去除混排，使用重排之后的前三个文档之和作为返回文档

src/server/get_similarity.py

@@ -81,44 +81,46 @@ class GetSimilarityWithExt:
         return result
 
     def get_rerank_with_doc(self, reranker: BgeRerank,split_docs_list:list):
-        top_k = self.get_doc_nums(len(split_docs_list))
+        # top_k = self.get_doc_nums(len(split_docs_list))
         question = '\n'.join(self.question)
         print(question)
-        rerank_docs1_hash = []
-        rerank_docs2_hash = []
-        m = {}
+        # rerank_docs1_hash = []
+        # rerank_docs2_hash = []
+        # m = {}
         result = []
         for split_doc in split_docs_list:
             start = time.time()
             rerank_docs1 = reranker.compress_documents(split_doc, question)
+            result.extend(rerank_docs1[:3])
             end = time.time()
             print('重排1 time: %s Seconds' % (end - start))
-            for doc in rerank_docs1:
-                m[hash(doc.page_content)] = doc
-                rerank_docs1_hash.append(hash(doc.page_content))
-            result.append((60, rerank_docs1_hash))
+            # for doc in rerank_docs1:
+            #     m[hash(doc.page_content)] = doc
+            #     rerank_docs1_hash.append(hash(doc.page_content))
+            # result.append((60, rerank_docs1_hash))
         start = time.time()
         rerank_docs2 = reranker.compress_documents(self.similarity_docs, question)
+        result.extend(rerank_docs2[:3])
         end = time.time()
         print('重排2 time: %s Seconds' % (end - start))
-        for doc in rerank_docs2:
-            m[hash(doc.page_content)] = doc
-            rerank_docs2_hash.append(hash(doc.page_content))
-
-        result.append((55,rerank_docs2_hash))
-        print(len(rerank_docs1_hash))
-        print(len(rerank_docs2_hash))
-        start = time.time()
-        rrf_doc = reciprocal_rank_fusion(result)
-        end = time.time()
-        print('混排 time: %s Seconds' % (end - start))
-        print("混排文档数量：", len(rrf_doc))
-        d_list = []
-        for key in rrf_doc:
-            d_list.append(m[key])
-        print("返回文档数量：",top_k)
-        self.rerank_docs = d_list[:top_k]
-        return self.join_document(d_list[:top_k])
+        # for doc in rerank_docs2:
+        #     m[hash(doc.page_content)] = doc
+        #     rerank_docs2_hash.append(hash(doc.page_content))
+        #
+        # result.append((55,rerank_docs2_hash))
+        # print(len(rerank_docs1_hash))
+        # print(len(rerank_docs2_hash))
+        # start = time.time()
+        # rrf_doc = reciprocal_rank_fusion(result)
+        # end = time.time()
+        # print('混排 time: %s Seconds' % (end - start))
+        # print("混排文档数量：", len(rrf_doc))
+        # d_list = []
+        # for key in rrf_doc:
+        #     d_list.append(m[key])
+        # print("返回文档数量：",top_k)
+        self.rerank_docs = result
+        return self.join_document(result)
 
     def get_similarity_doc(self):
         return self.similarity_doc_txt