将@关系批处理业务迁移到 value_rank 插件

- 从 MessageStorageDB 移除@抽取与社交图写入逻辑，消息层仅保留归档职责

- 从系统级任务移除 process_pending_mentions，取消 message_to_db 中对应入口

- 在 value_rank 插件新增定时动作 value_rank_mentions_extract（每10分钟）

- 在插件内实现窗口化批处理（默认10~20分钟前）、@提取、幂等写入明细/边表/日汇总及 unique_interactors 回填

- 新增插件侧可配置参数 mention_batch_size / mention_window_start_minutes / mention_window_end_minutes

db/message_storage.py

@@ -2,8 +2,6 @@
 
 from datetime import datetime
 import json
-import re
-import xml.etree.ElementTree as ET
 from typing import Dict, List, Optional
 
 from db.base import BaseDBOperator
@@ -91,400 +89,6 @@ class MessageStorageDB(BaseDBOperator):
             # 最后的保底策略：即使序列化失败，也确保字段有可追溯文本，避免丢失原始上下文。
             return str(msg)
 
-    def _extract_mentioned_user_ids(self, raw_xml: str) -> List[str]:
-        """从消息 XML 中提取被@用户ID列表，并返回去重后的列表。
-
-        解析策略：
-        1. 优先从 `msg.msg_source` 的 XML 里读取 `atuserlist` 节点；
-        2. 若 XML 解析失败，则退化为正则提取 `atuserlist` 文本；
-        3. 去重并过滤空值，保证输出稳定。
-
-        返回值示例：`["wxid_a", "wxid_b"]`
-        """
-        raw_xml = str(raw_xml or "")
-        if not raw_xml:
-            return []
-
-        at_user_list_text = ""
-        try:
-            root = ET.fromstring(raw_xml)
-            node = root.find(".//atuserlist")
-            if node is not None and node.text:
-                at_user_list_text = str(node.text).strip()
-        except Exception:
-            # 兼容异常格式 XML，采用正则兜底，确保尽量不丢数据。
-            match = re.search(r"<atuserlist><!\[CDATA\[(.*?)\]\]></atuserlist>", raw_xml, flags=re.IGNORECASE | re.DOTALL)
-            if match:
-                at_user_list_text = str(match.group(1) or "").strip()
-
-        if not at_user_list_text:
-            return []
-
-        # 微信 atuserlist 常见分隔符为 ','，但实际环境可能混入 ';' 或空白，这里统一兼容。
-        raw_ids = re.split(r"[,\s;]+", at_user_list_text)
-        seen = set()
-        result = []
-        for uid in raw_ids:
-            normalized_uid = str(uid or "").strip()
-            if not normalized_uid or normalized_uid in seen:
-                continue
-            seen.add(normalized_uid)
-            result.append(normalized_uid)
-
-        return result
-
-    def get_pending_mention_extract_messages(
-        self,
-        limit: int = 200,
-        window_start_minutes: int = 20,
-        window_end_minutes: int = 10,
-    ) -> List[Dict]:
-        """获取待处理 @ 抽取的消息批次。
-
-        筛选规则：
-        1. 群消息；
-        2. mentioned_user_ids 为空（表示还未处理）；
-        3. message_xml 非空；
-        4. 仅处理固定时间窗口（默认：10~20分钟前），降低扫描压力与热数据竞争。
-        """
-        # 兜底修正窗口参数，确保窗口有效：start > end >= 0
-        start_m = max(int(window_start_minutes), 1)
-        end_m = max(int(window_end_minutes), 0)
-        if start_m <= end_m:
-            start_m = end_m + 10
-            self.LOG.warning(
-                f"@抽取窗口参数异常，已自动修正: window_start_minutes={window_start_minutes}, "
-                f"window_end_minutes={window_end_minutes}, 修正后=[NOW-{start_m}m, NOW-{end_m}m)"
-            )
-
-        sql = """
-        SELECT message_id, group_id, sender, message_xml, timestamp
-        FROM messages
-        WHERE group_id IS NOT NULL
-          AND group_id <> ''
-          AND (mentioned_user_ids IS NULL OR mentioned_user_ids = '')
-          AND message_xml IS NOT NULL
-          AND message_xml <> ''
-          AND timestamp >= DATE_SUB(NOW(), INTERVAL %s MINUTE)
-          AND timestamp < DATE_SUB(NOW(), INTERVAL %s MINUTE)
-        ORDER BY timestamp ASC
-        LIMIT %s
-        """
-        rows = self.execute_query(sql, (start_m, end_m, limit)) or []
-        self.LOG.debug(
-            f"查询待抽取@消息: window=[NOW-{start_m}m, NOW-{end_m}m), limit={limit}, 命中={len(rows)}"
-        )
-        return rows
-
-    def process_pending_mentions(
-        self,
-        batch_size: int = 200,
-        window_start_minutes: int = 20,
-        window_end_minutes: int = 10,
-    ) -> Dict[str, int]:
-        """批量处理待抽取 @ 的消息，并同步社交图数据。
-
-        返回统计：
-        - total: 本批读取条数
-        - processed: 成功处理条数
-        - with_mentions: 提取到 @ 的消息条数
-        - failed: 失败条数
-        """
-        started_at = datetime.now()
-        self.LOG.info(
-            "开始执行@批处理: "
-            f"batch_size={batch_size}, window_start_minutes={window_start_minutes}, "
-            f"window_end_minutes={window_end_minutes}"
-        )
-
-        rows = self.get_pending_mention_extract_messages(
-            limit=batch_size,
-            window_start_minutes=window_start_minutes,
-            window_end_minutes=window_end_minutes,
-        )
-        if not rows:
-            elapsed_ms = int((datetime.now() - started_at).total_seconds() * 1000)
-            self.LOG.info(f"@批处理结束: 命中0条, 耗时={elapsed_ms}ms")
-            return {"total": 0, "processed": 0, "with_mentions": 0, "failed": 0}
-
-        processed = 0
-        with_mentions = 0
-        failed = 0
-        # 记录少量失败样本，便于快速定位问题，不刷屏。
-        fail_samples: List[str] = []
-
-        for idx, row in enumerate(rows, start=1):
-            try:
-                message_id = str(row.get("message_id") or "").strip()
-                group_id = str(row.get("group_id") or "").strip()
-                sender_id = str(row.get("sender") or "").strip()
-                raw_xml = str(row.get("message_xml") or "")
-                ts_raw = row.get("timestamp")
-                msg_time = self._safe_parse_message_time(ts_raw)
-
-                mentioned_user_ids = self._extract_mentioned_user_ids(raw_xml)
-                mentioned_user_ids_json = json.dumps(mentioned_user_ids, ensure_ascii=False)
-
-                self._update_message_mentioned_user_ids(
-                    message_id=message_id,
-                    group_id=group_id,
-                    sender_id=sender_id,
-                    mentioned_user_ids_json=mentioned_user_ids_json,
-                )
-                self._persist_mention_graph_data(
-                    group_id=group_id,
-                    sender_id=sender_id,
-                    message_id=message_id,
-                    mentioned_user_ids=mentioned_user_ids,
-                    msg_time=msg_time,
-                )
-
-                processed += 1
-                if mentioned_user_ids:
-                    with_mentions += 1
-                if idx <= 3:
-                    # 前3条打 debug 明细，便于确认当前批处理真实在工作。
-                    self.LOG.debug(
-                        f"@批处理样本[{idx}]: message_id={message_id}, group_id={group_id}, "
-                        f"sender={sender_id}, mentioned_count={len(mentioned_user_ids)}"
-                    )
-            except Exception as e:
-                failed += 1
-                self.LOG.error(f"处理待抽取@消息失败: message_id={row.get('message_id')}, error={e}")
-                if len(fail_samples) < 5:
-                    fail_samples.append(str(row.get("message_id") or ""))
-
-        elapsed_ms = int((datetime.now() - started_at).total_seconds() * 1000)
-        stats = {
-            "total": len(rows),
-            "processed": processed,
-            "with_mentions": with_mentions,
-            "failed": failed,
-        }
-        self.LOG.info(
-            f"@批处理结束: total={stats['total']}, processed={stats['processed']}, "
-            f"with_mentions={stats['with_mentions']}, failed={stats['failed']}, 耗时={elapsed_ms}ms, "
-            f"fail_samples={fail_samples}"
-        )
-        return stats
-
-    @staticmethod
-    def _safe_parse_message_time(value) -> datetime:
-        """安全解析消息时间，失败时回退到当前时间。"""
-        if isinstance(value, datetime):
-            return value
-        text = str(value or "").strip()
-        if not text:
-            return datetime.now()
-        try:
-            return datetime.strptime(text, "%Y-%m-%d %H:%M:%S")
-        except Exception:
-            return datetime.now()
-
-    def _update_message_mentioned_user_ids(
-        self,
-        message_id: str,
-        group_id: str,
-        sender_id: str,
-        mentioned_user_ids_json: str,
-    ) -> None:
-        """回填消息表的 mentioned_user_ids 字段。"""
-        self.execute_update(
-            """
-            UPDATE messages
-            SET mentioned_user_ids = %s
-            WHERE message_id = %s
-              AND group_id = %s
-              AND sender = %s
-            """,
-            (mentioned_user_ids_json, message_id, group_id, sender_id),
-        )
-
-    def _persist_mention_graph_data(
-        self,
-        group_id: str,
-        sender_id: str,
-        message_id: str,
-        mentioned_user_ids: List[str],
-        msg_time: datetime,
-    ) -> None:
-        """落盘社交图增量数据（明细 + 边 + 个人日汇总）。
-
-        设计原则：
-        1. 只在群消息中处理（group_id 为空直接忽略）；
-        2. 过滤无效 @ 目标（空值、@所有人、自己@自己）；
-        3. 统计写入失败不抛异常，不影响主消息归档。
-        """
-        # 非群消息或缺少关键字段时直接跳过，避免写入脏数据。
-        if not group_id or not sender_id or not message_id or not mentioned_user_ids:
-            return
-
-        # 统一清洗被@列表，避免重复统计。
-        invalid_mentions = {"notify@all", "all", "@all"}
-        clean_mentioned_ids: List[str] = []
-        seen = set()
-        for uid in mentioned_user_ids:
-            normalized_uid = str(uid or "").strip()
-            if (not normalized_uid or normalized_uid in invalid_mentions or
-                    normalized_uid == sender_id or normalized_uid in seen):
-                continue
-            seen.add(normalized_uid)
-            clean_mentioned_ids.append(normalized_uid)
-
-        if not clean_mentioned_ids:
-            return
-
-        try:
-            stat_date = msg_time.strftime("%Y-%m-%d")
-            msg_time_str = msg_time.strftime("%Y-%m-%d %H:%M:%S")
-
-            # 幂等控制：只处理“该消息中尚未写入明细表”的新增 @ 目标。
-            existed_rows = self.execute_query(
-                """
-                SELECT mentioned_user_id
-                FROM t_message_mentions
-                WHERE message_id = %s
-                  AND group_id = %s
-                  AND sender_id = %s
-                """,
-                (message_id, group_id, sender_id),
-            ) or []
-            existed_ids = {str(r.get("mentioned_user_id") or "").strip() for r in existed_rows}
-            newly_mentioned_ids = [uid for uid in clean_mentioned_ids if uid not in existed_ids]
-            if not newly_mentioned_ids:
-                self.LOG.debug(
-                    f"社交图写入跳过(无新增@关系): message_id={message_id}, group_id={group_id}, sender={sender_id}"
-                )
-                return
-
-            # 1) 写 @ 明细表：用于追溯“哪条消息 @ 了谁”。
-            mention_rows = [
-                (message_id, group_id, sender_id, mentioned_uid, stat_date, msg_time_str)
-                for mentioned_uid in newly_mentioned_ids
-            ]
-            self.execute_batch(
-                """
-                INSERT IGNORE INTO t_message_mentions
-                (message_id, group_id, sender_id, mentioned_user_id, stat_date, msg_time)
-                VALUES (%s, %s, %s, %s, %s, %s)
-                """,
-                mention_rows,
-            )
-
-            # 2) 写社交日边表：一条 @ 关系视为 sender -> mentioned_uid 的一条有向边增量。
-            edge_rows = [
-                (stat_date, group_id, sender_id, mentioned_uid, 1, 1.0)
-                for mentioned_uid in newly_mentioned_ids
-            ]
-            self.execute_batch(
-                """
-                INSERT INTO t_social_edges_daily
-                (stat_date, group_id, from_user_id, to_user_id, mention_count, interaction_score)
-                VALUES (%s, %s, %s, %s, %s, %s)
-                ON DUPLICATE KEY UPDATE
-                    mention_count = mention_count + VALUES(mention_count),
-                    interaction_score = interaction_score + VALUES(interaction_score),
-                    update_time = CURRENT_TIMESTAMP
-                """,
-                edge_rows,
-            )
-
-            # 3) 写个人日汇总：更新被@次数/主动@次数，供 value_rank 直接读取。
-            sender_social_row = (
-                stat_date, group_id, sender_id,
-                0, len(newly_mentioned_ids), 0,
-                float(len(newly_mentioned_ids)),
-            )
-            self.execute_update(
-                """
-                INSERT INTO t_value_rank_social_daily
-                (stat_date, group_id, user_id, mentioned_count, mention_others_count, unique_interactors, interaction_score)
-                VALUES (%s, %s, %s, %s, %s, %s, %s)
-                ON DUPLICATE KEY UPDATE
-                    mention_others_count = mention_others_count + VALUES(mention_others_count),
-                    interaction_score = interaction_score + VALUES(interaction_score),
-                    update_time = CURRENT_TIMESTAMP
-                """,
-                sender_social_row,
-            )
-
-            receiver_social_rows = [
-                (stat_date, group_id, mentioned_uid, 1, 0, 0, 1.0)
-                for mentioned_uid in newly_mentioned_ids
-            ]
-            self.execute_batch(
-                """
-                INSERT INTO t_value_rank_social_daily
-                (stat_date, group_id, user_id, mentioned_count, mention_others_count, unique_interactors, interaction_score)
-                VALUES (%s, %s, %s, %s, %s, %s, %s)
-                ON DUPLICATE KEY UPDATE
-                    mentioned_count = mentioned_count + VALUES(mentioned_count),
-                    interaction_score = interaction_score + VALUES(interaction_score),
-                    update_time = CURRENT_TIMESTAMP
-                """,
-                receiver_social_rows,
-            )
-
-            # 4) 回填 unique_interactors：针对本条消息受影响的用户实时重算“去重互动人数”。
-            affected_user_ids = [sender_id, *newly_mentioned_ids]
-            self._refresh_unique_interactors(stat_date, group_id, affected_user_ids)
-            self.LOG.debug(
-                f"社交图写入完成: message_id={message_id}, group_id={group_id}, sender={sender_id}, "
-                f"new_mentions={len(newly_mentioned_ids)}, affected_users={len(affected_user_ids)}"
-            )
-        except Exception as e:
-            # 社交图统计属于增强链路，不能反向影响主消息入库稳定性。
-            self.LOG.error(f"写入社交图增量数据失败: {e}")
-
-    def _refresh_unique_interactors(self, stat_date: str, group_id: str, user_ids: List[str]) -> None:
-        """重算并回填用户在指定日期内的去重互动人数。
-
-        定义：
-        - 某用户当天主动@过的人 + 被谁@过（去重并集）
-        """
-        if not user_ids:
-            return
-
-        deduped_user_ids = []
-        seen = set()
-        for uid in user_ids:
-            normalized_uid = str(uid or "").strip()
-            if not normalized_uid or normalized_uid in seen:
-                continue
-            seen.add(normalized_uid)
-            deduped_user_ids.append(normalized_uid)
-
-        for uid in deduped_user_ids:
-            try:
-                row = self.execute_query(
-                    """
-                    SELECT COUNT(DISTINCT partner_id) AS partner_count
-                    FROM (
-                        SELECT mentioned_user_id AS partner_id
-                        FROM t_message_mentions
-                        WHERE stat_date = %s AND group_id = %s AND sender_id = %s
-                        UNION
-                        SELECT sender_id AS partner_id
-                        FROM t_message_mentions
-                        WHERE stat_date = %s AND group_id = %s AND mentioned_user_id = %s
-                    ) t
-                    """,
-                    (stat_date, group_id, uid, stat_date, group_id, uid),
-                    fetch_one=True,
-                ) or {}
-                partner_count = int(row.get("partner_count") or 0)
-
-                self.execute_update(
-                    """
-                    UPDATE t_value_rank_social_daily
-                    SET unique_interactors = %s, update_time = CURRENT_TIMESTAMP
-                    WHERE stat_date = %s AND group_id = %s AND user_id = %s
-                    """,
-                    (partner_count, stat_date, group_id, uid),
-                )
-            except Exception as e:
-                self.LOG.error(f"回填 unique_interactors 失败: group={group_id}, user={uid}, err={e}")
-
     def get_recent_messages(self, group_id: str, hours_ago: int = 8, min_content_length: int = 6) -> List[Dict]:
         """获取最近的消息"""
         sql = """

plugins/value_rank/config.toml

@@ -28,3 +28,8 @@ base_score_scale = 1000
 # 排行默认展示数量
 default_rank_limit = 10
 max_rank_limit = 50
+
+# @关系批处理（插件定时任务）参数
+mention_batch_size = 200
+mention_window_start_minutes = 20
+mention_window_end_minutes = 10

plugins/value_rank/main.py

@@ -1,6 +1,8 @@
 # -*- coding: utf-8 -*-
 import json
 import math
+import re
+import xml.etree.ElementTree as ET
 from datetime import datetime, timedelta
 from typing import Any, Dict, List, Optional, Tuple
 
@@ -268,6 +270,145 @@ class ValueRankDB(BaseDBOperator):
             result[user_id] = float(row.get("score") or 0.0)
         return result
 
+    def get_pending_mention_extract_messages_for_group(
+        self,
+        group_id: str,
+        limit: int,
+        window_start_minutes: int,
+        window_end_minutes: int,
+    ) -> List[Dict[str, Any]]:
+        """按群读取待处理@抽取消息。"""
+        sql = """
+        SELECT message_id, group_id, sender, message_xml, timestamp
+        FROM messages
+        WHERE group_id = %s
+          AND (mentioned_user_ids IS NULL OR mentioned_user_ids = '')
+          AND message_xml IS NOT NULL
+          AND message_xml <> ''
+          AND timestamp >= DATE_SUB(NOW(), INTERVAL %s MINUTE)
+          AND timestamp < DATE_SUB(NOW(), INTERVAL %s MINUTE)
+        ORDER BY timestamp ASC
+        LIMIT %s
+        """
+        return self.execute_query(sql, (group_id, window_start_minutes, window_end_minutes, limit)) or []
+
+    def update_message_mentioned_user_ids(
+        self,
+        message_id: str,
+        group_id: str,
+        sender_id: str,
+        mentioned_user_ids_json: str,
+    ) -> bool:
+        """回填消息表的 mentioned_user_ids 字段。"""
+        return self.execute_update(
+            """
+            UPDATE messages
+            SET mentioned_user_ids = %s
+            WHERE message_id = %s
+              AND group_id = %s
+              AND sender = %s
+            """,
+            (mentioned_user_ids_json, message_id, group_id, sender_id),
+        )
+
+    def get_existing_mentions(self, message_id: str, group_id: str, sender_id: str) -> List[str]:
+        """查询某条消息已经入库的@关系，避免重复累加。"""
+        rows = self.execute_query(
+            """
+            SELECT mentioned_user_id
+            FROM t_message_mentions
+            WHERE message_id = %s
+              AND group_id = %s
+              AND sender_id = %s
+            """,
+            (message_id, group_id, sender_id),
+        ) or []
+        return [str(r.get("mentioned_user_id") or "").strip() for r in rows if str(r.get("mentioned_user_id") or "").strip()]
+
+    def insert_message_mentions(self, rows: List[Tuple[Any, ...]]) -> bool:
+        """批量写入@明细。"""
+        if not rows:
+            return True
+        return self.execute_batch(
+            """
+            INSERT IGNORE INTO t_message_mentions
+            (message_id, group_id, sender_id, mentioned_user_id, stat_date, msg_time)
+            VALUES (%s, %s, %s, %s, %s, %s)
+            """,
+            rows,
+        )
+
+    def upsert_social_edges_daily(self, rows: List[Tuple[Any, ...]]) -> bool:
+        """批量累加社交边。"""
+        if not rows:
+            return True
+        return self.execute_batch(
+            """
+            INSERT INTO t_social_edges_daily
+            (stat_date, group_id, from_user_id, to_user_id, mention_count, interaction_score)
+            VALUES (%s, %s, %s, %s, %s, %s)
+            ON DUPLICATE KEY UPDATE
+                mention_count = mention_count + VALUES(mention_count),
+                interaction_score = interaction_score + VALUES(interaction_score),
+                update_time = CURRENT_TIMESTAMP
+            """,
+            rows,
+        )
+
+    def upsert_social_daily_row(self, row: Tuple[Any, ...]) -> bool:
+        """写入或更新单个用户的社交日汇总。"""
+        return self.execute_update(
+            """
+            INSERT INTO t_value_rank_social_daily
+            (stat_date, group_id, user_id, mentioned_count, mention_others_count, unique_interactors, interaction_score)
+            VALUES (%s, %s, %s, %s, %s, %s, %s)
+            ON DUPLICATE KEY UPDATE
+                mentioned_count = mentioned_count + VALUES(mentioned_count),
+                mention_others_count = mention_others_count + VALUES(mention_others_count),
+                interaction_score = interaction_score + VALUES(interaction_score),
+                update_time = CURRENT_TIMESTAMP
+            """,
+            row,
+        )
+
+    def refresh_unique_interactors(self, stat_date: str, group_id: str, user_ids: List[str]) -> None:
+        """回填去重互动人数。"""
+        deduped = []
+        seen = set()
+        for uid in user_ids:
+            normalized = str(uid or "").strip()
+            if not normalized or normalized in seen:
+                continue
+            seen.add(normalized)
+            deduped.append(normalized)
+
+        for uid in deduped:
+            row = self.execute_query(
+                """
+                SELECT COUNT(DISTINCT partner_id) AS partner_count
+                FROM (
+                    SELECT mentioned_user_id AS partner_id
+                    FROM t_message_mentions
+                    WHERE stat_date = %s AND group_id = %s AND sender_id = %s
+                    UNION
+                    SELECT sender_id AS partner_id
+                    FROM t_message_mentions
+                    WHERE stat_date = %s AND group_id = %s AND mentioned_user_id = %s
+                ) t
+                """,
+                (stat_date, group_id, uid, stat_date, group_id, uid),
+                fetch_one=True,
+            ) or {}
+            partner_count = int(row.get("partner_count") or 0)
+            self.execute_update(
+                """
+                UPDATE t_value_rank_social_daily
+                SET unique_interactors = %s, update_time = CURRENT_TIMESTAMP
+                WHERE stat_date = %s AND group_id = %s AND user_id = %s
+                """,
+                (partner_count, stat_date, group_id, uid),
+            )
+
 
 class ValueRankPlugin(MessagePluginInterface):
     """群成员身价排行插件。
@@ -337,6 +478,9 @@ class ValueRankPlugin(MessagePluginInterface):
 
         self.default_rank_limit = 10
         self.max_rank_limit = 50
+        self.mention_batch_size = 200
+        self.mention_window_start_minutes = 20
+        self.mention_window_end_minutes = 10
 
     def initialize(self, context: Dict[str, Any]) -> bool:
         """初始化插件与配置。"""
@@ -361,6 +505,9 @@ class ValueRankPlugin(MessagePluginInterface):
 
         self.default_rank_limit = int(cfg.get("default_rank_limit", self.default_rank_limit))
         self.max_rank_limit = int(cfg.get("max_rank_limit", self.max_rank_limit))
+        self.mention_batch_size = int(cfg.get("mention_batch_size", self.mention_batch_size))
+        self.mention_window_start_minutes = int(cfg.get("mention_window_start_minutes", self.mention_window_start_minutes))
+        self.mention_window_end_minutes = int(cfg.get("mention_window_end_minutes", self.mention_window_end_minutes))
 
         # 权重归一化：避免配置误差导致总权重不为 1。
         weight_sum = self.points_weight + self.message_weight + self.active_days_weight + self.social_weight
@@ -470,6 +617,17 @@ class ValueRankPlugin(MessagePluginInterface):
                 "payload": {},
                 "default_enabled": True,
             },
+            {
+                "action_key": "value_rank_mentions_extract",
+                "name": "@关系批处理",
+                "description": "每10分钟批量抽取10-20分钟前的@关系并更新社交图",
+                "trigger_type": "every_seconds",
+                "trigger_config": {"seconds": 600},
+                "target_scope": "all_enabled_groups",
+                "target_config": {},
+                "payload": {},
+                "default_enabled": True,
+            },
             {
                 "action_key": "value_rank_weekly_report_push",
                 "name": "身价周报推送",
@@ -486,7 +644,7 @@ class ValueRankPlugin(MessagePluginInterface):
 
     async def run_scheduled_action(self, action_key: str, context: Dict[str, Any]) -> Dict[str, Any]:
         """执行调度动作。"""
-        if action_key not in {"value_rank_daily_recompute", "value_rank_weekly_report_push"}:
+        if action_key not in {"value_rank_daily_recompute", "value_rank_weekly_report_push", "value_rank_mentions_extract"}:
             return {"success": False, "summary": f"不支持动作: {action_key}", "detail": {}}
 
         target_groups = [str(g).strip() for g in (context.get("target_groups") or []) if str(g).strip()]
@@ -504,6 +662,35 @@ class ValueRankPlugin(MessagePluginInterface):
         stat_date = datetime.now().strftime("%Y-%m-%d")
         bot = context.get("bot") or getattr(self, "bot", None)
 
+        # @抽取任务不依赖 bot。
+        if action_key == "value_rank_mentions_extract":
+            total_stats = {"total": 0, "processed": 0, "with_mentions": 0, "failed": 0}
+            for gid in target_groups:
+                try:
+                    stats = self._process_pending_mentions_for_group(gid)
+                    total_stats["total"] += int(stats.get("total", 0))
+                    total_stats["processed"] += int(stats.get("processed", 0))
+                    total_stats["with_mentions"] += int(stats.get("with_mentions", 0))
+                    total_stats["failed"] += int(stats.get("failed", 0))
+                    success_groups.append(gid)
+                except Exception as e:
+                    failed_groups[gid] = str(e)
+
+            return {
+                "success": len(failed_groups) == 0,
+                "summary": (
+                    f"@关系批处理完成: 读取{total_stats['total']}条, "
+                    f"处理{total_stats['processed']}条, 含@{total_stats['with_mentions']}条, "
+                    f"失败{total_stats['failed']}条, 异常群{len(failed_groups)}个"
+                ),
+                "detail": {
+                    "window": f"[NOW-{self.mention_window_start_minutes}m, NOW-{self.mention_window_end_minutes}m)",
+                    "batch_size": self.mention_batch_size,
+                    "stats": total_stats,
+                    "failed_groups": failed_groups,
+                },
+            }
+
         # 周报任务先确保当日快照存在，再执行推送，避免“有报表命令但无数据”。
         if action_key == "value_rank_weekly_report_push" and not bot:
             return {"success": False, "summary": "周报推送失败：bot 未注入", "detail": {}}
@@ -864,6 +1051,172 @@ class ValueRankPlugin(MessagePluginInterface):
         lines.append("提示：分数由积分/发言/活跃/社交影响力综合计算。")
         return "\n".join(lines)
 
+    def _process_pending_mentions_for_group(self, group_id: str) -> Dict[str, int]:
+        """处理单群待抽取@消息（插件内定时业务）。"""
+        if not self.db:
+            return {"total": 0, "processed": 0, "with_mentions": 0, "failed": 0}
+
+        started_at = datetime.now()
+        window_start = max(int(self.mention_window_start_minutes), 1)
+        window_end = max(int(self.mention_window_end_minutes), 0)
+        if window_start <= window_end:
+            window_start = window_end + 10
+            self.LOG.warning(
+                f"[{self.name}] @窗口参数异常已修正: group={group_id}, "
+                f"window_start={self.mention_window_start_minutes}, "
+                f"window_end={self.mention_window_end_minutes}, fixed=[{window_start},{window_end}]"
+            )
+
+        rows = self.db.get_pending_mention_extract_messages_for_group(
+            group_id=group_id,
+            limit=self.mention_batch_size,
+            window_start_minutes=window_start,
+            window_end_minutes=window_end,
+        )
+        if not rows:
+            return {"total": 0, "processed": 0, "with_mentions": 0, "failed": 0}
+
+        processed, with_mentions, failed = 0, 0, 0
+        fail_samples: List[str] = []
+
+        for idx, row in enumerate(rows, start=1):
+            try:
+                message_id = str(row.get("message_id") or "").strip()
+                sender_id = str(row.get("sender") or "").strip()
+                raw_xml = str(row.get("message_xml") or "")
+                msg_time = self._safe_parse_message_time(row.get("timestamp"))
+
+                mentioned_ids = self._extract_mentioned_user_ids(raw_xml)
+                mentioned_ids_json = json.dumps(mentioned_ids, ensure_ascii=False)
+                self.db.update_message_mentioned_user_ids(
+                    message_id=message_id,
+                    group_id=group_id,
+                    sender_id=sender_id,
+                    mentioned_user_ids_json=mentioned_ids_json,
+                )
+
+                self._persist_mention_graph_data(
+                    group_id=group_id,
+                    sender_id=sender_id,
+                    message_id=message_id,
+                    mentioned_user_ids=mentioned_ids,
+                    msg_time=msg_time,
+                )
+
+                processed += 1
+                if mentioned_ids:
+                    with_mentions += 1
+                if idx <= 2:
+                    self.LOG.debug(
+                        f"[{self.name}] @抽取样本: group={group_id}, msg={message_id}, "
+                        f"sender={sender_id}, mentioned_count={len(mentioned_ids)}"
+                    )
+            except Exception as e:
+                failed += 1
+                if len(fail_samples) < 5:
+                    fail_samples.append(str(row.get("message_id") or ""))
+                self.LOG.error(f"[{self.name}] @抽取失败: group={group_id}, message_id={row.get('message_id')}, error={e}")
+
+        elapsed_ms = int((datetime.now() - started_at).total_seconds() * 1000)
+        stats = {"total": len(rows), "processed": processed, "with_mentions": with_mentions, "failed": failed}
+        self.LOG.info(
+            f"[{self.name}] @批处理完成: group={group_id}, total={stats['total']}, processed={processed}, "
+            f"with_mentions={with_mentions}, failed={failed}, cost={elapsed_ms}ms, fail_samples={fail_samples}"
+        )
+        return stats
+
+    @staticmethod
+    def _safe_parse_message_time(value: Any) -> datetime:
+        """安全解析消息时间，失败时回退到当前时间。"""
+        if isinstance(value, datetime):
+            return value
+        text = str(value or "").strip()
+        if not text:
+            return datetime.now()
+        try:
+            return datetime.strptime(text, "%Y-%m-%d %H:%M:%S")
+        except Exception:
+            return datetime.now()
+
+    @staticmethod
+    def _extract_mentioned_user_ids(raw_xml: str) -> List[str]:
+        """从消息 XML 提取@用户ID清单。"""
+        raw_xml = str(raw_xml or "")
+        if not raw_xml:
+            return []
+
+        at_user_list_text = ""
+        try:
+            root = ET.fromstring(raw_xml)
+            node = root.find(".//atuserlist")
+            if node is not None and node.text:
+                at_user_list_text = str(node.text).strip()
+        except Exception:
+            match = re.search(r"<atuserlist><!\[CDATA\[(.*?)\]\]></atuserlist>", raw_xml, flags=re.IGNORECASE | re.DOTALL)
+            if match:
+                at_user_list_text = str(match.group(1) or "").strip()
+
+        if not at_user_list_text:
+            return []
+
+        raw_ids = re.split(r"[,\s;]+", at_user_list_text)
+        seen = set()
+        result: List[str] = []
+        for uid in raw_ids:
+            normalized = str(uid or "").strip()
+            if not normalized or normalized in seen:
+                continue
+            seen.add(normalized)
+            result.append(normalized)
+        return result
+
+    def _persist_mention_graph_data(
+        self,
+        group_id: str,
+        sender_id: str,
+        message_id: str,
+        mentioned_user_ids: List[str],
+        msg_time: datetime,
+    ) -> None:
+        """落盘社交图增量数据（明细 + 边 + 个人日汇总）。"""
+        if not self.db or not group_id or not sender_id or not message_id:
+            return
+
+        invalid_mentions = {"notify@all", "all", "@all"}
+        clean_ids: List[str] = []
+        seen = set()
+        for uid in mentioned_user_ids:
+            normalized = str(uid or "").strip()
+            if (not normalized or normalized in invalid_mentions or normalized == sender_id or normalized in seen):
+                continue
+            seen.add(normalized)
+            clean_ids.append(normalized)
+
+        if not clean_ids:
+            return
+
+        existing = set(self.db.get_existing_mentions(message_id, group_id, sender_id))
+        new_ids = [uid for uid in clean_ids if uid not in existing]
+        if not new_ids:
+            return
+
+        stat_date = msg_time.strftime("%Y-%m-%d")
+        msg_time_text = msg_time.strftime("%Y-%m-%d %H:%M:%S")
+
+        mention_rows = [(message_id, group_id, sender_id, uid, stat_date, msg_time_text) for uid in new_ids]
+        self.db.insert_message_mentions(mention_rows)
+
+        edge_rows = [(stat_date, group_id, sender_id, uid, 1, 1.0) for uid in new_ids]
+        self.db.upsert_social_edges_daily(edge_rows)
+
+        # 发起方：主动@次数 + 互动分
+        self.db.upsert_social_daily_row((stat_date, group_id, sender_id, 0, len(new_ids), 0, float(len(new_ids))))
+        # 接收方：被@次数 + 互动分
+        for uid in new_ids:
+            self.db.upsert_social_daily_row((stat_date, group_id, uid, 1, 0, 0, 1.0))
+
+        self.db.refresh_unique_interactors(stat_date, group_id, [sender_id, *new_ids])
+
     def _build_explain_text(self) -> str:
         """输出算法说明文本。"""
         return (

utils/system_jobs.py

@@ -43,14 +43,6 @@ def get_system_job_definitions(robot) -> List[Dict[str, Any]]:
             "trigger_config": {"seconds": 300},
             "handler": _build_process_pending_images_handler(robot),
         },
-        {
-            "job_key": "process_pending_mentions",
-            "name": "待抽取@关系处理",
-            "description": "每 10 分钟处理一次待抽取@消息并更新社交图",
-            "trigger_type": "every_seconds",
-            "trigger_config": {"seconds": 600},
-            "handler": _build_process_pending_mentions_handler(robot),
-        },
     ]
 
 def _build_process_pending_images_handler(robot) -> Callable[[], Awaitable[None]]:
@@ -61,18 +53,6 @@ def _build_process_pending_images_handler(robot) -> Callable[[], Awaitable[None]
     return _handler
 
 
-def _build_process_pending_mentions_handler(robot) -> Callable[[], Awaitable[None]]:
-    async def _handler():
-        if hasattr(robot, "message_storage") and robot.message_storage:
-            await robot.message_storage.process_pending_mentions(
-                batch_size=200,
-                window_start_minutes=20,
-                window_end_minutes=10,
-            )
-
-    return _handler
-
-
 class SystemJobLoader:
     """系统任务加载器：从数据库读取调度配置并注册到 async_job。"""
 

utils/wechat/message_to_db.py

@@ -358,49 +358,6 @@ class MessageStorage:
         except Exception as e:
             logger.exception(f"定时处理媒体任务出错: {e}")
 
-    async def process_pending_mentions(
-        self,
-        batch_size: int = 200,
-        window_start_minutes: int = 20,
-        window_end_minutes: int = 10,
-    ):
-        """定时任务：批量处理待抽取 @ 的消息并写入社交图。
-
-        说明：
-        1. 该任务与主消息归档链路解耦，不阻塞实时收发；
-        2. 每次只处理有限批次，避免长事务和数据库抖动；
-        3. 重复执行安全：底层按 message_id + sender + mentioned_user_id 做幂等控制。
-        4. 默认只处理 10~20 分钟前的数据，减少对热数据区间的扫描压力。
-        """
-        try:
-            started_at = datetime.now()
-            logger.info(
-                "触发定时@抽取任务: "
-                f"batch_size={batch_size}, window=[NOW-{window_start_minutes}m, NOW-{window_end_minutes}m)"
-            )
-            stats = self.message_db.process_pending_mentions(
-                batch_size=batch_size,
-                window_start_minutes=window_start_minutes,
-                window_end_minutes=window_end_minutes,
-            )
-            total = int(stats.get("total", 0))
-            if total == 0:
-                elapsed_ms = int((datetime.now() - started_at).total_seconds() * 1000)
-                logger.info(f"定时@抽取任务结束: 无待处理数据, 耗时={elapsed_ms}ms")
-                return
-
-            elapsed_ms = int((datetime.now() - started_at).total_seconds() * 1000)
-            logger.info(
-                "定时@抽取任务结束: "
-                f"读取={stats.get('total', 0)}, "
-                f"处理={stats.get('processed', 0)}, "
-                f"含@={stats.get('with_mentions', 0)}, "
-                f"失败={stats.get('failed', 0)}, "
-                f"耗时={elapsed_ms}ms"
-            )
-        except Exception as e:
-            logger.exception(f"定时处理@抽取任务出错: {e}")
-
     def _process_image_done(self, future):
         """任务完成统一回调（极轻量）"""
         try: