优化DifyCompareTest类，添加DifyExporter实例以支持词条检索，更新DifyQueryRetrieval_api.py中的topk参数，增强DifyExporter类以从HTTP服务获取查询类型和点踩原因，简化构造函数，移除不必要的查询日志加载逻辑。

rag2_0/dify/DifyCompareTest.py

@@ -18,7 +18,8 @@ from langchain_core.output_parsers import JsonOutputParser
 sys.path.append(os.getcwd())
 from rag2_0.dify.dify_client import ChatClient
 from rag2_0.tool.ModelTool import OpenAiLLM
-
+from rag2_0.dify.dify_tool import PgSql, DifyTool
+from rag2_0.dify.export_new_dify import DifyExporter
 load_dotenv()
 # 创建日志目录
 log_dir = 'data/logs'
@@ -45,6 +46,7 @@ class DifyCompareTest:
         # 词条与工单同时检索
         self.both_wiki_worker_client = ChatClient(api_key=os.getenv("DIFY_APP_KEY"), base_url=os.getenv("DIFY_BSAE_URL"))
         self.llm = OpenAiLLM(base_url=os.getenv("OPENAI_API_BASE"), model=os.getenv("MODEL_NAME"))
+        self.exporter = DifyExporter()
 
     def llm_judge_answer(self, old_answer: str, now_answer: str):
         user_prompt = f"""
@@ -100,10 +102,11 @@ class DifyCompareTest:
                 answer = result.get('answer', "")
                 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=""
+                # 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', "")
@@ -117,6 +120,18 @@ class DifyCompareTest:
                     import time
                     time.sleep(10)  # 等待1秒后重试
 
+    def get_wiki_list_by_msgid(self,msg_id):
+        if msg_id is None or pd.isna(msg_id):
+            return ""
+        msg_debug_info = self.exporter.dify_tool.get_message_debug_info_by_id(msg_id)
+        if not msg_debug_info:
+            return ""
+        wiki_list = self.exporter.get_wiki_list(msg_debug_info)
+        if len(wiki_list) == 0:
+            return ""
+        else:
+            return "\n".join(list(set(wiki_list)))
+
     def process_single_row(self, index, row):
         """处理单行数据的方法"""
         try:
@@ -145,6 +160,7 @@ class DifyCompareTest:
             result_row["message_id"] = message_id
             result_row["回答"] = answer
             # result_row["词条与工单同时回答对比"] = judge_result
+            result_row["检索到的词条"] = self.get_wiki_list_by_msgid(message_id)
             logging.info(f"成功处理第 {index + 1} 行数据")
             return index, result_row
             
@@ -152,6 +168,7 @@ class DifyCompareTest:
             logging.error(f"处理第 {index + 1} 行数据时出错: {e}")
             result_row = row.copy()
             result_row["回答"] = ''
+            result_row["检索到的词条"] = ''
             result_row["message_id"] = ''
             return index, result_row
 
@@ -230,7 +247,7 @@ 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/第四轮问题-Part2.xlsx", "data/excel/第四轮问题-Part2-问答测试.xlsx")
+            ("data/excel/7.30数据导出.xlsx", "data/excel/7.30数据导出_问答测试.xlsx")
         ]
         
         for excel_path, save_path in excel_files: