'commit'

dsRag/Start.py

@@ -114,11 +114,11 @@ async def rag(request: fastapi.Request):
     query = data.get('query', '')
     query_tags = data.get('tags', [])
     # 调用es进行混合搜索
-    search_results = queryByEs(query, query_tags, logger)
+    search_results = EsSearchUtil.queryByEs(query, query_tags, logger)
     # 构建提示词
     context = "\n".join([
         f"结果{i + 1}: {res['tags']['full_content']}"
-        for i, res in enumerate(search_results['vector_results'] + search_results['text_results'])
+        for i, res in enumerate(search_results['text_results'])
     ])
     # 添加图片识别提示
     prompt = f"""

dsRag/Util/EsSearchUtil.py

@@ -201,47 +201,23 @@ class EsSearchUtil:
             )
             logger.info(f"2. 文本搜索结果数量: {len(text_results['hits']['hits'])}")
 
-            # 合并结果
-            logger.info("\n=== 最终搜索结果 ===")
-            logger.info(f"向量搜索结果: {len(vector_results['hits']['hits'])}条")
-            for i, hit in enumerate(vector_results['hits']['hits'], 1):
-                logger.info(f"   {i}. 文档ID: {hit['_id']}, 相似度分数: {hit['_score']:.2f}")
-                logger.info(f"      内容: {hit['_source']['user_input']}")
-
-            logger.info("文本精确搜索结果:")
-            for i, hit in enumerate(text_results['hits']['hits']):
-                logger.info(f"  {i + 1}. 文档ID: {hit['_id']}, 匹配分数: {hit['_score']:.2f}")
-                logger.info(f"      内容: {hit['_source']['user_input']}")
-
-            # 去重处理：去除vector_results和text_results中重复的user_input
-            vector_sources = [hit['_source'] for hit in vector_results['hits']['hits']]
-            text_sources = [hit['_source'] for hit in text_results['hits']['hits']]
-
-            # 构建去重后的结果
-            unique_text_sources = []
-            text_user_inputs = set()
-
-            # 先处理text_results，保留所有
-            for source in text_sources:
-                text_user_inputs.add(source['user_input'])
-                unique_text_sources.append(source)
-
-            # 处理vector_results，只保留不在text_results中的
-            unique_vector_sources = []
-            for source in vector_sources:
-                if source['user_input'] not in text_user_inputs:
-                    unique_vector_sources.append(source)
-
-            # 计算优化掉的记录数量和节约的tokens
-            removed_count = len(vector_sources) - len(unique_vector_sources)
-            saved_tokens = sum(len(source['user_input']) for source in vector_sources
-                               if source['user_input'] in text_user_inputs)
-
-            logger.info(f"优化掉 {removed_count} 条重复记录，节约约 {saved_tokens} tokens")
+            # 合并vector和text结果
+            all_sources = [hit['_source'] for hit in vector_results['hits']['hits']] + \
+                          [hit['_source'] for hit in text_results['hits']['hits']]
+            
+            # 去重处理
+            unique_sources = []
+            seen_user_inputs = set()
+            
+            for source in all_sources:
+                if source['user_input'] not in seen_user_inputs:
+                    seen_user_inputs.add(source['user_input'])
+                    unique_sources.append(source)
+            
+            logger.info(f"合并后去重结果数量: {len(unique_sources)}条")
 
             search_results = {
-                "vector_results": unique_vector_sources,
-                "text_results": unique_text_sources
+                "text_results": unique_sources
             }
             return search_results
         finally: