llm统计记录模型反应时间

src/chat/utils/statistic.py

@@ -90,6 +90,18 @@ STD_TIME_COST_BY_MODULE = "std_time_costs_by_module"
 ONLINE_TIME = "online_time"
 TOTAL_MSG_CNT = "total_messages"
 MSG_CNT_BY_CHAT = "messages_by_chat"
+TIME_COST_BY_TYPE = "time_costs_by_type"
+TIME_COST_BY_USER = "time_costs_by_user"
+TIME_COST_BY_MODEL = "time_costs_by_model"
+TIME_COST_BY_MODULE = "time_costs_by_module"
+AVG_TIME_COST_BY_TYPE = "avg_time_costs_by_type"
+AVG_TIME_COST_BY_USER = "avg_time_costs_by_user"
+AVG_TIME_COST_BY_MODEL = "avg_time_costs_by_model"
+AVG_TIME_COST_BY_MODULE = "avg_time_costs_by_module"
+STD_TIME_COST_BY_TYPE = "std_time_costs_by_type"
+STD_TIME_COST_BY_USER = "std_time_costs_by_user"
+STD_TIME_COST_BY_MODEL = "std_time_costs_by_model"
+STD_TIME_COST_BY_MODULE = "std_time_costs_by_module"
 
 
 class OnlineTimeRecordTask(AsyncTask):
@@ -418,7 +430,7 @@ class StatisticOutputTask(AsyncTask):
                         stats[period_key][COST_BY_USER][user_id] += cost
                         stats[period_key][COST_BY_MODEL][model_name] += cost
                         stats[period_key][COST_BY_MODULE][module_name] += cost
-
+                        
                         # 收集time_cost数据
                         time_cost = record.time_cost or 0.0
                         if time_cost > 0:  # 只记录有效的time_cost
@@ -427,8 +439,8 @@ class StatisticOutputTask(AsyncTask):
                             stats[period_key][TIME_COST_BY_MODEL][model_name].append(time_cost)
                             stats[period_key][TIME_COST_BY_MODULE][module_name].append(time_cost)
                     break
-
+                
-        # 计算平均耗时和标准差
+                        # 计算平均耗时和标准差
         for period_key in stats:
             for category in [REQ_CNT_BY_TYPE, REQ_CNT_BY_USER, REQ_CNT_BY_MODEL, REQ_CNT_BY_MODULE]:
                 time_cost_key = f"time_costs_by_{category.split('_')[-1]}"
@@ -452,7 +464,6 @@ class StatisticOutputTask(AsyncTask):
                     else:
                         stats[period_key][avg_key][item_name] = 0.0
                         stats[period_key][std_key][item_name] = 0.0
-
         return stats
 
     @staticmethod
@@ -687,7 +698,6 @@ class StatisticOutputTask(AsyncTask):
         data_fmt = "{:<32}  {:>10}  {:>12}  {:>12}  {:>12}  {:>9.4f}¥  {:>10}  {:>10}"
 
         output = [
-            "按模型分类统计:",
             " 模型名称                          调用次数    输入Token     输出Token     Token总量     累计花费    平均耗时(秒)  标准差(秒)",
         ]
         for model_name, count in sorted(stats[REQ_CNT_BY_MODEL].items()):
@@ -843,7 +853,7 @@ class StatisticOutputTask(AsyncTask):
                 
                 <h2>按模型分类统计</h2>
                 <table>
-                    <thead><tr><th>模型名称</th><th>调用次数</th><th>输入Token</th><th>输出Token</th><th>Token总量</th><th>累计花费</th><th>平均耗时(秒)</th><th>标准差(秒)</th></tr></thead>
+                    <tr><th>模块名称</th><th>调用次数</th><th>输入Token</th><th>输出Token</th><th>Token总量</th><th>累计花费</th><th>平均耗时(秒)</th><th>标准差(秒)</th></tr>
                     <tbody>
                         {model_rows}
                     </tbody>

src/common/database/sqlalchemy_models.py

@@ -29,6 +29,8 @@ def get_string_field(max_length=255, **kwargs):
         return String(max_length, **kwargs)
     else:
         return Text(**kwargs)
+    
+
 
 class SessionProxy:
     """线程安全的Session代理类，自动管理session生命周期"""
@@ -155,11 +157,14 @@ class LLMUsage(Base):
 
     id = Column(Integer, primary_key=True, autoincrement=True)
     model_name = Column(get_string_field(100), nullable=False, index=True)
+    model_assign_name = Column(get_string_field(100), index=True)  # 添加索引
+    model_api_provider = Column(get_string_field(100), index=True)  # 添加索引
     user_id = Column(get_string_field(50), nullable=False, index=True)
     request_type = Column(get_string_field(50), nullable=False, index=True)
     endpoint = Column(Text, nullable=False)
     prompt_tokens = Column(Integer, nullable=False)
     completion_tokens = Column(Integer, nullable=False)
+    time_cost = Column(Float, nullable=True)
     total_tokens = Column(Integer, nullable=False)
     cost = Column(Float, nullable=False)
     status = Column(Text, nullable=False)
@@ -167,6 +172,9 @@ class LLMUsage(Base):
 
     __table_args__ = (
         Index('idx_llmusage_model_name', 'model_name'),
+        Index('idx_llmusage_model_assign_name', 'model_assign_name'),
+        Index('idx_llmusage_model_api_provider', 'model_api_provider'),
+        Index('idx_llmusage_time_cost', 'time_cost'),
         Index('idx_llmusage_user_id', 'user_id'),
         Index('idx_llmusage_request_type', 'request_type'),
         Index('idx_llmusage_timestamp', 'timestamp'),

src/llm_models/utils_model.py

@@ -148,6 +148,7 @@ class LLMRequest:
                 model_info=model_info,
                 model_usage=usage,
                 user_id="system",
+                time_cost=time.time() - start_time,
                 request_type=self.request_type,
                 endpoint="/chat/completions",
                 time_cost=time.time() - start_time,
@@ -242,6 +243,7 @@ class LLMRequest:
     ) -> Tuple[str, Tuple[str, str, Optional[List[ToolCall]]]]:
         """执行单次请求"""
         # 模型选择和请求准备
+        start_time = time.time()
         model_info, api_provider, client = self._select_model()
         processed_prompt = self._apply_content_obfuscation(prompt, api_provider)
         
@@ -295,6 +297,7 @@ class LLMRequest:
                     llm_usage_recorder.record_usage_to_database(
                         model_info=model_info,
                         model_usage=usage,
+                        time_cost=time.time() - start_time,
                         user_id="system",
                         request_type=self.request_type,
                         endpoint="/chat/completions",
@@ -350,6 +353,7 @@ class LLMRequest:
         if usage := response.usage:
             llm_usage_recorder.record_usage_to_database(
                 model_info=model_info,
+                time_cost=time.time() - start_time,
                 model_usage=usage,
                 user_id="system",
                 request_type=self.request_type,