fix：通过计时定位LLM异常延时，移除memory模型

src/chat/express/expression_selector.py

@@ -36,11 +36,7 @@ def init_prompt():
 请以JSON格式输出，只需要输出选中的情境编号：
 例如：
 {{
-    "selected_situations": [2, 3, 5, 7, 19, 22, 25, 38, 39, 45, 48 , 64]
-}}
-例如：
-{{
-    "selected_situations": [1, 4, 7, 9, 23, 38, 44]
+    "selected_situations": [2, 3, 5, 7, 19, 22, 25, 38, 39, 45, 48, 64]
 }}
 
 请严格按照JSON格式输出，不要包含其他内容：
@@ -214,7 +210,7 @@ class ExpressionSelector:
         """使用LLM选择适合的表达方式"""
 
         # 1. 获取35个随机表达方式（现在按权重抽取）
-        style_exprs, grammar_exprs = self.get_random_expressions(chat_id, 50, 0.5, 0.5)
+        style_exprs, grammar_exprs = self.get_random_expressions(chat_id, 30, 0.5, 0.5)
 
         # 2. 构建所有表达方式的索引和情境列表
         all_expressions = []
@@ -264,7 +260,10 @@ class ExpressionSelector:
 
         # 4. 调用LLM
         try:
+            
+            start_time = time.time()
             content, (reasoning_content, model_name, _) = await self.llm_model.generate_response_async(prompt=prompt)
+            logger.info(f"LLM请求时间: {model_name}  {time.time() - start_time} \n{prompt}")
 
             # logger.info(f"模型名称: {model_name}")
             # logger.info(f"LLM返回结果: {content}")

src/chat/memory_system/Hippocampus.py

@@ -190,7 +190,7 @@ class MemoryGraph:
 class Hippocampus:
     def __init__(self):
         self.memory_graph = MemoryGraph()
-        self.model_summary: LLMRequest = None  # type: ignore
+        self.model_small: LLMRequest = None  # type: ignore
         self.entorhinal_cortex: EntorhinalCortex = None  # type: ignore
         self.parahippocampal_gyrus: ParahippocampalGyrus = None  # type: ignore
 
@@ -200,7 +200,7 @@ class Hippocampus:
         self.parahippocampal_gyrus = ParahippocampalGyrus(self)
         # 从数据库加载记忆图
         self.entorhinal_cortex.sync_memory_from_db()
-        self.model_summary = LLMRequest(model_set=model_config.model_task_config.memory, request_type="memory.builder")
+        self.model_small = LLMRequest(model_set=model_config.model_task_config.utils_small, request_type="memory.small")
 
     def get_all_node_names(self) -> list:
         """获取记忆图中所有节点的名字列表"""
@@ -340,7 +340,7 @@ class Hippocampus:
         else:
             topic_num = 5  # 51+字符: 5个关键词 (其余长文本)
 
-        topics_response, _ = await self.model_summary.generate_response_async(self.find_topic_llm(text, topic_num))
+        topics_response, _ = await self.model_small.generate_response_async(self.find_topic_llm(text, topic_num))
 
         # 提取关键词
         keywords = re.findall(r"<([^>]+)>", topics_response)

src/chat/memory_system/instant_memory.py

@@ -38,7 +38,7 @@ class InstantMemory:
         self.chat_id = chat_id
         self.last_view_time = time.time()
         self.summary_model = LLMRequest(
-            model_set=model_config.model_task_config.memory,
+            model_set=model_config.model_task_config.utils,
             request_type="memory.summary",
         )