Update RAG_Learning_Assistant_with_Streaming.py from CIV3283/CIV3283_admin

RAG_Learning_Assistant_with_Streaming.py

@@ -11,32 +11,27 @@ import re
 class RAGLearningAssistant:
     def __init__(self, api_key: str, model: str = "gpt-4.1-nano-2025-04-14", vector_db_path: str = ""):
         """
-        初始化RAG学习助手（适配学生Space）
+        初始化RAG学习助手
         
         Args:
             api_key: OpenAI API密钥（必需）
             model: 使用的模型名称
-            vector_db_path: 向量数据库所在目录路径（数据存储仓库的本地目录）
+            vector_db_path: 向量数据库路径
         """
         self.client = OpenAI(api_key=api_key)
-        
-        # 使用修改后的KnowledgeBaseVectorizer，指定vector_db_dir
         self.vectorizer = KnowledgeBaseVectorizer(
             api_key=api_key,
-            vector_db_dir=vector_db_path  # 传递数据存储仓库的本地目录
+            #data_path=os.path.join(vector_db_path, "knowledge_base.md")
+            data_path="knowledge_base.md"
         )
         
         # 预加载向量数据库到缓存
-        print("[RAGLearningAssistant] Preloading vector database...")
-        load_result = self.vectorizer.load_vector_database()
-        if load_result[0] is not None:
-            print(f"[RAGLearningAssistant] Vector database loaded successfully")
-        else:
-            print(f"[RAGLearningAssistant] Warning: Failed to load vector database")
+        print("预加载向量数据库...")
+        self.vectorizer.load_vector_database()
         
         # 模型配置
         self.model = model
-        self.temperature = 0.1
+        self.temperature = 0.2
         self.max_tokens = 2000
         
         # 系统提示词
@@ -51,6 +46,9 @@ You have access to a knowledge base of course materials. When answering question
 1. Stick to the provided context from the knowledge base.
 2. At the end of your response, provide students the 'title' & 'from' fields of the chunks that were used to answer the question. So that they can refer to the original source.
 3. If the knowledge base doesn't contain relevant information, say so. Students can go to the teaching team for further assistance.
+
+In the response, enclose full mathematical formulas with $$ for proper Markdown rendering. Do not enclose individual parameters or variables with $$.
+Bold key words if applicable.
 """
         
         # 查询重写的系统提示词 - 改进版本
@@ -121,7 +119,7 @@ Return the entities as a JSON array of strings. Only include the most important
             response = self.client.chat.completions.create(
                 model=self.model,
                 messages=messages,
-                temperature=0.1,  # 低温度确保一致性
+                temperature=0.3,  # 低温度确保一致性
                 max_tokens=2000
             )
             
@@ -175,22 +173,22 @@ Return the entities as a JSON array of strings. Only include the most important
             
             # 如果仍然没有获得有效结果，使用更简单的方法
             if not summary and self.conversation_history:
-                summary = "Continue previous discussion"
+                summary = "继续之前的讨论"
             
             if not rewritten or rewritten == query:
                 rewritten = query
             
-            print(f"[rewrite_query] Raw query: {query}")
-            print(f"[rewrite_query] Chat history summary: {summary}")
-            print(f"[rewrite_query] Rewritten query: {rewritten}")
+            print(f"Raw query: {query}")
+            print(f"Chat history summary: {summary}")
+            print(f"Rewrite query: {rewritten}")
             return summary, rewritten
             
         except Exception as e:
-            print(f"[rewrite_query] Query rewriting failed: {e}")
+            print(f"查询重写失败: {e}")
             # 生成简单的历史总结作为备用
             simple_summary = ""
             if self.conversation_history:
-                simple_summary = "Based on previous conversation content"
+                simple_summary = "基于之前的对话内容"
             return simple_summary, query  # 失败时返回简单总结和原始查询
     
     def extract_entities(self, original_query: str, summary: str, rewritten_query: str) -> List[str]:
@@ -231,8 +229,8 @@ Return the entities as a JSON array of strings. Only include the most important
             response = self.client.chat.completions.create(
                 model=self.model,
                 messages=messages,
-                temperature=self.temperature,
-                max_tokens=self.max_tokens
+                temperature=0.3,
+                max_tokens=200
             )
             
             content = response.choices[0].message.content.strip()
@@ -246,18 +244,18 @@ Return the entities as a JSON array of strings. Only include the most important
                 else:
                     entities = json.loads(content)
                 
-                print(f"[extract_entities] Extracted entities: {entities}")
+                print(f"Extracted entities: {entities}")
                 return entities
                 
             except json.JSONDecodeError:
                 # 如果JSON解析失败，尝试简单的文本处理
-                print(f"[extract_entities] JSON parsing failed, using backup method")
+                print(f"JSON解析失败，使用备用方法")
                 # 查找引号中的内容
                 entities = re.findall(r'"([^"]+)"', content)
                 return entities if entities else self.simple_entity_extraction(combined_text)
                 
         except Exception as e:
-            print(f"[extract_entities] Entity extraction failed: {e}")
+            print(f"实体提取失败: {e}")
             # 失败时使用简单的关键词提取
             return self.simple_entity_extraction(combined_text)
     
@@ -286,9 +284,9 @@ Return the entities as a JSON array of strings. Only include the most important
         entities.extend(special_terms)
         
         # 去重并返回
-        return list(set(entities))[:10]  # 最多返回5个实体
+        return list(set(entities))[:5]  # 最多返回5个实体
     
-    def enhanced_search(self, query: str, top_k: int = 5) -> Tuple[str, str, List[str], List[Tuple[Dict, float, Dict]]]:
+    def enhanced_search(self, query: str, top_k: int = 3) -> Tuple[str, str, List[str], List[Tuple[Dict, float, Dict]]]:
         """
         增强搜索：重写查询 -> 提取实体 -> 基于实体搜索（优化版本）
         
@@ -311,12 +309,12 @@ Return the entities as a JSON array of strings. Only include the most important
             search_results = self.vectorizer.search_with_entities_optimized(entities, top_k)
         else:
             # 如果没有提取到实体，使用重写后的查询进行搜索
-            print("[enhanced_search] No entities extracted, using full query search")
+            print("未提取到实体，使用完整查询搜索")
             search_results = self.vectorizer.search_similar(
                 rewritten_query, 
                 top_k=top_k,
-                title_weight=0.2,
-                content_weight=0.5,
+                title_weight=0.4,
+                content_weight=0.3,
                 full_weight=0.3
             )
         
@@ -340,6 +338,7 @@ Return the entities as a JSON array of strings. Only include the most important
             entry, combined_score, details = result
             # 只显示 title, source, content，不显示 id
             context_parts.append(
+                #f"[Source {i}]\n"
                 f"Title: {entry['title']}\n"
                 f"From: {entry['source']}\n"
                 f"Content: {entry['content']}\n"
@@ -386,7 +385,7 @@ Return the entities as a JSON array of strings. Only include the most important
             响应文本片段
         """
         # 1. 增强搜索（现在使用优化版本）
-        print("[generate_response_stream] Processing query...")
+        print("正在处理查询...")
         summary, rewritten_query, entities, search_results = self.enhanced_search(query)
         
         # 2. 格式化上下文
@@ -414,7 +413,7 @@ Return the entities as a JSON array of strings. Only include the most important
             if summary:
                 search_info += f"- Summary of history: {summary}\n"
             if rewritten_query != query:
-                search_info += f"- Rewritten query: {rewritten_query}\n"
+                search_info += f"- Rewrite query: {rewritten_query}\n"
             search_info += f"- Key entities: {', '.join(entities) if entities else 'No specific entities extracted'}\n"
             
             if search_results:
@@ -430,7 +429,7 @@ Return the entities as a JSON array of strings. Only include the most important
             # 添加缓存信息（调试用）
             cache_info = self.vectorizer.get_cache_info()
             if cache_info['is_cached']:
-                search_info += f"💡 Vector database cached with {cache_info['cache_size']} entries\n\n"
+                search_info += f"The vector db has been cached, containing {cache_info['cache_size']} entries\n\n"
 
             yield search_info
             
@@ -446,7 +445,7 @@ Return the entities as a JSON array of strings. Only include the most important
             self.conversation_history.append({"role": "assistant", "content": full_response})
             
         except Exception as e:
-            yield f"\n\nError: Problem occurred while generating response - {str(e)}"
+            yield f"\n\n错误：生成响应时出现问题 - {str(e)}"
     
     def generate_response(self, query: str) -> str:
         """
@@ -466,18 +465,18 @@ Return the entities as a JSON array of strings. Only include the most important
     def clear_history(self):
         """清除对话历史"""
         self.conversation_history = []
-        print("[clear_history] Conversation history cleared")
+        print("对话历史已清除")
     
     def clear_vector_cache(self):
         """清除向量数据库缓存"""
         self.vectorizer.clear_cache()
-        print("[clear_vector_cache] Vector database cache cleared")
+        print("向量数据库缓存已清除")
     
     def reload_vector_database(self):
         """重新加载向量数据库"""
-        print("[reload_vector_database] Reloading vector database...")
+        print("重新加载向量数据库...")
         self.vectorizer.load_vector_database(force_reload=True)
-        print("[reload_vector_database] Vector database reload completed")
+        print("向量数据库重新加载完成")
     
     def get_system_status(self) -> Dict:
         """
@@ -515,4 +514,5 @@ Return the entities as a JSON array of strings. Only include the most important
         with open(filepath, 'w', encoding='utf-8') as f:
             json.dump(conversation_data, f, ensure_ascii=False, indent=2)
         
-        print(f"[save_conversation] Conversation saved to: {filepath}")
+        print(f"对话已保存到: {filepath}")
+