !!focusLayer }" ref="bodyRef" :style="bodyZoomStyle">
         <!-- 四层泳道（纵向排列，固定顺序：财务→客户→流程→学习） -->
         <div
           v-for="(cfg, idx) in orderedLayerConfigs"
@@ -116,6 +136,10 @@
           v-show="visibleLayers[cfg.key]"
           :ref="el => setLayerRef(cfg.key, el)"
           class="layer-swimlane-wrapper"
+          :class="{
+            'is-focused': focusLayer === cfg.key,
+            'is-compact': !!focusLayer && focusLayer !== cfg.key,
+          }"
           :style="{ borderLeft: '4px solid ' + layerColor(cfg.key) }"
         >
           <StrategyLayer
@@ -126,6 +150,10 @@
             :color="layerColor(cfg.key)"
             :objectives="getLayerNodes(cfg.key)"
             :collapsed="!!collapsedLayers[cfg.key]"
+            :compact="!!focusLayer && focusLayer !== cfg.key"
+            :focused="focusLayer === cfg.key"
+            :layer-ratio="layerStats[cfg.key] ? layerStats[cfg.key].ratio : null"
+            @focus="toggleFocusLayer(cfg.key)"
             :dim-active="filterActive"
             :dimmed-keys="matchedKeyList"
             @toggle-collapse="toggleLayerCollapse(cfg.key)"
@@ -174,10 +202,40 @@
           <span class="lg-item"><i class="lg-dot lv-gray"></i>无数据</span>
           <span class="lg-note">达成率按指标方向归一，取目标关联 KPI 均值</span>
         </div>
-      </div><!-- end map-canvas-col -->
 
+        <!-- 跨层关系摘要条（C 方案）：数字由 connections 的 from/to 现算，不写死 -->
+        <div class="cross-summary" aria-label="跨层关系摘要">
+          <span class="cs-title">跨层关系摘要</span>
+          <span v-for="item in crossSummary" :key="item.key" class="cs-item">
+            <i class="cs-dot" :style="{ background: item.color }"></i>
+            <span class="cs-label">{{ item.label }}</span>
+            <b class="cs-num mono">{{ item.count }}</b>
+          </span>
+          <span class="cs-total">
+            合计 <b class="mono">{{ connections.length }}</b> 条 · 跨层 <b class="mono">{{ crossLayerCount }}</b> 条 · 同层 <b class="mono">{{ sameLayerCount }}</b> 条
+          </span>
+        </div>
+      </div><!-- end map-canvas-col -->
+    </div><!-- end map-body-row -->
+
+    <!-- 详情浮层（C 方案）：右栏不再常驻占位，画布按视口 100% 宽；浮层可拖动、可关闭 -->
+    <div
+      v-if="selectedNode"
+      class="detail-float"
+      :style="{ left: floatPos.x + 'px', top: floatPos.y + 'px' }"
+      role="dialog"
+      aria-label="目标详情浮层"
+    >
+      <div class="float-bar" @mousedown="startFloatDrag">
+        <span class="float-bar-grip" aria-hidden="true">
+          <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M8 8h.01M8 12h.01M8 16h.01M16 8h.01M16 12h.01M16 16h.01" /></svg>
+        </span>
+        <span class="float-bar-title">拖动可移动</span>
+        <button type="button" class="float-bar-close" aria-label="关闭详情浮层" @click="closeDetail">
+          <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" aria-hidden="true"><path d="M6 6l12 12M18 6L6 18" /></svg>
+        </button>
+      </div>
       <MapNodeDetail
-        v-if="selectedNode"
         :obj="selectedNode.obj"
         :dim-key="selectedNode.dimKey"
         :index="selectedNode.index"
@@ -196,13 +254,7 @@
         @select-key="selectByKey"
         @go-kpi-library="goKpiLibrary"
       />
-      <aside v-else class="map-detail-empty" aria-label="目标详情占位">
-        <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.5" aria-hidden="true">
-          <rect x="3" y="4" width="18" height="16" rx="2" /><path d="M7 9h10M7 13h7M7 17h4" />
-        </svg>
-        <p>选中左侧目标或画布节点<br />查看达成率、KPI 与因果上下游</p>
-      </aside>
-    </div><!-- end map-body-row -->
+    </div>
 
     <!-- SVG 连线层（覆盖整个画布，绘制跨层+同层箭头） -->
     <ConnectionLines
@@ -210,7 +262,8 @@
       :layer-refs="layerRefs"
       :node-refs="compiledNodeRefs"
       :layer-keys="layerKeysOrdered"
-      :connections="connections"
+      :conn-keys="connKeys"
+      :connections="renderedConnections"
       :temp-line="tempLine"
       :container-key="recalcTrigger"
       :view-direction="viewDirection"
@@ -319,6 +372,8 @@ import MapTargetTree from "../components/strategy-map/MapTargetTree.vue"
 import MapNodeDetail from "../components/strategy-map/MapNodeDetail.vue"
 import MapCompleteness from "../components/strategy-map/MapCompleteness.vue"
 import ConnectionLines from "../components/strategy/ConnectionLines.vue"
+// P3：连线端点解析键（`层-序号`）—— 目标改名后线仍指向同一节点
+import { buildKeyIndex, type ConnectionKeyEntry } from "../utils/connKeys"
 import MapCanvasToolbar from "../components/map-canvas/MapCanvasToolbar.vue"
 import MapDimensionFilter from "../components/map-canvas/MapDimensionFilter.vue"
 import CaliberBadge from "../components/CaliberBadge.vue"
@@ -387,7 +442,35 @@ const filterMode = ref<'all' | 'risk' | 'nodata'>('all')
 const visibleLayers = reactive<Record<string, boolean>>({ finance: true, customer: true, process: true, learning: true })
 const collapsedLayers = reactive<Record<string, boolean>>({})
 const collapsedGroups = reactive<Record<string, boolean>>({})
-const treeOpen = ref(true)
+
+/* ── C 方案：目标树默认收起 + 记住上次选择（localStorage 键名 cma.map.canvas.treeOpen）── */
+const TREE_OPEN_STORAGE_KEY = 'cma.map.canvas.treeOpen'
+function readTreeOpenPref(): boolean {
+  try {
+    return localStorage.getItem(TREE_OPEN_STORAGE_KEY) === '1'
+  } catch {
+    return false
+  }
+}
+const treeOpen = ref<boolean>(readTreeOpenPref())
+watch(treeOpen, (v) => {
+  try {
+    localStorage.setItem(TREE_OPEN_STORAGE_KEY, v ? '1' : '0')
+  } catch {
+    /* 隐私模式/配额满：静默降级为不记忆，不影响画布功能 */
+  }
+})
+
+/* ── C 方案：聚焦模式（null = 四层全景；非 null = 该层展开、其余三层压成单行）── */
+const focusLayer = ref<string | null>(null)
+function toggleFocusLayer(key: string | null) {
+  focusLayer.value = !key || focusLayer.value === key ? null : key
+  nextTick(() => {
+    updateGridMode()
+    triggerRecalc()
+    updateZoomWrapHeight()
+  })
+}
 
 const selectedNode = ref<{ dimKey: string; index: number; obj: any } | null>(null)
 const selectedKey = computed(() => (selectedNode.value ? `${selectedNode.value.dimKey}-${selectedNode.value.index}` : null))
@@ -403,6 +486,8 @@ const undoStack = ref<{ dimensions: any; connections: any[] }[]>([])
 
 // 连线
 const connections = ref<any[]>([])
+/** P3：服务端解析键（与 connections 逐下标对齐）；空 ⇒ 回落原标识串（历史/降级） */
+const connKeys = ref<ConnectionKeyEntry[]>([])
 const linkingFrom = ref<{ key: string; obj: any } | null>(null)
 const tempLine = ref<{ x1: number; y1: number; x2: number; y2: number } | null>(null)
 const dragConnectTarget = ref<string | null>(null)
@@ -520,6 +605,112 @@ const achievementCount = computed(() =>
   Object.values(kpiProgressMap).filter((p: any) => p.hasData && p.ratio != null).length
 )
 
+/** 各层达成率均值（聚焦压缩态色条用）：仅统计有实测值的目标，无数据 ⇒ null（界面显式「—」） */
+const layerStats = computed<Record<string, { ratio: number | null; count: number }>>(() => {
+  const out: Record<string, { ratio: number | null; count: number }> = {}
+  for (const cfg of orderedLayerConfigs) {
+    const objs = getLayerNodes(cfg.key)
+    const vals: number[] = []
+    objs.forEach((_: any, i: number) => {
+      const p: any = kpiProgressMap[`${cfg.key}-${i}`]
+      if (p && p.hasData && p.ratio != null && !Number.isNaN(Number(p.ratio))) vals.push(Number(p.ratio))
+    })
+    out[cfg.key] = { ratio: vals.length ? vals.reduce((a, b) => a + b, 0) / vals.length : null, count: objs.length }
+  }
+  return out
+})
+
+/* ── C 方案：跨层关系摘要（全部由 connections 的 from/to 现算，禁写死）──
+ * 连线端点兼容两种键：渲染键 `finance-0` 与数据键 `learning:目标名`（StrategyLayer 双键注册）。
+ */
+function layerOfNodeKey(key: string | null | undefined): string | null {
+  if (!key || typeof key !== 'string') return null
+  const sep = key.includes(':') ? ':' : '-'
+  const layer = key.split(sep)[0]
+  return LAYER_KEYS.includes(layer as any) ? layer : null
+}
+const layerHue = (key: string) => LAYER_VISUAL[key]?.color || LAYER_CONFIG[key]?.color || '#635BFF'
+/** 聚焦态只绘制聚焦层内部的同层连线（跨层线跨向被压缩层时无锚点）；全景态全量绘制 */
+const renderedConnections = computed<any[]>(() => {
+  const f = focusLayer.value
+  if (!f) return connections.value
+  return connections.value.filter(
+    (c: any) => layerOfNodeKey(c?.from) === f && layerOfNodeKey(c?.to) === f
+  )
+})
+const crossLayerCount = computed(() =>
+  connections.value.filter((c: any) => {
+    const f = layerOfNodeKey(c?.from)
+    const t = layerOfNodeKey(c?.to)
+    return !!f && !!t && f !== t
+  }).length
+)
+const sameLayerCount = computed(() => connections.value.length - crossLayerCount.value)
+/** BSC 三段因果 + 「其他跨层」兜底桶 + 同层；合计必须等于 connections.length */
+const crossSummary = computed(() => {
+  const dirs: { key: string; label: string; from: string; to: string }[] = [
+    { key: 'learn2proc', label: '学习→流程', from: 'learning', to: 'process' },
+    { key: 'proc2cust', label: '流程→客户', from: 'process', to: 'customer' },
+    { key: 'cust2fin', label: '客户→财务', from: 'customer', to: 'finance' },
+  ]
+  const counts: Record<string, number> = {}
+  let other = 0
+  for (const c of connections.value as any[]) {
+    const f = layerOfNodeKey(c?.from)
+    const t = layerOfNodeKey(c?.to)
+    if (!f || !t || f === t) continue
+    const hit = dirs.find(d => d.from === f && d.to === t)
+    if (hit) {
+      counts[hit.key] = (counts[hit.key] || 0) + 1
+    } else {
+      other++
+    }
+  }
+  const items = dirs.map(d => ({
+    key: d.key,
+    label: d.label,
+    count: counts[d.key] || 0,
+    color: layerHue(d.to),
+  }))
+  if (other > 0) items.push({ key: 'other', label: '其他跨层', count: other, color: '#94A3B8' })
+  return items
+})
+
+/* ── C 方案：详情浮层位置与拖动（自绘 div + transform，非 el-dialog）── */
+const floatPos = ref<{ x: number; y: number }>({ x: 0, y: 0 })
+const floatDragging = ref(false)
+let floatDragOffset = { x: 0, y: 0 }
+const FLOAT_W = 360
+function clampFloat(x: number, y: number) {
+  const maxX = Math.max(8, window.innerWidth - FLOAT_W - 12)
+  const maxY = Math.max(8, window.innerHeight - 120)
+  return { x: Math.min(Math.max(8, x), maxX), y: Math.min(Math.max(8, y), maxY) }
+}
+/** 默认落位：右侧靠上（不遮挡左侧层名与中部卡片） */
+function placeFloatDefault() {
+  floatPos.value = clampFloat(window.innerWidth - FLOAT_W - 24, 132)
+}
+function startFloatDrag(e: MouseEvent) {
+  floatDragging.value = true
+  floatDragOffset = { x: e.clientX - floatPos.value.x, y: e.clientY - floatPos.value.y }
+  window.addEventListener('mousemove', onFloatDragMove)
+  window.addEventListener('mouseup', stopFloatDrag)
+  e.preventDefault()
+}
+function onFloatDragMove(e: MouseEvent) {
+  if (!floatDragging.value) return
+  floatPos.value = clampFloat(e.clientX - floatDragOffset.x, e.clientY - floatDragOffset.y)
+}
+function stopFloatDrag() {
+  floatDragging.value = false
+  window.removeEventListener('mousemove', onFloatDragMove)
+  window.removeEventListener('mouseup', stopFloatDrag)
+}
+watch(selectedNode, (v) => {
+  if (v && !floatPos.value.x && !floatPos.value.y) placeFloatDefault()
+})
+onUnmounted(() => stopFloatDrag())
+
 /** 搜索/筛选是否激活（激活时画布非匹配节点变淡） */
 const filterActive = computed(() => searchText.value.trim() !== '' || filterMode.value !== 'all')
 
@@ -797,6 +988,22 @@ watch(zoomLevel, () => {
   nextTick(() => connectionLinesRef.value?.recalcAll())
 })
 
+// ── P3：服务端解析键 ──
+// 为什么必须取服务端：canvas_data 里存的是写入时刻的标识串（`层:中文名`），
+// 目标改名后名字串失效；服务端能由 from/to_objective_id 反查当前 `${层}-${序号}`。
+// 取不到（接口异常/离线）⇒ 置空，画布回落原标识串，行为与改造前一致（不更坏）。
+async function refreshConnKeys() {
+  const id = selectedMap.value
+  if (!id) { connKeys.value = []; return }
+  try {
+    const r: any = await api.get(`/maps/${id}/connections`)
+    const serverConns = r?.data?.connections || []
+    connKeys.value = buildKeyIndex(loadedCanvas || connections.value || [], serverConns)
+  } catch (e) {
+    connKeys.value = []
+  }
+}
+
 // ── 画布加载 ──
 function loadCanvas() {
   caliberFacts.value = caliberGapFor(currentCaliberEntityId())
@@ -842,6 +1049,7 @@ function loadCanvas() {
     loadedDims = deepClone(dimensions)
     connections.value = map?.canvas_data?.connections || []
     loading.value = false
+    refreshConnKeys()
     nextTick(() => {
       triggerRecalc()
       updateGridMode()
@@ -982,6 +1190,7 @@ function undoLast() {
   connections.value = JSON.parse(JSON.stringify(snap.connections))
   selectedNode.value = null
   markDirty()
+  refreshConnKeys()   // P3：撤销可能改动目标顺序/名字，解析键随之刷新
   nextTick(() => { triggerRecalc(); updateZoomWrapHeight(); loadObjectiveLevels() })
   ElMessage.success('已撤销上一步')
 }
@@ -994,10 +1203,21 @@ function onTreeDragStart(payload: { dimKey: string; index: number; obj: any }) {
 }
 function onTreeDragOver(e: DragEvent) {
   if (!treeDragSource.value) return
-  const el = document.elementFromPoint(e.clientX, e.clientY)?.closest('[data-node-key]') as HTMLElement | null
-  const key = el?.getAttribute('data-node-key') || null
+  // 命中判定改用节点 rect 包含判断：elementFromPoint 会被覆盖层（窄屏时的知识面板 .kp-panel、
+  // 连线 SVG 等）挡住 ⇒ 拖拽在窄屏下永远取不到目标节点。rect 判定与层级无关。
+  const key = hitTestNodeKey(e.clientX, e.clientY)
   dragConnectTarget.value = key && key !== treeDragSource.value.key ? key : null
 }
+function hitTestNodeKey(x: number, y: number): string | null {
+  const nodes = document.querySelectorAll('[data-node-key]') as NodeListOf<HTMLElement>
+  for (const el of Array.from(nodes)) {
+    const r = el.getBoundingClientRect()
+    if (r.width > 0 && r.height > 0 && x >= r.left && x <= r.right && y >= r.top && y <= r.bottom) {
+      return el.getAttribute('data-node-key')
+    }
+  }
+  return null
+}
 function onTreeDragLeave(e: DragEvent) {
   const related = (e as any).relatedTarget as Node | null
   const box = scrollWrapRef.value
@@ -1705,6 +1925,9 @@ async function saveCanvas(silent?: boolean) {
     if (res?.updated_at) baseUpdatedAt.value = res.updated_at   // 采纳服务端新基准
     refreshBaselines(res?.dimensions)                           // 基线回到服务端现值（响应体为权威）
     markClean()
+    // P3：保存后刷新解析键（改名/调序都会让 `${层}-${序号}` 变，线要跟着节点走）
+    if (payload.canvas_data) loadedCanvas = deepClone(connections.value)
+    refreshConnKeys()
     if (!silent) ElMessage.success("已保存")
   } catch (e: any) {
     if (e?.response?.status === 409) {
@@ -1731,6 +1954,8 @@ async function saveConnections() {
       base_updated_at: currentBaseStamp(),
     })
     if (res?.updated_at) baseUpdatedAt.value = res.updated_at
+    loadedCanvas = deepClone(connections.value)
+    refreshConnKeys()   // P3：连线新增/删除后服务端重算解析键
   } catch (e: any) {
     if (e?.response?.status === 409) {
       autoSaveBlocked.value = true
@@ -2137,6 +2362,140 @@ onUnmounted(() => {
   margin-bottom: 4px;
 }
 
+/* ── C 方案：聚焦模式 ── */
+/* 聚焦层：田字格下占满整行，卡片网格展开 */
+.canvas-body.grid-mode .layer-swimlane-wrapper.is-focused {
+  grid-column: 1 / -1;
+}
+/* 其余三层：压成单行（≤48px，仅层头 + 目标数 + 达成率色条） */
+.canvas-body.focus-active .layer-swimlane-wrapper.is-compact {
+  height: 48px;
+  min-height: 48px;
+  max-height: 48px;
+  overflow: hidden;
+  align-self: start;
+}
+.canvas-body.focus-active .layer-swimlane-wrapper.is-compact :deep(.strategy-layer) {
+  min-height: 0;
+  height: 48px;
+}
+.canvas-body.focus-active .layer-swimlane-wrapper.is-compact :deep(.layer-body) {
+  display: none;
+}
+.canvas-body.focus-active .layer-swimlane-wrapper.is-compact :deep(.layer-header) {
+  padding: 6px 14px;
+  gap: 8px;
+}
+.canvas-body.focus-active .layer-swimlane-wrapper.is-focused :deep(.strategy-layer) {
+  box-shadow: 0 6px 18px rgba(15, 23, 42, 0.10);
+}
+
+/* ── 目标树收起后的 46px 图标条 ── */
+.tree-rail {
+  width: 46px;
+  flex: 0 0 46px;
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  gap: 8px;
+  padding: 8px 0;
+  border: 1px solid var(--bh-border, #E5E7EB);
+  border-radius: 8px;
+  background: var(--bh-bg-card, #fff);
+}
+.rail-btn {
+  width: 30px;
+  height: 30px;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  border: 1px solid var(--bh-border, #E5E7EB);
+  border-radius: 6px;
+  background: #F8FAFC;
+  color: #475569;
+  cursor: pointer;
+}
+.rail-btn:hover { background: #EEF2F7; color: #0F172A; }
+.rail-btn svg { width: 16px; height: 16px; }
+.rail-count {
+  font-size: 11px;
+  color: var(--bh-text-secondary, #64748B);
+  font-variant-numeric: tabular-nums;
+}
+
+/* ── 详情浮层（自绘 div，可拖动，非 el-dialog） ── */
+.detail-float {
+  position: absolute;
+  z-index: 60;
+  width: 360px;
+  max-height: calc(100vh - 180px);
+  display: flex;
+  flex-direction: column;
+  background: var(--bh-bg-card, #fff);
+  border: 1px solid #E5E7EB;
+  border-radius: 10px;
+  box-shadow: 0 12px 32px rgba(15, 23, 42, 0.16), 0 2px 6px rgba(15, 23, 42, 0.08);
+  overflow: hidden;
+}
+.detail-float :deep(.node-detail) {
+  width: 100%;
+  flex: 1 1 auto;
+  border: 0;
+  border-radius: 0;
+  overflow: auto;
+}
+.float-bar {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+  height: 28px;
+  padding: 0 8px;
+  background: #F8FAFC;
+  border-bottom: 1px solid #E5E7EB;
+  cursor: move;
+  user-select: none;
+  flex-shrink: 0;
+}
+.float-bar-grip { display: inline-flex; color: #94A3B8; }
+.float-bar-grip svg { width: 14px; height: 14px; }
+.float-bar-title { font-size: 11px; color: #94A3B8; flex: 1; }
+.float-bar-close {
+  width: 20px;
+  height: 20px;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  border: 0;
+  background: transparent;
+  color: #64748B;
+  border-radius: 4px;
+  cursor: pointer;
+}
+.float-bar-close:hover { background: #E2E8F0; color: #0F172A; }
+.float-bar-close svg { width: 12px; height: 12px; }
+
+/* ── 跨层关系摘要条 ── */
+.cross-summary {
+  flex-shrink: 0;
+  display: flex;
+  align-items: center;
+  gap: 14px;
+  flex-wrap: nowrap;
+  overflow-x: auto;
+  padding: 6px 2px 0;
+  font-size: 11px;
+  color: #475569;
+  border-top: 1px dashed var(--bh-border, #E5E7EB);
+  margin-top: 4px;
+}
+.cs-title { color: #94A3B8; flex-shrink: 0; }
+.cs-item { display: inline-flex; align-items: center; gap: 4px; flex-shrink: 0; }
+.cs-dot { width: 7px; height: 7px; border-radius: 2px; display: inline-block; }
+.cs-label { color: #64748B; }
+.cs-num { color: #0F172A; font-variant-numeric: tabular-nums; }
+.cs-total { margin-left: auto; color: #64748B; flex-shrink: 0; }
+.cs-total b { color: #0F172A; }
+
 /* ── 跨层箭头指示器 ── */
 .cross-layer-arrow {
   display: flex;

frontend/src/views/StrategyCharter.vue

@@ -0,0 +1,288 @@
+<template>
+  <div class="charter-page">
+    <div class="charter-head">
+      <div class="charter-title">
+        <h2>战略章程</h2>
+        <p>P 战略基座 · 使命 / 愿景（年度 + 可验证状态）/ 价值观 —— 全公司战略口径的唯一起点</p>
+      </div>
+      <div class="charter-actions">
+        <el-button
+          v-if="canEdit"
+          type="primary"
+          :icon="Edit"
+          :disabled="loading"
+          @click="openEdit"
+        >{{ filled ? '修订章程' : '制定章程' }}</el-button>
+        <el-button v-else :icon="Lock" disabled title="仅经营层（总经理 / 财务部）可修订章程">只读</el-button>
+      </div>
+    </div>
+
+    <el-skeleton v-if="loading" :rows="6" animated />
+
+    <!-- 空态：库里没有章程行 —— 不显示任何示例文本（派单红线） -->
+    <div v-else-if="!filled" class="charter-empty">
+      <el-icon :size="40"><Opportunity /></el-icon>
+      <h3>尚未填写战略章程</h3>
+      <p>本账套还没有使命 / 愿景 / 价值观记录。请由经营层录入第一条，系统不做任何预填。</p>
+      <el-button v-if="canEdit" type="primary" :icon="Edit" @click="openEdit">录入章程</el-button>
+    </div>
+
+    <template v-else>
+      <div class="charter-cards">
+        <section class="charter-card">
+          <div class="card-label"><el-icon><Aim /></el-icon>使命<span>Mission</span></div>
+          <p class="card-text">{{ data?.mission }}</p>
+        </section>
+        <section class="charter-card">
+          <div class="card-label">
+            <el-icon><Sunrise /></el-icon>愿景<span>Vision</span>
+            <el-tag size="small" effect="plain" class="year-tag">{{ data?.vision_year }}</el-tag>
+          </div>
+          <p class="card-text">{{ data?.vision }}</p>
+          <div class="measurable">
+            <span class="measurable-k">可验证状态</span>
+            <span class="measurable-v">{{ data?.vision_measurable }}</span>
+          </div>
+        </section>
+        <section class="charter-card charter-card--wide">
+          <div class="card-label"><el-icon><Star /></el-icon>价值观<span>Core Values</span></div>
+          <p class="card-text">{{ data?.values_text }}</p>
+          <ul class="value-list">
+            <li v-for="(item, idx) in data?.value_items || []" :key="idx">
+              <span class="value-name">{{ item.name }}</span>
+              <span class="value-meaning">{{ item.meaning }}</span>
+            </li>
+          </ul>
+        </section>
+      </div>
+
+      <div class="charter-meta">
+        <div class="meta-item">
+          <span class="meta-k">最后修订</span>
+          <span class="meta-v">{{ data?.updated_by_name || '—' }} · {{ data?.updated_at || '—' }}</span>
+        </div>
+        <div class="meta-item">
+          <span class="meta-k">创建时间</span>
+          <span class="meta-v">{{ data?.created_at || '—' }}</span>
+        </div>
+      </div>
+    </template>
+
+    <!-- 项目铁律：业务弹窗一律 MyDialog（禁 el-dialog / ElMessageBox 确认） -->
+    <MyDialog v-model="dialogVisible" :title="filled ? '修订战略章程' : '制定战略章程'" :width="760">
+      <el-form :model="form" label-width="110px" label-position="left" class="charter-form">
+        <el-form-item label="使命" required>
+          <el-input
+            v-model="form.mission"
+            type="textarea"
+            :rows="2"
+            maxlength="1000"
+            show-word-limit
+            placeholder="我们为何存在：做什么、为谁做"
+          />
+        </el-form-item>
+        <el-form-item label="愿景" required>
+          <el-input
+            v-model="form.vision"
+            type="textarea"
+            :rows="2"
+            maxlength="1000"
+            show-word-limit
+            placeholder="要成为什么"
+          />
+        </el-form-item>
+        <el-form-item label="愿景年度" required>
+          <el-input v-model="form.vision_year" maxlength="4" placeholder="如 2030" class="year-input" />
+          <span class="form-hint">不可空；4 位年份</span>
+        </el-form-item>
+        <el-form-item label="可验证状态" required>
+          <el-input
+            v-model="form.vision_measurable"
+            type="textarea"
+            :rows="2"
+            maxlength="1000"
+            show-word-limit
+            placeholder="用可核对的规模/业务口径描述，如「2030 年服务 300 家付费企业、续约率 ≥85%」"
+          />
+          <span class="form-hint">不可空：没有可核对口径的愿景无法验证</span>
+        </el-form-item>
+        <el-form-item label="价值观总述" required>
+          <el-input
+            v-model="form.values_text"
+            type="textarea"
+            :rows="2"
+            maxlength="2000"
+            show-word-limit
+            placeholder="不可交易的底线，一句话总述"
+          />
+        </el-form-item>
+        <el-form-item label="价值观条目" required>
+          <div class="value-editor">
+            <div v-for="(item, idx) in form.value_items" :key="idx" class="value-row">
+              <el-input v-model="item.name" maxlength="50" placeholder="名称（如 客户成功）" class="value-row-name" />
+              <el-input v-model="item.meaning" maxlength="300" placeholder="含义：决策冲突时如何裁决" class="value-row-meaning" />
+              <el-button
+                :icon="Delete"
+                text
+                type="danger"
+                :disabled="form.value_items.length <= 1"
+                title="删除该条"
+                @click="form.value_items.splice(idx, 1)"
+              />
+            </div>
+            <el-button :icon="Plus" text type="primary" @click="addValueItem">新增条目</el-button>
+          </div>
+        </el-form-item>
+      </el-form>
+      <template #footer>
+        <el-button @click="dialogVisible = false" :disabled="saving">取消</el-button>
+        <el-button type="primary" :loading="saving" @click="save">保存</el-button>
+      </template>
+    </MyDialog>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed, onMounted, reactive, ref } from 'vue'
+import { ElMessage } from 'element-plus'
+import { Aim, Delete, Edit, Lock, Opportunity, Plus, Star, Sunrise } from '@element-plus/icons-vue'
+import MyDialog from '@/components/MyDialog.vue'
+import { charterApi } from '@/api'
+
+interface ValueItem { name: string; meaning: string }
+
+const loading = ref(false)
+const saving = ref(false)
+const filled = ref(false)
+const data = ref<any>(null)
+const dialogVisible = ref(false)
+
+const userRole = (() => {
+  try { return JSON.parse(localStorage.getItem('cma_user') || '{}').role || '' } catch { return '' }
+})()
+// 与后端 PUT 角色白名单保持一致（ceo / finance）
+const canEdit = computed(() => ['ceo', 'finance'].includes(userRole))
+
+const form = reactive({
+  mission: '',
+  vision: '',
+  vision_year: '',
+  vision_measurable: '',
+  values_text: '',
+  value_items: [{ name: '', meaning: '' }] as ValueItem[],
+})
+
+async function load() {
+  loading.value = true
+  try {
+    const res: any = await charterApi.get()
+    filled.value = !!res.filled
+    data.value = res.data
+  } catch (e) {
+    ElMessage.error('章程加载失败')
+  } finally {
+    loading.value = false
+  }
+}
+
+function openEdit() {
+  form.mission = data.value?.mission || ''
+  form.vision = data.value?.vision || ''
+  form.vision_year = data.value?.vision_year || ''
+  form.vision_measurable = data.value?.vision_measurable || ''
+  form.values_text = data.value?.values_text || ''
+  const items: ValueItem[] = (data.value?.value_items || []).map((i: ValueItem) => ({
+    name: i.name, meaning: i.meaning,
+  }))
+  form.value_items = items.length ? items : [{ name: '', meaning: '' }]
+  dialogVisible.value = true
+}
+
+function addValueItem() {
+  form.value_items.push({ name: '', meaning: '' })
+}
+
+function validate(): string | null {
+  if (!form.mission.trim()) return '使命不能为空'
+  if (!form.vision.trim()) return '愿景不能为空'
+  if (!/^\d{4}$/.test(form.vision_year.trim())) return '愿景年度须为 4 位年份（如 2030）'
+  if (!form.vision_measurable.trim()) return '可验证状态不能为空'
+  if (!form.values_text.trim()) return '价值观总述不能为空'
+  const items = form.value_items.filter(i => i.name.trim() || i.meaning.trim())
+  if (!items.length) return '价值观条目至少 1 条'
+  if (items.some(i => !i.name.trim() || !i.meaning.trim())) return '价值观条目的名称与含义都要填'
+  return null
+}
+
+async function save() {
+  const err = validate()
+  if (err) { ElMessage.warning(err); return }
+  saving.value = true
+  try {
+    // 只提交后端字段白名单内的字段（entity_id 由登录账套决定，不手写）
+    const payload = {
+      mission: form.mission.trim(),
+      vision: form.vision.trim(),
+      vision_year: form.vision_year.trim(),
+      vision_measurable: form.vision_measurable.trim(),
+      values_text: form.values_text.trim(),
+      value_items: form.value_items
+        .filter(i => i.name.trim() && i.meaning.trim())
+        .map(i => ({ name: i.name.trim(), meaning: i.meaning.trim() })),
+    }
+    const body: any = await charterApi.save(payload)
+    ElMessage.success('战略章程已保存')
+    dialogVisible.value = false
+    filled.value = true
+    data.value = body.data
+  } catch (e: any) {
+    ElMessage.error(e?.response?.data?.detail || '保存失败，请稍后重试')
+  } finally {
+    saving.value = false
+  }
+}
+
+onMounted(load)
+</script>
+
+<style scoped>
+.charter-page { padding: var(--bh-space-xl, 20px); background: var(--bh-bg-main); min-height: 100%; }
+.charter-head { display: flex; justify-content: space-between; align-items: flex-start; gap: var(--bh-space-lg, 16px); margin-bottom: var(--bh-space-xl, 20px); flex-wrap: wrap; }
+.charter-title h2 { margin: 0 0 6px; font-size: 20px; color: var(--bh-text-primary); font-weight: 600; }
+.charter-title p { margin: 0; font-size: 13px; color: var(--bh-text-secondary); }
+.charter-actions { display: flex; align-items: center; gap: var(--bh-space-md, 12px); }
+
+.charter-empty { background: var(--bh-bg-card); border: 1px dashed var(--bh-border); border-radius: var(--bh-radius-md, 8px); padding: 56px 24px; text-align: center; color: var(--bh-text-secondary); }
+.charter-empty h3 { margin: 12px 0 6px; color: var(--bh-text-primary); font-size: 16px; }
+.charter-empty p { margin: 0 0 18px; font-size: 13px; }
+
+.charter-cards { display: grid; grid-template-columns: 1fr 1fr; gap: var(--bh-space-lg, 16px); }
+.charter-card { background: var(--bh-bg-card); border-radius: var(--bh-radius-md, 8px); box-shadow: var(--bh-shadow-sm); padding: var(--bh-space-xl, 20px); border-left: 3px solid var(--bh-primary); }
+.charter-card--wide { grid-column: 1 / -1; border-left-color: var(--bh-success); }
+.card-label { display: flex; align-items: center; gap: 6px; font-size: 13px; font-weight: 600; color: var(--bh-text-regular); }
+.card-label span { font-weight: 400; font-size: 12px; color: var(--bh-text-placeholder); margin-left: 4px; letter-spacing: .3px; }
+.year-tag { margin-left: auto; }
+.card-text { margin: 12px 0 0; font-size: 15px; line-height: 1.75; color: var(--bh-text-primary); white-space: pre-wrap; word-break: break-word; }
+
+.measurable { margin-top: 14px; padding-top: 12px; border-top: 1px dashed var(--bh-border-light); display: flex; flex-direction: column; gap: 4px; }
+.measurable-k { font-size: 12px; color: var(--bh-text-secondary); }
+.measurable-v { font-size: 13px; color: var(--bh-text-regular); line-height: 1.6; word-break: break-word; }
+
+.value-list { list-style: none; margin: 14px 0 0; padding: 0; display: flex; flex-direction: column; gap: 10px; }
+.value-list li { display: flex; gap: 10px; align-items: baseline; padding: 8px 12px; background: var(--bh-bg-main); border-radius: var(--bh-radius-sm, 4px); }
+.value-name { flex: 0 0 110px; font-size: 13px; font-weight: 600; color: var(--bh-text-primary); }
+.value-meaning { font-size: 13px; color: var(--bh-text-regular); line-height: 1.6; }
+
+.charter-meta { margin-top: var(--bh-space-lg, 16px); background: var(--bh-bg-card); border-radius: var(--bh-radius-md, 8px); box-shadow: var(--bh-shadow-sm); padding: var(--bh-space-lg, 16px) var(--bh-space-xl, 20px); display: flex; flex-wrap: wrap; gap: var(--bh-space-xl, 20px); }
+.meta-item { display: flex; flex-direction: column; gap: 4px; min-width: 160px; }
+.meta-k { font-size: 12px; color: var(--bh-text-secondary); }
+.meta-v { font-size: 13px; color: var(--bh-text-primary); }
+
+.charter-form :deep(.el-form-item) { margin-bottom: 16px; }
+.form-hint { margin-left: 10px; font-size: 12px; color: var(--bh-text-placeholder); }
+.year-input { width: 140px; }
+.value-editor { width: 100%; display: flex; flex-direction: column; gap: 8px; }
+.value-row { display: flex; gap: 8px; align-items: center; }
+.value-row-name { width: 150px; flex: 0 0 150px; }
+.value-row-meaning { flex: 1; }
+</style>

frontend/src/views/StrategyMap.vue

@@ -25,6 +25,23 @@
       </div>
     </div>
 
+    <!-- ── 战略章程（P 战略基座：地图最顶层输入，只读；空态不显示，禁示例文案）── -->
+    <div v-if="charter" class="charter-strip">
+      <div class="cs-item cs-mission">
+        <span class="cs-k">使命</span>
+        <span class="cs-v">{{ charter.mission }}</span>
+      </div>
+      <div class="cs-item">
+        <span class="cs-k">愿景 · {{ charter.vision_year }}</span>
+        <span class="cs-v">{{ charter.vision }}</span>
+      </div>
+      <div class="cs-item cs-measurable">
+        <span class="cs-k">可验证状态</span>
+        <span class="cs-v">{{ charter.vision_measurable }}</span>
+      </div>
+      <router-link class="cs-link" to="/strategy-charter">查看章程</router-link>
+    </div>
+
     <!-- ── 口径/统计条 ── -->
     <div class="meta-bar">
       <!-- 口径标识位（常驻，非 hover）：与看板/KPI 列表共用同一组件与文案常量 -->
@@ -167,6 +184,7 @@ import {
   Money, User, Setting, Reading, TrendCharts,
 } from '@element-plus/icons-vue'
 import { mapApi } from '../api/index'
+import { charterApi } from '../api/index'
 import CaliberBadge from '../components/CaliberBadge.vue'
 import { caliberGapFor, currentCaliberEntityId, isCaliberKpiCode, type CaliberGapFacts } from '../constants/caliber'
 
@@ -175,6 +193,8 @@ const boardRef = ref<HTMLElement | null>(null)
 const loading = ref(false)
 const errorMsg = ref('')
 const board = ref<any>(null)
+// 战略章程（P 基座最顶层输入，只读；无行 → null，不显示任何文案）
+const charter = ref<any>(null)
 const period = ref<string>('')
 const periods = ref<string[]>([])
 const boardSize = ref({ w: 0, h: 0 })
@@ -317,6 +337,10 @@ function onResize() {
 }
 
 onMounted(async () => {
+  // 章程与地图并行取；章程失败不阻断地图（只读装饰项）
+  charterApi.get()
+    .then((res: any) => { if (res?.filled) charter.value = res.data })
+    .catch(() => { charter.value = null })
   await load()
   window.addEventListener('resize', onResize)
 })
@@ -367,6 +391,16 @@ function doPrint() {
 <style scoped>
 .strategy-map-page { padding: 16px; background: #f5f7fa; min-height: 100%; }
 
+/* 战略章程条（P 基座最顶层输入，只读） */
+.charter-strip { display: flex; align-items: center; gap: 18px; flex-wrap: wrap; padding: 10px 14px; margin-bottom: 10px; background: var(--bh-bg-card, #fff); border-left: 3px solid var(--bh-primary, #409eff); border-radius: var(--bh-radius-sm, 4px); box-shadow: var(--bh-shadow-sm, 0 1px 2px rgba(0,0,0,.06)); }
+.cs-item { display: flex; align-items: baseline; gap: 8px; min-width: 0; }
+.cs-k { flex: 0 0 auto; font-size: 12px; color: var(--bh-text-secondary, #909399); }
+.cs-v { font-size: 13px; color: var(--bh-text-primary, #303133); max-width: 420px; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
+.cs-mission .cs-v { font-weight: 600; }
+.cs-measurable .cs-v { color: var(--bh-text-regular, #606266); }
+.cs-link { margin-left: auto; font-size: 12px; color: var(--bh-primary, #409eff); text-decoration: none; }
+.cs-link:hover { text-decoration: underline; }
+
 .page-header { display: flex; align-items: center; justify-content: space-between; flex-wrap: wrap; gap: 10px; margin-bottom: 10px; }
 .page-subtitle { display: flex; align-items: center; gap: 8px; font-size: 14px; font-weight: 500; color: #64748B; margin: 0; }
 .title-icon { color: #64748B; font-size: 16px; }