优化成员画像初始化状态管理,避免空群重复回扫历史窗口

- 新增成员摘要群级状态表,记录每个群的初始化状态、最近一次初始化时间、初始化窗口天数、生成摘要数量与触达成员数量
- 将成员画像群日摘要逻辑拆分为初始化模式与日常增量模式
- 首次运行且群内尚无日摘要时,自动按 bootstrap_days 回补最近已结束日期
- 初始化完成后写入群级状态:有产出记为 done,无产出记为 empty
- 后续调度或手动刷新时,若群级状态已是 done/empty,则不再重复按 60 天历史窗口回扫,改为仅按日常增量窗口处理前一天数据
- 解决冷群、空群、长期低活跃群在每天定时任务中反复初始化扫描的问题,减少无意义数据库扫描与等待时间
- 调整刷新群画像逻辑,即使最近 72 小时无活跃成员,首次初始化也能先尝试补历史摘要,再决定是否需要刷新成员画像
- 初始化模式下,会把历史窗口中真正产出过日摘要的成员补充进候选刷新集合,避免只依赖最近 72 小时活跃成员导致历史初始化不完整
This commit is contained in:
liuwei
2026-04-02 13:54:24 +08:00
parent 78ab3e65c2
commit 75dc3b289d
3 changed files with 141 additions and 4 deletions

View File

@@ -48,23 +48,64 @@ class MemberDigestService:
def ensure_recent_group_daily_digests(self, chatroom_id: str, days: Optional[int] = None,
force: bool = False) -> Dict:
days = days or self.group_digest_days
bootstrap_mode = False
if days is None:
bootstrap_mode = self._should_bootstrap_group(chatroom_id)
days = self.bootstrap_days if bootstrap_mode else self.group_digest_days
built_daily = 0
touched_members = set()
processed_dates = []
for offset in range(days):
target_date = (datetime.now() - timedelta(days=offset + 1)).strftime("%Y-%m-%d")
digests = self._build_group_daily_digests(chatroom_id, target_date, force=force)
processed_dates.append(target_date)
for digest in digests:
self.digest_db.save_digest(digest)
built_daily += 1
touched_members.add(digest.get("wxid"))
mode = "bootstrap" if bootstrap_mode else "incremental"
if bootstrap_mode:
bootstrap_status = "done" if built_daily > 0 or touched_members else "empty"
self.digest_db.save_group_state(
chatroom_id,
bootstrap_status=bootstrap_status,
bootstrap_days=days,
built_daily_count=built_daily,
touched_member_count=len(touched_members),
extra={"processed_dates": processed_dates},
)
self.LOG.info(
f"[成员交互摘要][群日摘要窗口] group={chatroom_id}, mode={mode}, days={days}, "
f"built_daily={built_daily}, touched_members={len(touched_members)}"
)
return {
"built_daily": built_daily,
"touched_members": list(touched_members),
"days": days,
"mode": mode,
"processed_dates": processed_dates,
}
def _should_bootstrap_group(self, chatroom_id: str) -> bool:
try:
group_state = self.digest_db.get_group_state(chatroom_id)
if group_state and group_state.get("bootstrap_status") in {"done", "empty"}:
return False
sql = """
SELECT 1
FROM t_member_digest
WHERE chatroom_id = %s AND digest_type = 'daily'
LIMIT 1
"""
row = self.digest_db.execute_query(sql, (chatroom_id,), fetch_one=True)
return not bool(row)
except Exception as e:
self.LOG.warning(f"[成员交互摘要] 检查群初始化状态失败,按增量处理: group={chatroom_id}, error={e}")
return False
def ensure_member_digest_pipeline(self, chatroom_id: str, wxid: str, force: bool = False) -> Dict:
member = self.contacts_db.get_chatroom_member_info(chatroom_id, wxid) or {}
display_name = member.get("display_name") or member.get("nick_name") or wxid

View File

@@ -236,9 +236,6 @@ class MemberContextService:
return {"refreshed": 0, "skipped": 0, "disabled": True}
active_members = self._get_recent_active_members(chatroom_id)
if not active_members:
self.LOG.info(f"{chatroom_id} 最近没有满足条件的活跃成员,跳过刷新")
return {"refreshed": 0, "skipped": 0, "disabled": False, "active_candidates": 0}
members = self.contacts_db.get_chatroom_member_list(chatroom_id) or []
enabled_members = {
@@ -257,10 +254,38 @@ class MemberContextService:
group_digest_stats = self.digest_service.ensure_recent_group_daily_digests(chatroom_id)
self.LOG.info(
f"[成员交互摘要] 群日摘要批处理完成: group={chatroom_id}, "
f"mode={group_digest_stats.get('mode', '')}, "
f"days={group_digest_stats.get('days', 0)}, "
f"built_daily={group_digest_stats.get('built_daily', 0)}, "
f"touched_members={len(group_digest_stats.get('touched_members', []))}"
)
candidate_map = {
(member.get("wxid") or ""): dict(member)
for member in active_members
if member.get("wxid")
}
if group_digest_stats.get("mode") == "bootstrap":
for wxid in group_digest_stats.get("touched_members", []):
if not wxid or wxid in candidate_map:
continue
candidate_map[wxid] = {
"wxid": wxid,
"msg_count": 0,
"latest_message_time": "",
}
self.LOG.info(
f"[成员交互摘要] 初始化模式补充成员: group={chatroom_id}, "
f"bootstrap_members={len(group_digest_stats.get('touched_members', []))}, "
f"candidate_total={len(candidate_map)}"
)
active_members = list(candidate_map.values())
total = len(active_members)
if not active_members:
self.LOG.info(f"{chatroom_id} 初始化/增量后仍无可刷新成员,跳过刷新")
return {"refreshed": 0, "skipped": 0, "disabled": False, "active_candidates": 0}
for index, active_member in enumerate(active_members, start=1):
wxid = active_member.get("wxid")
if wxid not in enabled_members: