在heli_db_to_excel.py中添加SEQUENCE_ID字段到数据处理结果，并新增根据会话ID查询SEQUENCE_ID的方法；在DifyCompareTest.py中优化回答判断逻辑，增加对空回答的处理，调整文件处理的最大并发数。

rag2_0/demo/heli_db_to_excel.py

@@ -277,6 +277,7 @@ class DataProcessor:
                 "消息发送者": record["message_sender"],
                 "发送者昵称": record["sender_nickname"],
                 "创建时间": record["CREATE_TIME"],
+                "SEQUENCE_ID": record["SEQUENCE_ID"],
             })
         
         return result
@@ -337,11 +338,21 @@ class MariaDBClient:
         df, _ = self.execute_query(sql, (start_date, end_date))
         return df
     
+    def get_sequence_id_by_session_id(self, session_id: str) -> Optional[pd.DataFrame]:
+        """根据会话ID查询消息详情"""
+        sql = """
+        SELECT SEQUENCE_ID
+        FROM crm_hlyj.crm_hlyj_dmri
+        WHERE SESSION_ID = %s
+        """
+        df, _ = self.execute_query(sql, (session_id,))
+        return df
+
     def query_messages_by_session_id(self, session_id: str) -> Optional[pd.DataFrame]:
         """根据会话ID查询消息详情"""
         sql = """
         SELECT CREATE_TIME, CUS_NICK_NAME, MODE, MSG_TYPE, AGENT_NAME, CONTENT, 
-               SESSION_ID, ACCOUNT, SYSTEM_MODE_MESSAGE_TYPE
+               SESSION_ID, ACCOUNT, SYSTEM_MODE_MESSAGE_TYPE, SEQUENCE_ID
         FROM crm_hlyj.crm_hlyj_dmri
         WHERE SESSION_ID = %s
         ORDER BY CREATE_TIME

rag2_0/dify/DifyCompareTest.py

@@ -69,7 +69,8 @@ class DifyCompareTest:
 
         max_retries = 3
         retry_count = 0
-        
+        if len(old_answer) == 0 or len(now_answer) == 0:
+            return "回答基本不相同"
         while retry_count < max_retries:
             try:
                 response = self.llm.invoke(user_prompt=user_prompt, need_retry=False, response_format={"type": "json_object"})    
@@ -100,15 +101,14 @@ class DifyCompareTest:
                 )
                 result = response.json()
                 answer = result.get('answer', "")
+                answer = answer.split("----------------------------------------")[0].strip()
                 if len(answer) == 0:
                     raise Exception(f"回答为空: {result}")
                 if old_answer:
                     judge_result = self.llm_judge_answer(old_answer=old_answer, now_answer=answer)
                 else:
                     judge_result=""
-                judge_result=""
                 # 只取回答的前半部分
-                answer = answer.split("----------------------------------------")[0].strip()
                 message_id = result.get('message_id', "")
                 return answer, judge_result, message_id
             except Exception as e:
@@ -247,14 +247,14 @@ if __name__ == "__main__":
         
         # 处理第一个文件
         excel_files = [
-            # ("data/excel/5月.xlsx", "data/excel/5月问答对比.xlsx"),
+            ("data/excel/300专业提问.xlsx", "data/excel/300专业提问_问答测试.xlsx"),
-            ("data/excel/第四轮点踩问题.xlsx", "data/excel/第四轮点踩问题_问答测试.xlsx")
+            # ("data/excel/有知识的.xlsx", "data/excel/有知识的_问答测试.xlsx")
         ]
         
         for excel_path, save_path in excel_files:
             logging.info(f"开始处理文件: {excel_path}")
             try:
-                dify_compare_test.run(excel_path=excel_path, save_path=save_path, max_workers=3)
+                dify_compare_test.run(excel_path=excel_path, save_path=save_path, max_workers=10)
                 logging.info(f"文件处理完成: {excel_path}")
             except Exception as e:
                 logging.error(f"处理文件 {excel_path} 时出错: {e}")