Merge branch 'dev' of https://github.com/Windpicker-owo/MaiBot into dev

src/chat/chat_loop/heartFC_chat.py

@@ -88,11 +88,6 @@ class HeartFChatting:
 
         self.loop_mode = ChatMode.NORMAL  # 初始循环模式为普通模式
 
-        # 新增：消息计数器和疲惫阈值
-        self._message_count = 0  # 发送的消息计数
-        self._message_threshold = max(10, int(30 * global_config.chat.focus_value))
-        self._fatigue_triggered = False  # 是否已触发疲惫退出
-
         self.action_manager = ActionManager()
         self.action_planner = ActionPlanner(chat_id=self.stream_id, action_manager=self.action_manager)
         self.action_modifier = ActionModifier(action_manager=self.action_manager, chat_id=self.stream_id)
@@ -112,7 +107,6 @@ class HeartFChatting:
 
         self.last_read_time = time.time() - 1
 
-        self.willing_amplifier = 1
         self.willing_manager = get_willing_manager()
 
         logger.info(f"{self.log_prefix} HeartFChatting 初始化完成")
@@ -182,6 +176,9 @@ class HeartFChatting:
             if self.loop_mode == ChatMode.NORMAL:
                 self.energy_value -= 0.3
                 self.energy_value = max(self.energy_value, 0.3)
+            if self.loop_mode == ChatMode.FOCUS:
+                self.energy_value -= 0.6
+                self.energy_value = max(self.energy_value, 0.3)
 
     def print_cycle_info(self, cycle_timers):
         # 记录循环信息和计时器结果
@@ -200,9 +197,9 @@ class HeartFChatting:
     async def _loopbody(self):
         if self.loop_mode == ChatMode.FOCUS:
             if await self._observe():
-                self.energy_value -= 1 * global_config.chat.focus_value
+                self.energy_value -= 1 / global_config.chat.focus_value
             else:
-                self.energy_value -= 3 * global_config.chat.focus_value
+                self.energy_value -= 3 / global_config.chat.focus_value
             if self.energy_value <= 1:
                 self.energy_value = 1
                 self.loop_mode = ChatMode.NORMAL
@@ -218,15 +215,15 @@ class HeartFChatting:
                 limit_mode="earliest",
                 filter_bot=True,
             )
+            if global_config.chat.focus_value != 0:
+                if len(new_messages_data) > 3 / pow(global_config.chat.focus_value,0.5):
+                    self.loop_mode = ChatMode.FOCUS
+                    self.energy_value = 10 + (len(new_messages_data) / (3 / pow(global_config.chat.focus_value,0.5))) * 10
+                    return True
 
-            if len(new_messages_data) > 3 * global_config.chat.focus_value:
-                self.loop_mode = ChatMode.FOCUS
-                self.energy_value = 10 + (len(new_messages_data) / (3 * global_config.chat.focus_value)) * 10
-                return True
-
-            if self.energy_value >= 30 * global_config.chat.focus_value:
-                self.loop_mode = ChatMode.FOCUS
-                return True
+                if self.energy_value >= 30:
+                    self.loop_mode = ChatMode.FOCUS
+                    return True
 
             if new_messages_data:
                 earliest_messages_data = new_messages_data[0]
@@ -235,10 +232,10 @@ class HeartFChatting:
                 if_think = await self.normal_response(earliest_messages_data)
                 if if_think:
                     factor = max(global_config.chat.focus_value, 0.1)
-                    self.energy_value *= 1.1 / factor
+                    self.energy_value *= 1.1 * factor
                     logger.info(f"{self.log_prefix} 进行了思考，能量值按倍数增加，当前能量值：{self.energy_value:.1f}")
                 else:
-                    self.energy_value += 0.1 / global_config.chat.focus_value
+                    self.energy_value += 0.1 * global_config.chat.focus_value
                     logger.debug(f"{self.log_prefix} 没有进行思考，能量值线性增加，当前能量值：{self.energy_value:.1f}")
 
                 logger.debug(f"{self.log_prefix} 当前能量值：{self.energy_value:.1f}")
@@ -330,13 +327,13 @@ class HeartFChatting:
 
             if self.loop_mode == ChatMode.NORMAL:
                 if action_type == "no_action":
-                    logger.info(f"[{self.log_prefix}] {global_config.bot.nickname} 决定进行回复")
+                    logger.info(f"{self.log_prefix}{global_config.bot.nickname} 决定进行回复")
                 elif is_parallel:
                     logger.info(
-                        f"[{self.log_prefix}] {global_config.bot.nickname} 决定进行回复, 同时执行{action_type}动作"
+                        f"{self.log_prefix}{global_config.bot.nickname} 决定进行回复, 同时执行{action_type}动作"
                     )
                 else:
-                    logger.info(f"[{self.log_prefix}] {global_config.bot.nickname} 决定执行{action_type}动作")
+                    logger.info(f"{self.log_prefix}{global_config.bot.nickname} 决定执行{action_type}动作")
 
             if action_type == "no_action":
                 # 等待回复生成完毕
@@ -351,15 +348,15 @@ class HeartFChatting:
 
                 # 模型炸了，没有回复内容生成
                 if not response_set:
-                    logger.warning(f"[{self.log_prefix}] 模型未生成回复内容")
+                    logger.warning(f"{self.log_prefix}模型未生成回复内容")
                     return False
                 elif action_type not in ["no_action"] and not is_parallel:
                     logger.info(
-                        f"[{self.log_prefix}] {global_config.bot.nickname} 原本想要回复：{content}，但选择执行{action_type}，不发表回复"
+                        f"{self.log_prefix}{global_config.bot.nickname} 原本想要回复：{content}，但选择执行{action_type}，不发表回复"
                     )
                     return False
 
-                logger.info(f"[{self.log_prefix}] {global_config.bot.nickname} 决定的回复内容: {content}")
+                logger.info(f"{self.log_prefix}{global_config.bot.nickname} 决定的回复内容: {content}")
 
                 # 发送回复 (不再需要传入 chat)
                 reply_text = await self._send_response(response_set, reply_to_str, loop_start_time,message_data)
@@ -406,8 +403,18 @@ class HeartFChatting:
         if self.loop_mode == ChatMode.NORMAL:
             await self.willing_manager.after_generate_reply_handle(message_data.get("message_id", ""))
 
+        # 管理no_reply计数器：当执行了非no_reply动作时，重置计数器
         if action_type != "no_reply" and action_type != "no_action":
+            # 导入NoReplyAction并重置计数器
+            from src.plugins.built_in.core_actions.no_reply import NoReplyAction
+            NoReplyAction.reset_consecutive_count()
+            logger.info(f"{self.log_prefix} 执行了{action_type}动作，重置no_reply计数器")
             return True
+        elif action_type == "no_action":
+            # 当执行回复动作时，也重置no_reply计数器
+            from src.plugins.built_in.core_actions.no_reply import NoReplyAction
+            NoReplyAction.reset_consecutive_count()
+            logger.info(f"{self.log_prefix} 执行了回复动作，重置no_reply计数器")
 
         return True
 
@@ -501,7 +508,7 @@ class HeartFChatting:
         在"兴趣"模式下，判断是否回复并生成内容。
         """
 
-        interested_rate = (message_data.get("interest_value") or 0.0) * self.willing_amplifier
+        interested_rate = (message_data.get("interest_value") or 0.0) * global_config.chat.willing_amplifier
 
         self.willing_manager.setup(message_data, self.chat_stream)
 
@@ -515,8 +522,8 @@ class HeartFChatting:
                 reply_probability += additional_config["maimcore_reply_probability_gain"]
                 reply_probability = min(max(reply_probability, 0), 1)  # 确保概率在 0-1 之间
 
-            talk_frequency = global_config.chat.get_current_talk_frequency(self.stream_id)
-            reply_probability = talk_frequency * reply_probability
+        talk_frequency = global_config.chat.get_current_talk_frequency(self.stream_id)
+        reply_probability = talk_frequency * reply_probability
 
         # 处理表情包
         if message_data.get("is_emoji") or message_data.get("is_picid"):
@@ -563,7 +570,7 @@ class HeartFChatting:
             return reply_set
 
         except Exception as e:
-            logger.error(f"[{self.log_prefix}] 回复生成出现错误：{str(e)} {traceback.format_exc()}")
+            logger.error(f"{self.log_prefix}回复生成出现错误：{str(e)} {traceback.format_exc()}")
             return None
 
     async def _send_response(self, reply_set, reply_to, thinking_start_time, message_data):

src/chat/emoji_system/emoji_manager.py

@@ -525,9 +525,9 @@ class EmojiManager:
         如果文件已被删除，则执行对象的删除方法并从列表中移除
         """
         try:
-            if not self.emoji_objects:
-                logger.warning("[检查] emoji_objects为空，跳过完整性检查")
-                return
+            # if not self.emoji_objects:
+            #     logger.warning("[检查] emoji_objects为空，跳过完整性检查")
+            #     return
 
             total_count = len(self.emoji_objects)
             self.emoji_num = total_count
@@ -707,6 +707,38 @@ class EmojiManager:
                 return emoji
         return None  # 如果循环结束还没找到，则返回 None
 
+    async def get_emoji_description_by_hash(self, emoji_hash: str) -> Optional[str]:
+        """根据哈希值获取已注册表情包的描述
+        
+        Args:
+            emoji_hash: 表情包的哈希值
+            
+        Returns:
+            Optional[str]: 表情包描述，如果未找到则返回None
+        """
+        try:
+            # 先从内存中查找
+            emoji = await self.get_emoji_from_manager(emoji_hash)
+            if emoji and emoji.description:
+                logger.info(f"[缓存命中] 从内存获取表情包描述: {emoji.description[:50]}...")
+                return emoji.description
+            
+            # 如果内存中没有，从数据库查找
+            self._ensure_db()
+            try:
+                emoji_record = Emoji.get_or_none(Emoji.emoji_hash == emoji_hash)
+                if emoji_record and emoji_record.description:
+                    logger.info(f"[缓存命中] 从数据库获取表情包描述: {emoji_record.description[:50]}...")
+                    return emoji_record.description
+            except Exception as e:
+                logger.error(f"从数据库查询表情包描述时出错: {e}")
+            
+            return None
+            
+        except Exception as e:
+            logger.error(f"获取表情包描述失败 (Hash: {emoji_hash}): {str(e)}")
+            return None
+
     async def delete_emoji(self, emoji_hash: str) -> bool:
         """根据哈希值删除表情包
 

src/chat/express/expression_learner.py

@@ -330,48 +330,8 @@ class ExpressionLearner:
         """
         current_time = time.time()
 
-        # 全局衰减所有已存储的表达方式
-        for type in ["style", "grammar"]:
-            base_dir = os.path.join("data", "expression", f"learnt_{type}")
-            if not os.path.exists(base_dir):
-                logger.debug(f"目录不存在，跳过衰减: {base_dir}")
-                continue
-
-            try:
-                chat_ids = os.listdir(base_dir)
-                logger.debug(f"在 {base_dir} 中找到 {len(chat_ids)} 个聊天ID目录进行衰减")
-            except Exception as e:
-                logger.error(f"读取目录失败 {base_dir}: {e}")
-                continue
-
-            for chat_id in chat_ids:
-                file_path = os.path.join(base_dir, chat_id, "expressions.json")
-                if not os.path.exists(file_path):
-                    continue
-
-                try:
-                    with open(file_path, "r", encoding="utf-8") as f:
-                        expressions = json.load(f)
-
-                    if not isinstance(expressions, list):
-                        logger.warning(f"表达方式文件格式错误，跳过衰减: {file_path}")
-                        continue
-
-                    # 应用全局衰减
-                    decayed_expressions = self.apply_decay_to_expressions(expressions, current_time)
-
-                    # 保存衰减后的结果
-                    with open(file_path, "w", encoding="utf-8") as f:
-                        json.dump(decayed_expressions, f, ensure_ascii=False, indent=2)
-                        
-                    logger.debug(f"已对 {file_path} 应用衰减，剩余 {len(decayed_expressions)} 个表达方式")
-                except json.JSONDecodeError as e:
-                    logger.error(f"JSON解析失败，跳过衰减 {file_path}: {e}")
-                except PermissionError as e:
-                    logger.error(f"权限不足，无法更新 {file_path}: {e}")
-                except Exception as e:
-                    logger.error(f"全局衰减{type}表达方式失败 {file_path}: {e}")
-                    continue
+        # 全局衰减所有已存储的表达方式（直接操作数据库）
+        self._apply_global_decay_to_database(current_time)
 
         learnt_style: Optional[List[Tuple[str, str, str]]] = []
         learnt_grammar: Optional[List[Tuple[str, str, str]]] = []
@@ -388,6 +348,42 @@ class ExpressionLearner:
 
         return learnt_style, learnt_grammar
 
+    def _apply_global_decay_to_database(self, current_time: float) -> None:
+        """
+        对数据库中的所有表达方式应用全局衰减
+        """
+        try:
+            # 获取所有表达方式
+            all_expressions = Expression.select()
+            
+            updated_count = 0
+            deleted_count = 0
+            
+            for expr in all_expressions:
+                # 计算时间差
+                last_active = expr.last_active_time
+                time_diff_days = (current_time - last_active) / (24 * 3600)  # 转换为天
+                
+                # 计算衰减值
+                decay_value = self.calculate_decay_factor(time_diff_days)
+                new_count = max(0.01, expr.count - decay_value)
+                
+                if new_count <= 0.01:
+                    # 如果count太小，删除这个表达方式
+                    expr.delete_instance()
+                    deleted_count += 1
+                else:
+                    # 更新count
+                    expr.count = new_count
+                    expr.save()
+                    updated_count += 1
+            
+            if updated_count > 0 or deleted_count > 0:
+                logger.info(f"全局衰减完成：更新了 {updated_count} 个表达方式，删除了 {deleted_count} 个表达方式")
+                
+        except Exception as e:
+            logger.error(f"数据库全局衰减失败: {e}")
+
     def calculate_decay_factor(self, time_diff_days: float) -> float:
         """
         计算衰减值
@@ -410,30 +406,6 @@ class ExpressionLearner:
 
         return min(0.01, decay)
 
-    def apply_decay_to_expressions(
-        self, expressions: List[Dict[str, Any]], current_time: float
-    ) -> List[Dict[str, Any]]:
-        """
-        对表达式列表应用衰减
-        返回衰减后的表达式列表，移除count小于0的项
-        """
-        result = []
-        for expr in expressions:
-            # 确保last_active_time存在，如果不存在则使用current_time
-            if "last_active_time" not in expr:
-                expr["last_active_time"] = current_time
-
-            last_active = expr["last_active_time"]
-            time_diff_days = (current_time - last_active) / (24 * 3600)  # 转换为天
-
-            decay_value = self.calculate_decay_factor(time_diff_days)
-            expr["count"] = max(0.01, expr.get("count", 1) - decay_value)
-
-            if expr["count"] > 0:
-                result.append(expr)
-
-        return result
-
     async def learn_and_store(self, type: str, num: int = 10) -> List[Tuple[str, str, str]]:
         # sourcery skip: use-join
         """

src/chat/express/expression_selector.py

@@ -2,7 +2,7 @@ import json
 import time
 import random
 
-from typing import List, Dict, Tuple, Optional
+from typing import List, Dict, Tuple, Optional, Any
 from json_repair import repair_json
 
 from src.llm_models.utils_model import LLMRequest
@@ -117,36 +117,42 @@ class ExpressionSelector:
 
     def get_random_expressions(
         self, chat_id: str, total_num: int, style_percentage: float, grammar_percentage: float
-    ) -> Tuple[List[Dict[str, str]], List[Dict[str, str]]]:
+    ) -> Tuple[List[Dict[str, Any]], List[Dict[str, Any]]]:
         # 支持多chat_id合并抽选
         related_chat_ids = self.get_related_chat_ids(chat_id)
-        style_exprs = []
-        grammar_exprs = []
-        for cid in related_chat_ids:
-            style_query = Expression.select().where((Expression.chat_id == cid) & (Expression.type == "style"))
-            grammar_query = Expression.select().where((Expression.chat_id == cid) & (Expression.type == "grammar"))
-            style_exprs.extend([
-                {
-                    "situation": expr.situation,
-                    "style": expr.style,
-                    "count": expr.count,
-                    "last_active_time": expr.last_active_time,
-                    "source_id": cid,
-                    "type": "style",
-                    "create_date": expr.create_date if expr.create_date is not None else expr.last_active_time,
-                } for expr in style_query
-            ])
-            grammar_exprs.extend([
-                {
-                    "situation": expr.situation,
-                    "style": expr.style,
-                    "count": expr.count,
-                    "last_active_time": expr.last_active_time,
-                    "source_id": cid,
-                    "type": "grammar",
-                    "create_date": expr.create_date if expr.create_date is not None else expr.last_active_time,
-                } for expr in grammar_query
-            ])
+        
+        # 优化：一次性查询所有相关chat_id的表达方式
+        style_query = Expression.select().where(
+            (Expression.chat_id.in_(related_chat_ids)) & (Expression.type == "style")
+        )
+        grammar_query = Expression.select().where(
+            (Expression.chat_id.in_(related_chat_ids)) & (Expression.type == "grammar")
+        )
+        
+        style_exprs = [
+            {
+                "situation": expr.situation,
+                "style": expr.style,
+                "count": expr.count,
+                "last_active_time": expr.last_active_time,
+                "source_id": expr.chat_id,
+                "type": "style",
+                "create_date": expr.create_date if expr.create_date is not None else expr.last_active_time,
+            } for expr in style_query
+        ]
+        
+        grammar_exprs = [
+            {
+                "situation": expr.situation,
+                "style": expr.style,
+                "count": expr.count,
+                "last_active_time": expr.last_active_time,
+                "source_id": expr.chat_id,
+                "type": "grammar",
+                "create_date": expr.create_date if expr.create_date is not None else expr.last_active_time,
+            } for expr in grammar_query
+        ]
+        
         style_num = int(total_num * style_percentage)
         grammar_num = int(total_num * grammar_percentage)
         # 按权重抽样（使用count作为权重）
@@ -162,7 +168,7 @@ class ExpressionSelector:
             selected_grammar = []
         return selected_style, selected_grammar
 
-    def update_expressions_count_batch(self, expressions_to_update: List[Dict[str, str]], increment: float = 0.1):
+    def update_expressions_count_batch(self, expressions_to_update: List[Dict[str, Any]], increment: float = 0.1):
         """对一批表达方式更新count值，按chat_id+type分组后一次性写入数据库"""
         if not expressions_to_update:
             return
@@ -203,7 +209,7 @@ class ExpressionSelector:
         max_num: int = 10,
         min_num: int = 5,
         target_message: Optional[str] = None,
-    ) -> List[Dict[str, str]]:
+    ) -> List[Dict[str, Any]]:
         # sourcery skip: inline-variable, list-comprehension
         """使用LLM选择适合的表达方式"""
 

src/chat/heart_flow/heartflow_message_processor.py

@@ -12,6 +12,7 @@ from src.chat.message_receive.storage import MessageStorage
 from src.chat.heart_flow.heartflow import heartflow
 from src.chat.utils.utils import is_mentioned_bot_in_message
 from src.chat.utils.timer_calculator import Timer
+from src.chat.utils.chat_message_builder import replace_user_references_sync
 from src.common.logger import get_logger
 from src.person_info.relationship_manager import get_relationship_manager
 from src.mood.mood_manager import mood_manager
@@ -56,16 +57,41 @@ async def _calculate_interest(message: MessageRecv) -> Tuple[float, bool]:
     with Timer("记忆激活"):
         interested_rate = await hippocampus_manager.get_activate_from_text(
             message.processed_plain_text,
+            max_depth= 5,
             fast_retrieval=False,
         )
         logger.debug(f"记忆激活率: {interested_rate:.2f}")
 
     text_len = len(message.processed_plain_text)
-    # 根据文本长度调整兴趣度，长度越大兴趣度越高，但增长率递减，最低0.01，最高0.05
-    # 采用对数函数实现递减增长
-
-    base_interest = 0.01 + (0.05 - 0.01) * (math.log10(text_len + 1) / math.log10(1000 + 1))
-    base_interest = min(max(base_interest, 0.01), 0.05)
+    # 根据文本长度分布调整兴趣度，采用分段函数实现更精确的兴趣度计算
+    # 基于实际分布：0-5字符(26.57%), 6-10字符(27.18%), 11-20字符(22.76%), 21-30字符(10.33%), 31+字符(13.86%)
+    
+    if text_len == 0:
+        base_interest = 0.01  # 空消息最低兴趣度
+    elif text_len <= 5:
+        # 1-5字符：线性增长 0.01 -> 0.03
+        base_interest = 0.01 + (text_len - 1) * (0.03 - 0.01) / 4
+    elif text_len <= 10:
+        # 6-10字符：线性增长 0.03 -> 0.06
+        base_interest = 0.03 + (text_len - 5) * (0.06 - 0.03) / 5
+    elif text_len <= 20:
+        # 11-20字符：线性增长 0.06 -> 0.12
+        base_interest = 0.06 + (text_len - 10) * (0.12 - 0.06) / 10
+    elif text_len <= 30:
+        # 21-30字符：线性增长 0.12 -> 0.18
+        base_interest = 0.12 + (text_len - 20) * (0.18 - 0.12) / 10
+    elif text_len <= 50:
+        # 31-50字符：线性增长 0.18 -> 0.22
+        base_interest = 0.18 + (text_len - 30) * (0.22 - 0.18) / 20
+    elif text_len <= 100:
+        # 51-100字符：线性增长 0.22 -> 0.26
+        base_interest = 0.22 + (text_len - 50) * (0.26 - 0.22) / 50
+    else:
+        # 100+字符：对数增长 0.26 -> 0.3，增长率递减
+        base_interest = 0.26 + (0.3 - 0.26) * (math.log10(text_len - 99) / math.log10(901))  # 1000-99=901
+    
+    # 确保在范围内
+    base_interest = min(max(base_interest, 0.01), 0.3)
 
     interested_rate += base_interest
 
@@ -123,6 +149,13 @@ class HeartFCMessageReceiver:
             # 如果消息中包含图片标识，则将 [picid:...] 替换为 [图片]
             picid_pattern = r"\[picid:([^\]]+)\]"
             processed_plain_text = re.sub(picid_pattern, "[图片]", message.processed_plain_text)
+            
+            # 应用用户引用格式替换，将回复<aaa:bbb>和@<aaa:bbb>格式转换为可读格式
+            processed_plain_text = replace_user_references_sync(
+                processed_plain_text,
+                message.message_info.platform, # type: ignore
+                replace_bot_name=True
+            )
 
             logger.info(f"[{mes_name}]{userinfo.user_nickname}:{processed_plain_text}[兴趣度：{interested_rate:.2f}]")  # type: ignore
 

src/chat/memory_system/Hippocampus.py

@@ -224,10 +224,16 @@ class Hippocampus:
         return hash((source, target))
 
     @staticmethod
-    def find_topic_llm(text, topic_num):
+    def find_topic_llm(text: str, topic_num: int | list[int]):
         # sourcery skip: inline-immediately-returned-variable
+        topic_num_str = ""
+        if isinstance(topic_num, list):
+            topic_num_str = f"{topic_num[0]}-{topic_num[1]}"
+        else:
+            topic_num_str = topic_num
+
         prompt = (
-            f"这是一段文字：\n{text}\n\n请你从这段话中总结出最多{topic_num}个关键的概念，可以是名词，动词，或者特定人物，帮我列出来，"
+            f"这是一段文字：\n{text}\n\n请你从这段话中总结出最多{topic_num_str}个关键的概念，可以是名词，动词，或者特定人物，帮我列出来，"
             f"将主题用逗号隔开，并加上<>,例如<主题1>,<主题2>......尽可能精简。只需要列举最多{topic_num}个话题就好，不要有序号，不要告诉我其他内容。"
             f"如果确定找不出主题或者没有明显主题，返回<none>。"
         )
@@ -300,6 +306,60 @@ class Hippocampus:
         memories.sort(key=lambda x: x[2], reverse=True)
         return memories
 
+    async def get_keywords_from_text(self, text: str) -> list:
+        """从文本中提取关键词。
+
+        Args:
+            text (str): 输入文本
+            fast_retrieval (bool, optional): 是否使用快速检索。默认为False。
+                如果为True，使用jieba分词提取关键词，速度更快但可能不够准确。
+                如果为False，使用LLM提取关键词，速度较慢但更准确。
+        """
+        if not text:
+            return []
+
+        # 使用LLM提取关键词 - 根据详细文本长度分布优化topic_num计算
+        text_length = len(text)
+        topic_num: int | list[int] = 0
+        if text_length <= 5:
+            words = jieba.cut(text)
+            keywords = [word for word in words if len(word) > 1]
+            keywords = list(set(keywords))[:3]  # 限制最多3个关键词
+            if keywords:
+                logger.info(f"提取关键词: {keywords}")
+            return keywords
+        elif text_length <= 10:
+            topic_num = [1, 3]  # 6-10字符: 1个关键词 (27.18%的文本)
+        elif text_length <= 20:
+            topic_num = [2, 4]  # 11-20字符: 2个关键词 (22.76%的文本)
+        elif text_length <= 30:
+            topic_num = [3, 5]  # 21-30字符: 3个关键词 (10.33%的文本)
+        elif text_length <= 50:
+            topic_num = [4, 5]  # 31-50字符: 4个关键词 (9.79%的文本)
+        else:
+            topic_num = 5  # 51+字符: 5个关键词 (其余长文本)
+
+        topics_response, (reasoning_content, model_name) = await self.model_summary.generate_response_async(
+            self.find_topic_llm(text, topic_num)
+        )
+
+        # 提取关键词
+        keywords = re.findall(r"<([^>]+)>", topics_response)
+        if not keywords:
+            keywords = []
+        else:
+            keywords = [
+                keyword.strip()
+                for keyword in ",".join(keywords).replace("，", ",").replace("、", ",").replace(" ", ",").split(",")
+                if keyword.strip()
+            ]
+        
+        if keywords:
+            logger.info(f"提取关键词: {keywords}")
+        
+        return keywords 
+        
+
     async def get_memory_from_text(
         self,
         text: str,
@@ -325,39 +385,7 @@ class Hippocampus:
                 - memory_items: list, 该主题下的记忆项列表
                 - similarity: float, 与文本的相似度
         """
-        if not text:
-            return []
-
-        if fast_retrieval:
-            # 使用jieba分词提取关键词
-            words = jieba.cut(text)
-            # 过滤掉停用词和单字词
-            keywords = [word for word in words if len(word) > 1]
-            # 去重
-            keywords = list(set(keywords))
-            # 限制关键词数量
-            logger.debug(f"提取关键词: {keywords}")
-
-        else:
-            # 使用LLM提取关键词
-            topic_num = min(5, max(1, int(len(text) * 0.1)))  # 根据文本长度动态调整关键词数量
-            # logger.info(f"提取关键词数量: {topic_num}")
-            topics_response, (reasoning_content, model_name) = await self.model_summary.generate_response_async(
-                self.find_topic_llm(text, topic_num)
-            )
-
-            # 提取关键词
-            keywords = re.findall(r"<([^>]+)>", topics_response)
-            if not keywords:
-                keywords = []
-            else:
-                keywords = [
-                    keyword.strip()
-                    for keyword in ",".join(keywords).replace("，", ",").replace("、", ",").replace(" ", ",").split(",")
-                    if keyword.strip()
-                ]
-
-        # logger.info(f"提取的关键词: {', '.join(keywords)}")
+        keywords = await self.get_keywords_from_text(text)
 
         # 过滤掉不存在于记忆图中的关键词
         valid_keywords = [keyword for keyword in keywords if keyword in self.memory_graph.G]
@@ -679,38 +707,7 @@ class Hippocampus:
         Returns:
             float: 激活节点数与总节点数的比值
         """
-        if not text:
-            return 0
-
-        if fast_retrieval:
-            # 使用jieba分词提取关键词
-            words = jieba.cut(text)
-            # 过滤掉停用词和单字词
-            keywords = [word for word in words if len(word) > 1]
-            # 去重
-            keywords = list(set(keywords))
-            # 限制关键词数量
-            keywords = keywords[:5]
-        else:
-            # 使用LLM提取关键词
-            topic_num = min(5, max(1, int(len(text) * 0.1)))  # 根据文本长度动态调整关键词数量
-            # logger.info(f"提取关键词数量: {topic_num}")
-            topics_response, (reasoning_content, model_name) = await self.model_summary.generate_response_async(
-                self.find_topic_llm(text, topic_num)
-            )
-
-            # 提取关键词
-            keywords = re.findall(r"<([^>]+)>", topics_response)
-            if not keywords:
-                keywords = []
-            else:
-                keywords = [
-                    keyword.strip()
-                    for keyword in ",".join(keywords).replace("，", ",").replace("、", ",").replace(" ", ",").split(",")
-                    if keyword.strip()
-                ]
-
-        # logger.info(f"提取的关键词: {', '.join(keywords)}")
+        keywords = await self.get_keywords_from_text(text)
 
         # 过滤掉不存在于记忆图中的关键词
         valid_keywords = [keyword for keyword in keywords if keyword in self.memory_graph.G]
@@ -727,7 +724,7 @@ class Hippocampus:
         for keyword in valid_keywords:
             logger.debug(f"开始以关键词 '{keyword}' 为中心进行扩散检索 (最大深度: {max_depth}):")
             # 初始化激活值
-            activation_values = {keyword: 1.0}
+            activation_values = {keyword: 1.5}
             # 记录已访问的节点
             visited_nodes = {keyword}
             # 待处理的节点队列，每个元素是(节点, 激活值, 当前深度)
@@ -1315,6 +1312,7 @@ class ParahippocampalGyrus:
         return compressed_memory, similar_topics_dict
 
     async def operation_build_memory(self):
+        # sourcery skip: merge-list-appends-into-extend
         logger.info("------------------------------------开始构建记忆--------------------------------------")
         start_time = time.time()
         memory_samples = self.hippocampus.entorhinal_cortex.get_memory_sample()

src/chat/replyer/default_generator.py

@@ -17,7 +17,11 @@ from src.chat.message_receive.uni_message_sender import HeartFCSender
 from src.chat.utils.timer_calculator import Timer  # <--- Import Timer
 from src.chat.utils.utils import get_chat_type_and_target_info
 from src.chat.utils.prompt_builder import Prompt, global_prompt_manager
-from src.chat.utils.chat_message_builder import build_readable_messages, get_raw_msg_before_timestamp_with_chat
+from src.chat.utils.chat_message_builder import (
+    build_readable_messages,
+    get_raw_msg_before_timestamp_with_chat,
+    replace_user_references_sync,
+)
 from src.chat.express.expression_selector import expression_selector
 from src.chat.knowledge.knowledge_lib import qa_manager
 from src.chat.memory_system.memory_activator import MemoryActivator
@@ -30,6 +34,7 @@ from src.plugin_system.base.component_types import ActionInfo
 
 logger = get_logger("replyer")
 
+
 def init_prompt():
     Prompt("你正在qq群里聊天，下面是群里在聊的内容：", "chat_target_group1")
     Prompt("你正在和{sender_name}聊天，这是你们之前聊的内容：", "chat_target_private1")
@@ -356,17 +361,20 @@ class DefaultReplyer:
         expression_habits_block = ""
         expression_habits_title = ""
         if style_habits_str.strip():
-            expression_habits_title = "你可以参考以下的语言习惯，当情景合适就使用，但不要生硬使用，以合理的方式结合到你的回复中："
+            expression_habits_title = (
+                "你可以参考以下的语言习惯，当情景合适就使用，但不要生硬使用，以合理的方式结合到你的回复中："
+            )
             expression_habits_block += f"{style_habits_str}\n"
         if grammar_habits_str.strip():
-            expression_habits_title = "你可以选择下面的句法进行回复，如果情景合适就使用，不要盲目使用,不要生硬使用，以合理的方式使用："
+            expression_habits_title = (
+                "你可以选择下面的句法进行回复，如果情景合适就使用，不要盲目使用,不要生硬使用，以合理的方式使用："
+            )
             expression_habits_block += f"{grammar_habits_str}\n"
-            
+
         if style_habits_str.strip() and grammar_habits_str.strip():
             expression_habits_title = "你可以参考以下的语言习惯和句法，如果情景合适就使用，不要盲目使用,不要生硬使用，以合理的方式结合到你的回复中："
-            
+
         expression_habits_block = f"{expression_habits_title}\n{expression_habits_block}"
-        
 
         return expression_habits_block
 
@@ -375,27 +383,27 @@ class DefaultReplyer:
             return ""
 
         instant_memory = None
-        
+
         running_memories = await self.memory_activator.activate_memory_with_chat_history(
             target_message=target, chat_history_prompt=chat_history
         )
-        
+
         if global_config.memory.enable_instant_memory:
             asyncio.create_task(self.instant_memory.create_and_store_memory(chat_history))
 
             instant_memory = await self.instant_memory.get_memory(target)
             logger.info(f"即时记忆：{instant_memory}")
-            
+
         if not running_memories:
             return ""
 
         memory_str = "以下是当前在聊天中，你回忆起的记忆：\n"
         for running_memory in running_memories:
             memory_str += f"- {running_memory['content']}\n"
-        
+
         if instant_memory:
             memory_str += f"- {instant_memory}\n"
-            
+
         return memory_str
 
     async def build_tool_info(self, chat_history, reply_data: Optional[Dict], enable_tool: bool = True):
@@ -438,7 +446,7 @@ class DefaultReplyer:
 
                 tool_info_str += "以上是你获取到的实时信息，请在回复时参考这些信息。"
                 logger.info(f"获取到 {len(tool_results)} 个工具结果")
-                
+
                 return tool_info_str
             else:
                 logger.debug("未获取到任何工具结果")
@@ -469,7 +477,7 @@ class DefaultReplyer:
             # 添加None检查，防止NoneType错误
             if target is None:
                 return keywords_reaction_prompt
-                
+
             # 处理关键词规则
             for rule in global_config.keyword_reaction.keyword_rules:
                 if any(keyword in target for keyword in rule.keywords):
@@ -621,7 +629,7 @@ class DefaultReplyer:
         is_group_chat = bool(chat_stream.group_info)
         reply_to = reply_data.get("reply_to", "none")
         extra_info_block = reply_data.get("extra_info", "") or reply_data.get("extra_info_block", "")
-            
+
         if global_config.mood.enable_mood:
             chat_mood = mood_manager.get_mood_by_chat_id(chat_id)
             mood_prompt = chat_mood.mood_state
@@ -630,6 +638,8 @@ class DefaultReplyer:
 
         sender, target = self._parse_reply_target(reply_to)
 
+        target = replace_user_references_sync(target, chat_stream.platform, replace_bot_name=True)
+
         # 构建action描述 (如果启用planner)
         action_descriptions = ""
         if available_actions:
@@ -679,25 +689,21 @@ class DefaultReplyer:
             self._time_and_run_task(
                 self.build_expression_habits(chat_talking_prompt_short, target), "expression_habits"
             ),
-            self._time_and_run_task(
-                self.build_relation_info(reply_data), "relation_info"
-            ),
+            self._time_and_run_task(self.build_relation_info(reply_data), "relation_info"),
             self._time_and_run_task(self.build_memory_block(chat_talking_prompt_short, target), "memory_block"),
             self._time_and_run_task(
                 self.build_tool_info(chat_talking_prompt_short, reply_data, enable_tool=enable_tool), "tool_info"
             ),
-            self._time_and_run_task(
-                get_prompt_info(target, threshold=0.38), "prompt_info"
-            ),
+            self._time_and_run_task(get_prompt_info(target, threshold=0.38), "prompt_info"),
         )
 
         # 任务名称中英文映射
         task_name_mapping = {
             "expression_habits": "选取表达方式",
-            "relation_info": "感受关系", 
+            "relation_info": "感受关系",
             "memory_block": "回忆",
             "tool_info": "使用工具",
-            "prompt_info": "获取知识"
+            "prompt_info": "获取知识",
         }
 
         # 处理结果
@@ -790,7 +796,7 @@ class DefaultReplyer:
             core_dialogue_prompt, background_dialogue_prompt = self.build_s4u_chat_history_prompts(
                 message_list_before_now_long, target_user_id
             )
-            
+
             self.build_mai_think_context(
                 chat_id=chat_id,
                 memory_block=memory_block,
@@ -807,9 +813,8 @@ class DefaultReplyer:
 --------------------------------
 {time_block}
 这是你和{sender}的对话，你们正在交流中：
-{core_dialogue_prompt}"""
+{core_dialogue_prompt}""",
             )
-            
 
             # 使用 s4u 风格的模板
             template_name = "s4u_style_prompt"
@@ -847,9 +852,9 @@ class DefaultReplyer:
                 identity_block=identity_block,
                 sender=sender,
                 target=target,
-                chat_info=chat_talking_prompt
+                chat_info=chat_talking_prompt,
             )
-            
+
             # 使用原有的模式
             return await global_prompt_manager.format_prompt(
                 template_name,
@@ -1071,9 +1076,11 @@ async def get_prompt_info(message: str, threshold: float):
             related_info += found_knowledge_from_lpmm
             logger.debug(f"获取知识库内容耗时: {(end_time - start_time):.3f}秒")
             logger.debug(f"获取知识库内容，相关信息：{related_info[:100]}...，信息长度: {len(related_info)}")
-            
+
             # 格式化知识信息
-            formatted_prompt_info = await global_prompt_manager.format_prompt("knowledge_prompt", prompt_info=related_info)
+            formatted_prompt_info = await global_prompt_manager.format_prompt(
+                "knowledge_prompt", prompt_info=related_info
+            )
             return formatted_prompt_info
         else:
             logger.debug("从LPMM知识库获取知识失败，可能是从未导入过知识，返回空知识...")

src/chat/utils/chat_message_builder.py

@@ -2,7 +2,7 @@ import time  # 导入 time 模块以获取当前时间
 import random
 import re
 
-from typing import List, Dict, Any, Tuple, Optional
+from typing import List, Dict, Any, Tuple, Optional, Callable
 from rich.traceback import install
 
 from src.config.config import global_config
@@ -10,11 +10,161 @@ from src.common.message_repository import find_messages, count_messages
 from src.common.database.database_model import ActionRecords
 from src.common.database.database_model import Images
 from src.person_info.person_info import PersonInfoManager, get_person_info_manager
-from src.chat.utils.utils import translate_timestamp_to_human_readable,assign_message_ids
+from src.chat.utils.utils import translate_timestamp_to_human_readable, assign_message_ids
 
 install(extra_lines=3)
 
 
+def replace_user_references_sync(
+    content: str,
+    platform: str,
+    name_resolver: Optional[Callable[[str, str], str]] = None,
+    replace_bot_name: bool = True,
+) -> str:
+    """
+    替换内容中的用户引用格式，包括回复<aaa:bbb>和@<aaa:bbb>格式
+
+    Args:
+        content: 要处理的内容字符串
+        platform: 平台标识
+        name_resolver: 名称解析函数，接收(platform, user_id)参数，返回用户名称
+                       如果为None，则使用默认的person_info_manager
+        replace_bot_name: 是否将机器人的user_id替换为"机器人昵称(你)"
+
+    Returns:
+        str: 处理后的内容字符串
+    """
+    if name_resolver is None:
+        person_info_manager = get_person_info_manager()
+
+        def default_resolver(platform: str, user_id: str) -> str:
+            # 检查是否是机器人自己
+            if replace_bot_name and user_id == global_config.bot.qq_account:
+                return f"{global_config.bot.nickname}(你)"
+            person_id = PersonInfoManager.get_person_id(platform, user_id)
+            return person_info_manager.get_value_sync(person_id, "person_name") or user_id  # type: ignore
+
+        name_resolver = default_resolver
+
+    # 处理回复<aaa:bbb>格式
+    reply_pattern = r"回复<([^:<>]+):([^:<>]+)>"
+    match = re.search(reply_pattern, content)
+    if match:
+        aaa = match[1]
+        bbb = match[2]
+        try:
+            # 检查是否是机器人自己
+            if replace_bot_name and bbb == global_config.bot.qq_account:
+                reply_person_name = f"{global_config.bot.nickname}(你)"
+            else:
+                reply_person_name = name_resolver(platform, bbb) or aaa
+            content = re.sub(reply_pattern, f"回复 {reply_person_name}", content, count=1)
+        except Exception:
+            # 如果解析失败，使用原始昵称
+            content = re.sub(reply_pattern, f"回复 {aaa}", content, count=1)
+
+    # 处理@<aaa:bbb>格式
+    at_pattern = r"@<([^:<>]+):([^:<>]+)>"
+    at_matches = list(re.finditer(at_pattern, content))
+    if at_matches:
+        new_content = ""
+        last_end = 0
+        for m in at_matches:
+            new_content += content[last_end : m.start()]
+            aaa = m.group(1)
+            bbb = m.group(2)
+            try:
+                # 检查是否是机器人自己
+                if replace_bot_name and bbb == global_config.bot.qq_account:
+                    at_person_name = f"{global_config.bot.nickname}(你)"
+                else:
+                    at_person_name = name_resolver(platform, bbb) or aaa
+                new_content += f"@{at_person_name}"
+            except Exception:
+                # 如果解析失败，使用原始昵称
+                new_content += f"@{aaa}"
+            last_end = m.end()
+        new_content += content[last_end:]
+        content = new_content
+
+    return content
+
+
+async def replace_user_references_async(
+    content: str,
+    platform: str,
+    name_resolver: Optional[Callable[[str, str], Any]] = None,
+    replace_bot_name: bool = True,
+) -> str:
+    """
+    替换内容中的用户引用格式，包括回复<aaa:bbb>和@<aaa:bbb>格式
+
+    Args:
+        content: 要处理的内容字符串
+        platform: 平台标识
+        name_resolver: 名称解析函数，接收(platform, user_id)参数，返回用户名称
+                       如果为None，则使用默认的person_info_manager
+        replace_bot_name: 是否将机器人的user_id替换为"机器人昵称(你)"
+
+    Returns:
+        str: 处理后的内容字符串
+    """
+    if name_resolver is None:
+        person_info_manager = get_person_info_manager()
+
+        async def default_resolver(platform: str, user_id: str) -> str:
+            # 检查是否是机器人自己
+            if replace_bot_name and user_id == global_config.bot.qq_account:
+                return f"{global_config.bot.nickname}(你)"
+            person_id = PersonInfoManager.get_person_id(platform, user_id)
+            return await person_info_manager.get_value(person_id, "person_name") or user_id  # type: ignore
+
+        name_resolver = default_resolver
+
+    # 处理回复<aaa:bbb>格式
+    reply_pattern = r"回复<([^:<>]+):([^:<>]+)>"
+    match = re.search(reply_pattern, content)
+    if match:
+        aaa = match.group(1)
+        bbb = match.group(2)
+        try:
+            # 检查是否是机器人自己
+            if replace_bot_name and bbb == global_config.bot.qq_account:
+                reply_person_name = f"{global_config.bot.nickname}(你)"
+            else:
+                reply_person_name = await name_resolver(platform, bbb) or aaa
+            content = re.sub(reply_pattern, f"回复 {reply_person_name}", content, count=1)
+        except Exception:
+            # 如果解析失败，使用原始昵称
+            content = re.sub(reply_pattern, f"回复 {aaa}", content, count=1)
+
+    # 处理@<aaa:bbb>格式
+    at_pattern = r"@<([^:<>]+):([^:<>]+)>"
+    at_matches = list(re.finditer(at_pattern, content))
+    if at_matches:
+        new_content = ""
+        last_end = 0
+        for m in at_matches:
+            new_content += content[last_end : m.start()]
+            aaa = m.group(1)
+            bbb = m.group(2)
+            try:
+                # 检查是否是机器人自己
+                if replace_bot_name and bbb == global_config.bot.qq_account:
+                    at_person_name = f"{global_config.bot.nickname}(你)"
+                else:
+                    at_person_name = await name_resolver(platform, bbb) or aaa
+                new_content += f"@{at_person_name}"
+            except Exception:
+                # 如果解析失败，使用原始昵称
+                new_content += f"@{aaa}"
+            last_end = m.end()
+        new_content += content[last_end:]
+        content = new_content
+
+    return content
+
+
 def get_raw_msg_by_timestamp(
     timestamp_start: float, timestamp_end: float, limit: int = 0, limit_mode: str = "latest"
 ) -> List[Dict[str, Any]]:
@@ -374,33 +524,8 @@ def _build_readable_messages_internal(
             else:
                 person_name = "某人"
 
-        # 检查是否有 回复<aaa:bbb> 字段
-        reply_pattern = r"回复<([^:<>]+):([^:<>]+)>"
-        match = re.search(reply_pattern, content)
-        if match:
-            aaa: str = match[1]
-            bbb: str = match[2]
-            reply_person_id = PersonInfoManager.get_person_id(platform, bbb)
-            reply_person_name = person_info_manager.get_value_sync(reply_person_id, "person_name") or aaa
-            # 在内容前加上回复信息
-            content = re.sub(reply_pattern, lambda m, name=reply_person_name: f"回复 {name}", content, count=1)
-
-        # 检查是否有 @<aaa:bbb> 字段 @<{member_info.get('nickname')}:{member_info.get('user_id')}>
-        at_pattern = r"@<([^:<>]+):([^:<>]+)>"
-        at_matches = list(re.finditer(at_pattern, content))
-        if at_matches:
-            new_content = ""
-            last_end = 0
-            for m in at_matches:
-                new_content += content[last_end : m.start()]
-                aaa = m.group(1)
-                bbb = m.group(2)
-                at_person_id = PersonInfoManager.get_person_id(platform, bbb)
-                at_person_name = person_info_manager.get_value_sync(at_person_id, "person_name") or aaa
-                new_content += f"@{at_person_name}"
-                last_end = m.end()
-            new_content += content[last_end:]
-            content = new_content
+        # 使用独立函数处理用户引用格式
+        content = replace_user_references_sync(content, platform, replace_bot_name=replace_bot_name)
 
         target_str = "这是QQ的一个功能，用于提及某人，但没那么明显"
         if target_str in content and random.random() < 0.6:
@@ -654,6 +779,7 @@ async def build_readable_messages_with_list(
 
     return formatted_string, details_list
 
+
 def build_readable_messages_with_id(
     messages: List[Dict[str, Any]],
     replace_bot_name: bool = True,
@@ -669,9 +795,9 @@ def build_readable_messages_with_id(
     允许通过参数控制格式化行为。
     """
     message_id_list = assign_message_ids(messages)
-    
+
     formatted_string = build_readable_messages(
-        messages = messages,
+        messages=messages,
         replace_bot_name=replace_bot_name,
         merge_messages=merge_messages,
         timestamp_mode=timestamp_mode,
@@ -682,10 +808,7 @@ def build_readable_messages_with_id(
         message_id_list=message_id_list,
     )
 
-    
-    
-    
-    return formatted_string , message_id_list
+    return formatted_string, message_id_list
 
 
 def build_readable_messages(
@@ -770,7 +893,13 @@ def build_readable_messages(
     if read_mark <= 0:
         # 没有有效的 read_mark，直接格式化所有消息
         formatted_string, _, pic_id_mapping, _ = _build_readable_messages_internal(
-            copy_messages, replace_bot_name, merge_messages, timestamp_mode, truncate, show_pic=show_pic, message_id_list=message_id_list
+            copy_messages,
+            replace_bot_name,
+            merge_messages,
+            timestamp_mode,
+            truncate,
+            show_pic=show_pic,
+            message_id_list=message_id_list,
         )
 
         # 生成图片映射信息并添加到最前面
@@ -893,7 +1022,7 @@ async def build_anonymous_messages(messages: List[Dict[str, Any]]) -> str:
 
     for msg in messages:
         try:
-            platform = msg.get("chat_info_platform")
+            platform: str = msg.get("chat_info_platform")  # type: ignore
             user_id = msg.get("user_id")
             _timestamp = msg.get("time")
             content: str = ""
@@ -916,38 +1045,14 @@ async def build_anonymous_messages(messages: List[Dict[str, Any]]) -> str:
             anon_name = get_anon_name(platform, user_id)
             # print(f"anon_name:{anon_name}")
 
-            # 处理 回复<aaa:bbb>
-            reply_pattern = r"回复<([^:<>]+):([^:<>]+)>"
-            match = re.search(reply_pattern, content)
-            if match:
-                # print(f"发现回复match:{match}")
-                bbb = match.group(2)
+            # 使用独立函数处理用户引用格式，传入自定义的匿名名称解析器
+            def anon_name_resolver(platform: str, user_id: str) -> str:
                 try:
-                    anon_reply = get_anon_name(platform, bbb)
-                    # print(f"anon_reply:{anon_reply}")
+                    return get_anon_name(platform, user_id)
                 except Exception:
-                    anon_reply = "?"
-                content = re.sub(reply_pattern, f"回复 {anon_reply}", content, count=1)
+                    return "?"
 
-            # 处理 @<aaa:bbb>，无嵌套def
-            at_pattern = r"@<([^:<>]+):([^:<>]+)>"
-            at_matches = list(re.finditer(at_pattern, content))
-            if at_matches:
-                # print(f"发现@match:{at_matches}")
-                new_content = ""
-                last_end = 0
-                for m in at_matches:
-                    new_content += content[last_end : m.start()]
-                    bbb = m.group(2)
-                    try:
-                        anon_at = get_anon_name(platform, bbb)
-                        # print(f"anon_at:{anon_at}")
-                    except Exception:
-                        anon_at = "?"
-                    new_content += f"@{anon_at}"
-                    last_end = m.end()
-                new_content += content[last_end:]
-                content = new_content
+            content = replace_user_references_sync(content, platform, anon_name_resolver, replace_bot_name=False)
 
             header = f"{anon_name}说 "
             output_lines.append(header)

src/chat/utils/utils_image.py

@@ -37,7 +37,7 @@ class ImageManager:
             self._ensure_image_dir()
 
             self._initialized = True
-            self._llm = LLMRequest(model=global_config.model.vlm, temperature=0.4, max_tokens=300, request_type="image")
+            self.vlm = LLMRequest(model=global_config.model.vlm, temperature=0.4, max_tokens=300, request_type="image")
 
             try:
                 db.connect(reuse_if_open=True)
@@ -94,7 +94,7 @@ class ImageManager:
             logger.error(f"保存描述到数据库失败 (Peewee): {str(e)}")
 
     async def get_emoji_description(self, image_base64: str) -> str:
-        """获取表情包描述，使用二步走识别并带缓存优化"""
+        """获取表情包描述，优先使用Emoji表中的缓存数据"""
         try:
             # 计算图片哈希
             # 确保base64字符串只包含ASCII字符
@@ -104,9 +104,21 @@ class ImageManager:
             image_hash = hashlib.md5(image_bytes).hexdigest()
             image_format = Image.open(io.BytesIO(image_bytes)).format.lower()  # type: ignore
 
-            # 查询缓存的描述
+            # 优先使用EmojiManager查询已注册表情包的描述
+            try:
+                from src.chat.emoji_system.emoji_manager import get_emoji_manager
+                emoji_manager = get_emoji_manager()
+                cached_emoji_description = await emoji_manager.get_emoji_description_by_hash(image_hash)
+                if cached_emoji_description:
+                    logger.info(f"[缓存命中] 使用已注册表情包描述: {cached_emoji_description[:50]}...")
+                    return cached_emoji_description
+            except Exception as e:
+                logger.debug(f"查询EmojiManager时出错: {e}")
+
+            # 查询ImageDescriptions表的缓存描述
             cached_description = self._get_description_from_db(image_hash, "emoji")
             if cached_description:
+                logger.info(f"[缓存命中] 使用ImageDescriptions表中的描述: {cached_description[:50]}...")
                 return f"[表情包：{cached_description}]"
 
             # === 二步走识别流程 ===
@@ -118,10 +130,10 @@ class ImageManager:
                     logger.warning("GIF转换失败，无法获取描述")
                     return "[表情包(GIF处理失败)]"
                 vlm_prompt = "这是一个动态图表情包，每一张图代表了动态图的某一帧，黑色背景代表透明，描述一下表情包表达的情感和内容，描述细节，从互联网梗,meme的角度去分析"
-                detailed_description, _ = await self._llm.generate_response_for_image(vlm_prompt, image_base64_processed, "jpg")
+                detailed_description, _ = await self.vlm.generate_response_for_image(vlm_prompt, image_base64_processed, "jpg")
             else:
                 vlm_prompt = "这是一个表情包，请详细描述一下表情包所表达的情感和内容，描述细节，从互联网梗,meme的角度去分析"
-                detailed_description, _ = await self._llm.generate_response_for_image(vlm_prompt, image_base64, image_format)
+                detailed_description, _ = await self.vlm.generate_response_for_image(vlm_prompt, image_base64, image_format)
 
             if detailed_description is None:
                 logger.warning("VLM未能生成表情包详细描述")
@@ -158,7 +170,7 @@ class ImageManager:
             if len(emotions) > 1 and emotions[1] != emotions[0]:
                 final_emotion = f"{emotions[0]}，{emotions[1]}"
 
-            logger.info(f"[二步走识别] 详细描述: {detailed_description[:50]}... -> 情感标签: {final_emotion}")
+            logger.info(f"[emoji识别] 详细描述: {detailed_description[:50]}... -> 情感标签: {final_emotion}")
 
             # 再次检查缓存，防止并发写入时重复生成
             cached_description = self._get_description_from_db(image_hash, "emoji")
@@ -201,13 +213,13 @@ class ImageManager:
             self._save_description_to_db(image_hash, final_emotion, "emoji")
 
             return f"[表情包：{final_emotion}]"
-            
+
         except Exception as e:
             logger.error(f"获取表情包描述失败: {str(e)}")
-            return "[表情包]"
+            return "[表情包(处理失败)]"
 
     async def get_image_description(self, image_base64: str) -> str:
-        """获取普通图片描述，带查重和保存功能"""
+        """获取普通图片描述，优先使用Images表中的缓存数据"""
         try:
             # 计算图片哈希
             if isinstance(image_base64, str):
@@ -215,7 +227,7 @@ class ImageManager:
             image_bytes = base64.b64decode(image_base64)
             image_hash = hashlib.md5(image_bytes).hexdigest()
 
-            # 检查图片是否已存在
+            # 优先检查Images表中是否已有完整的描述
             existing_image = Images.get_or_none(Images.emoji_hash == image_hash)
             if existing_image:
                 # 更新计数
@@ -227,18 +239,20 @@ class ImageManager:
 
                 # 如果已有描述，直接返回
                 if existing_image.description:
+                    logger.debug(f"[缓存命中] 使用Images表中的图片描述: {existing_image.description[:50]}...")
                     return f"[图片：{existing_image.description}]"
 
-            # 查询缓存的描述
+            # 查询ImageDescriptions表的缓存描述
             cached_description = self._get_description_from_db(image_hash, "image")
             if cached_description:
-                logger.debug(f"图片描述缓存中 {cached_description}")
+                logger.debug(f"[缓存命中] 使用ImageDescriptions表中的描述: {cached_description[:50]}...")
                 return f"[图片：{cached_description}]"
 
             # 调用AI获取描述
             image_format = Image.open(io.BytesIO(image_bytes)).format.lower()  # type: ignore
             prompt = global_config.custom_prompt.image_prompt
-            description, _ = await self._llm.generate_response_for_image(prompt, image_base64, image_format)
+            logger.info(f"[VLM调用] 为图片生成新描述 (Hash: {image_hash[:8]}...)")
+            description, _ = await self.vlm.generate_response_for_image(prompt, image_base64, image_format)
 
             if description is None:
                 logger.warning("AI未能生成图片描述")
@@ -266,6 +280,7 @@ class ImageManager:
                     if not hasattr(existing_image, "vlm_processed") or existing_image.vlm_processed is None:
                         existing_image.vlm_processed = True
                     existing_image.save()
+                    logger.debug(f"[数据库] 更新已有图片记录: {image_hash[:8]}...")
                 else:
                     Images.create(
                         image_id=str(uuid.uuid4()),
@@ -277,16 +292,18 @@ class ImageManager:
                         vlm_processed=True,
                         count=1,
                     )
+                    logger.debug(f"[数据库] 创建新图片记录: {image_hash[:8]}...")
             except Exception as e:
                 logger.error(f"保存图片文件或元数据失败: {str(e)}")
 
-            # 保存描述到ImageDescriptions表
+            # 保存描述到ImageDescriptions表作为备用缓存
             self._save_description_to_db(image_hash, description, "image")
 
+            logger.info(f"[VLM完成] 图片描述生成: {description[:50]}...")
             return f"[图片：{description}]"
         except Exception as e:
             logger.error(f"获取图片描述失败: {str(e)}")
-            return "[图片]"
+            return "[图片(处理失败)]"
 
     @staticmethod
     def transform_gif(gif_base64: str, similarity_threshold: float = 1000.0, max_frames: int = 15) -> Optional[str]:
@@ -502,12 +519,28 @@ class ImageManager:
             image_bytes = base64.b64decode(image_base64)
             image_hash = hashlib.md5(image_bytes).hexdigest()
 
-            # 先检查缓存的描述
+            # 获取当前图片记录
+            image = Images.get(Images.image_id == image_id)
+
+            # 优先检查是否已有其他相同哈希的图片记录包含描述
+            existing_with_description = Images.get_or_none(
+                (Images.emoji_hash == image_hash) & 
+                (Images.description.is_null(False)) & 
+                (Images.description != "")
+            )
+            if existing_with_description and existing_with_description.id != image.id:
+                logger.debug(f"[缓存复用] 从其他相同图片记录复用描述: {existing_with_description.description[:50]}...")
+                image.description = existing_with_description.description
+                image.vlm_processed = True
+                image.save()
+                # 同时保存到ImageDescriptions表作为备用缓存
+                self._save_description_to_db(image_hash, existing_with_description.description, "image")
+                return
+
+            # 检查ImageDescriptions表的缓存描述
             cached_description = self._get_description_from_db(image_hash, "image")
             if cached_description:
-                logger.debug(f"VLM处理时发现缓存描述: {cached_description}")
-                # 更新数据库
-                image = Images.get(Images.image_id == image_id)
+                logger.debug(f"[缓存复用] 从ImageDescriptions表复用描述: {cached_description[:50]}...")
                 image.description = cached_description
                 image.vlm_processed = True
                 image.save()
@@ -520,7 +553,8 @@ class ImageManager:
             prompt = global_config.custom_prompt.image_prompt
 
             # 获取VLM描述
-            description, _ = await self._llm.generate_response_for_image(prompt, image_base64, image_format)
+            logger.info(f"[VLM异步调用] 为图片生成描述 (ID: {image_id}, Hash: {image_hash[:8]}...)")
+            description, _ = await self.vlm.generate_response_for_image(prompt, image_base64, image_format)
 
             if description is None:
                 logger.warning("VLM未能生成图片描述")
@@ -533,14 +567,15 @@ class ImageManager:
                 description = cached_description
 
             # 更新数据库
-            image = Images.get(Images.image_id == image_id)
             image.description = description
             image.vlm_processed = True
             image.save()
 
-            # 保存描述到ImageDescriptions表
+            # 保存描述到ImageDescriptions表作为备用缓存
             self._save_description_to_db(image_hash, description, "image")
 
+            logger.info(f"[VLM异步完成] 图片描述生成: {description[:50]}...")
+
         except Exception as e:
             logger.error(f"VLM处理图片失败: {str(e)}")
 

src/chat/willing/mode_classical.py

@@ -28,7 +28,7 @@ class ClassicalWillingManager(BaseWillingManager):
         
         # print(f"[{chat_id}] 回复意愿: {current_willing}")
 
-        interested_rate = willing_info.interested_rate * global_config.normal_chat.response_interested_rate_amplifier
+        interested_rate = willing_info.interested_rate
         
         # print(f"[{chat_id}] 兴趣值: {interested_rate}")
 
@@ -36,20 +36,18 @@ class ClassicalWillingManager(BaseWillingManager):
             current_willing += interested_rate - 0.2
 
         if willing_info.is_mentioned_bot and global_config.chat.mentioned_bot_inevitable_reply and current_willing < 2:
-            current_willing += 1 if current_willing < 1.0 else 0.05
+            current_willing += 1 if current_willing < 1.0 else 0.2
 
         self.chat_reply_willing[chat_id] = min(current_willing, 1.0)
         
-        reply_probability = min(max((current_willing - 0.5), 0.01) * 2, 1)
+        reply_probability = min(max((current_willing - 0.5), 0.01) * 2, 1.5)
         
         # print(f"[{chat_id}] 回复概率: {reply_probability}")
         
         return reply_probability
 
     async def before_generate_reply_handle(self, message_id):
-        chat_id = self.ongoing_messages[message_id].chat_id
-        current_willing = self.chat_reply_willing.get(chat_id, 0)
-        self.chat_reply_willing[chat_id] = max(0.0, current_willing - 1.8)
+        pass
 
     async def after_generate_reply_handle(self, message_id):
         if message_id not in self.ongoing_messages:
@@ -58,7 +56,7 @@ class ClassicalWillingManager(BaseWillingManager):
         chat_id = self.ongoing_messages[message_id].chat_id
         current_willing = self.chat_reply_willing.get(chat_id, 0)
         if current_willing < 1:
-            self.chat_reply_willing[chat_id] = min(1.0, current_willing + 0.4)
+            self.chat_reply_willing[chat_id] = min(1.0, current_willing + 0.3)
 
     async def not_reply_handle(self, message_id):
         return await super().not_reply_handle(message_id)