删除多余代码

rag2_0/intent_recognition/IntentRecognition.py

@@ -11,6 +11,7 @@ import logging
 import os
 import asyncio
 from langchain.output_parsers import PydanticOutputParser
+from langchain_core.output_parsers import JsonOutputParser
 import json
 from typing import List, Tuple, Dict, Any, Optional
 import re
@@ -23,11 +24,6 @@ from .PromptTemplates import (classification_prompt, query_rewrite_prompt_pro,
                               slot_filling_prompt, step_back_prompt, 
                               follow_up_questions_prompt, hyde_prompt, multi_questions_prompt)
 
-from .Multi_PromptTemplates import (
-    intent_and_slot_prompt, output_example, 
-    generate_slot_mapping_doc
-)
-
 from .DataModels import (
     Classification, QueryRewrite, Term, TermList,
     SoftwareFunctionSlots, SoftwareTroubleShootingSlots, ProfessionalConsultingSlots, 
@@ -197,7 +193,7 @@ class AsyncIntentRecognizer:
         
         return filtered_tokens
     
-    async def _extract_keywords_with_llm_async(self, query: str, use_jieba: bool = False) -> List[Term]:
+    async def _extract_keywords_async(self, query: str, use_jieba: bool = False) -> List[Term]:
         """
         异步使用LLM从用户查询中提取专业关键词
         
@@ -289,7 +285,7 @@ class AsyncIntentRecognizer:
         # 步骤1: 使用LLM提取查询中的关键词
         try:
             llm_start_time = time.time()
-            extracted_terms = await self._extract_keywords_with_llm_async(query, use_jieba)
+            extracted_terms = await self._extract_keywords_async(query, use_jieba)
             for term in extracted_terms:
                 query_keys.append(term.name)
             llm_end_time = time.time()
@@ -328,7 +324,8 @@ class AsyncIntentRecognizer:
             
             # 合并所有结果
             for result in keyword_results:
-                matched_terms.extend(result)
+                if len(result) > 0:
+                    matched_terms.extend(result)
                 
             vector_end_time = time.time()
             vector_time = vector_end_time - vector_start_time
@@ -383,26 +380,6 @@ class AsyncIntentRecognizer:
         except Exception as e:
             raise RuntimeError(f"解析问题改写结果时出错: {e}") from e
     
-    def _judge_define_suffix(self, input_str: str) -> Tuple[bool, List[str]]:
-        """
-        判断输入字符串是否包含定义的后缀，并返回所有匹配到的后缀名列表
-        
-        Args:
-            input_str: 输入字符串
-            
-        Returns:
-            Tuple[bool, List[str]]: (是否包含定义的后缀, 匹配到的后缀名列表)
-        """
-        
-        # 构建正则表达式模式，匹配大小写不敏感且前面可能带有.
-        pattern = r'(?:\.?)(' + '|'.join(re.escape(field.get('name')) for field in self._suffix_keywords) + r')'
-        
-        # 使用 re.IGNORECASE 标志来忽略大小写，findall找到所有匹配
-        matches = re.finditer(pattern, input_str, re.IGNORECASE)
-        matched_suffixes = [match.group(1) for match in matches]
-        
-        return bool(matched_suffixes), matched_suffixes
-
     def _process_lock_related_query(self, query: str) -> str:
         """
         特殊处理锁相关咨询
@@ -786,7 +763,8 @@ class AsyncIntentRecognizer:
             # 解析JSON响应
             try:
                 wiki_names = []
-                json_response = json.loads(response.content)
+                json_parser = JsonOutputParser()
+                json_response = json_parser.parse(response.content)
                 for match in json_response:
                     wiki_names.append(match["content"])
                 logging.debug(f"软件文档匹配耗时: {end_time - start_time:.2f}秒")
@@ -838,97 +816,4 @@ class AsyncIntentRecognizer:
         except Exception as e:
             # 如果解析失败，返回原始查询作为唯一子问题
             logging.error(f"异步多角度问题生成失败: {e}", exc_info=True)
-            return MultiQuestions(original_query=query, sub_questions=[query])
-            
-    async def _process_intent_and_slot_async(self, user_input: str, conversation_context: str = "", 
-                              chat_history: List[Dict[str, str]] = None, 
-                              previous_slots: Dict[str, Any] = None) -> Dict[str, Any]:
-        """
-        异步使用统一提示词同时进行意图识别和槽位填充
-        
-        Args:
-            user_input: 当前用户输入
-            conversation_context: 会话背景信息
-            chat_history: 历史对话记录，格式为[{"role": "user", "content": "..."}, {"role": "assistant", "content": "..."}]
-            previous_slots: 历史槽位信息
-            
-        Returns:
-            包含意图分类和槽位填充结果的字典
-        """
-        # 初始化默认值
-        if chat_history is None:
-            chat_history = []
-        
-        if previous_slots is None:
-            previous_slots = {}
-            
-        # 生成槽位映射文档
-        slot_mapping_doc = generate_slot_mapping_doc()
-        
-        # 准备提示词
-        parser = PydanticOutputParser(pydantic_object=IntentAndSlotResult)
-        formatted_prompt = intent_and_slot_prompt.format(
-            conversation_context=conversation_context,
-            chat_history=json.dumps(chat_history, ensure_ascii=False),
-            previous_slots=json.dumps(previous_slots, ensure_ascii=False),
-            user_input=user_input,
-            slot_mapping_doc=slot_mapping_doc,
-            output_format=parser.get_format_instructions(),
-            classification_info=classification_info
-        )
-        
-        # 异步调用LLM
-        llm_start_time = time.time()
-        response = await self._llm.invoke_async(formatted_prompt + output_example, False)
-        llm_end_time = time.time()
-        llm_time = llm_end_time - llm_start_time
-
-        try:
-            # 解析LLM响应为JSON
-            response.content = response.content.strip()
-            clean_output = re.sub(r'<think>.*?</think>', '', response.content, flags=re.DOTALL)
-            result_json = parser.parse(clean_output)
-            classification = result_json.classification
-            slot_filling = result_json.slots
-            is_complete, missing_slots = slot_filling.check_required_slots()
-            expected_slot_model = self._get_slot_model(classification)
-
-            # 添加容错处理，发生概率较低，但仍需处理
-            if expected_slot_model is None:
-                # 添加容错处理，应对LLM返回错误分类信息,一级分类跟二级分类错乱
-                # 重新分类
-                classification = await self._classify_intent_async(user_input, conversation_context, chat_history, previous_slots)
-                fill_slots = await self._fill_slots_async(user_input, classification, conversation_context, chat_history, previous_slots)
-                
-                result = {
-                    "classification": classification.model_dump(),
-                    "slot_filling": fill_slots
-                }
-                logging.warning(f"异步重新分类与槽点填充")
-                return result
-            elif expected_slot_model.__name__ != type(slot_filling).__name__:
-                # 添加容错处理，应对LLM槽位与分类不匹配。重新填充槽位
-                slot_filling = await self._fill_slots_async(user_input, classification, conversation_context, chat_history, previous_slots)
-                result = {
-                    "classification": classification.model_dump(),
-                    "slot_filling": slot_filling
-                }
-                logging.warning(f"异步重新填充槽点")
-                return result
-            
-            logging.info(f"异步意图识别+槽位LLM调用耗时: {llm_time:.2f}秒")
-
-            # 构建最终结果
-            result = {
-                "classification": classification.model_dump(),
-                "slot_filling": {
-                    "is_complete": is_complete,
-                    "missing_slots": missing_slots,
-                    "filled_data": slot_filling.model_dump()
-                }
-            }
-            
-            return result
-            
-        except Exception as e:
-            raise RuntimeError(f"异步process_intent_and_slot error:{e}") from e
+            return MultiQuestions(original_query=query, sub_questions=[query])