perf(api): 会话/5数字弹窗超时修复——日粒度汇总层 msg_daily_stats

根因(服务器实测):sessions 对 owner 全量 message GROUP BY,重账号 5.2万条
3~5s;staff-chat-stats 全时段窗 6.9s——128MB buffer pool 下被推过前端
15s 超时(ECONNABORTED)。EXPLAIN ANALYZE 证实最优 covering 索引也要
18.9s,瓶颈是磁盘不是查询形状。

修复(预聚合,云客同款思路):
- msg_daily_stats 表:每「渠道×账号×会话×上海自然日」一行(收发计数/
  当日最后时间与方向);迁移 a7d3e9f1c2b4(幂等建表)
- 写路径:repos.upsert_messages 落库自动重算本批触及的会话×天(所有
  抓取路径汇到该函数,一处挂全;失败不阻塞写入)
- 读路径:sessions/staff-chat-stats 表有行时读汇总(namedtuple 同形,
  端点主体零改动),空表守卫回退旧聚合(部署回填前不空白)
- 自愈:CLI daily-stats 全量重建(幂等);full_sync_all.sh 夜间追加第 11
  步——顺带修掉尾部 sqlite3 直连 smoke.db 的硬编码(服务器 MySQL 上会炸)
- 口径统一:day 一律 Asia/Shanghai Python 计算(旧 from_unixtime 跟随
  DB 会话时区与今日数割裂);统计窗口对齐自然日(文档化语义变化)

测试:12 新增(汇总层 8 + 端点判别/对拍 4,判别手法=回填后删光 message
仍出完整数据),全量 512 绿。

Co-Authored-By: Claude Code <noreply@anthropic.com>

yunke-server/scripts/full_sync_all.sh

@@ -45,30 +45,54 @@ echo "$(date) --- 9c/10 Moments (员工朋友圈) ---"
 $PY scripts/scrape_via_page.py --entity moments
 echo "$(date) --- 10/10 Media download (audio+images) ---"
 $PY scripts/scrape_via_page.py --entity media
+# 11:会话日汇总全量重建(夜间自愈:绕过 upsert 钩子的写入由它纠正;
+# sessions/staff-chat-stats 读路径依赖此表,2026-09-03)
+echo "$(date) --- 11/11 Daily-stats rebuild ---"
+$PY -m app.sync.cli daily-stats
 
 echo "=============================================="
 echo "$(date) === FULL RE-SYNC COMPLETE ==="
 echo "=============================================="
 
-$PY -c "
-import sqlite3
-con = sqlite3.connect('smoke.db')
-c = con.cursor()
-print()
-print('=== FINAL DATA SUMMARY ===')
-for label, sql in [
-    ('wechat accounts', \"SELECT COUNT(*) FROM wechat_account\"),
-    ('whatsapp accounts', \"SELECT COUNT(*) FROM whatsapp_account\"),
-    ('wechat customers', \"SELECT COUNT(*) FROM customer WHERE channel='wechat'\"),
-    ('whatsapp customers', \"SELECT COUNT(*) FROM customer WHERE channel='whatsapp'\"),
-    ('wechat groups', \"SELECT COUNT(*) FROM chat_group WHERE channel='wechat'\"),
-    ('whatsapp groups', \"SELECT COUNT(*) FROM chat_group WHERE channel='whatsapp'\"),
-    ('wechat messages', \"SELECT COUNT(*) FROM message WHERE channel='wechat'\"),
-    ('whatsapp messages', \"SELECT COUNT(*) FROM message WHERE channel='whatsapp'\"),
-    ('voice calls', \"SELECT COUNT(*) FROM phone_call WHERE channel IN ('voice_call','wa_voice_call')\"),
-    ('voice clips', 'SELECT COUNT(*) FROM voice_clip'),
-    ('moments posts', 'SELECT COUNT(*) FROM moments_post'),
-    ('media files', 'SELECT COUNT(*) FROM media'),
-]:
-    print(f'  {label:22s} {c.execute(sql).fetchone()[0]:>6}')
-"
+# 终局数据快照:走 app 引擎(DB_URL 可移植,SQLite/MySQL 通吃;原 sqlite3
+# 直连 smoke.db 的写法在服务器 MySQL 上会炸,2026-09-03 修)
+$PY - <<'PYEOF'
+import asyncio
+
+from sqlalchemy import text
+from sqlalchemy.ext.asyncio import create_async_engine
+
+from app.config import get_settings
+
+TABLES = [
+    ("wechat accounts", "SELECT COUNT(*) FROM wechat_account"),
+    ("whatsapp accounts", "SELECT COUNT(*) FROM whatsapp_account"),
+    ("wechat customers", "SELECT COUNT(*) FROM customer WHERE channel='wechat'"),
+    ("whatsapp customers", "SELECT COUNT(*) FROM customer WHERE channel='whatsapp'"),
+    ("wechat groups", "SELECT COUNT(*) FROM chat_group WHERE channel='wechat'"),
+    ("whatsapp groups", "SELECT COUNT(*) FROM chat_group WHERE channel='whatsapp'"),
+    ("wechat messages", "SELECT COUNT(*) FROM message WHERE channel='wechat'"),
+    ("whatsapp messages", "SELECT COUNT(*) FROM message WHERE channel='whatsapp'"),
+    ("daily stats rows", "SELECT COUNT(*) FROM msg_daily_stats"),
+    ("voice calls", "SELECT COUNT(*) FROM phone_call WHERE channel IN ('voice_call','wa_voice_call')"),
+    ("voice clips", "SELECT COUNT(*) FROM voice_clip"),
+    ("moments posts", "SELECT COUNT(*) FROM moments_post"),
+    ("media files", "SELECT COUNT(*) FROM media"),
+]
+
+
+async def main() -> None:
+    engine = create_async_engine(get_settings().db_url)
+    try:
+        async with engine.connect() as conn:
+            print()
+            print("=== FINAL DATA SUMMARY ===")
+            for label, sql in TABLES:
+                n = (await conn.execute(text(sql))).scalar_one()
+                print(f"  {label:22s} {n:>6}")
+    finally:
+        await engine.dispose()
+
+
+asyncio.run(main())
+PYEOF