更新API密钥，优化意图识别示例，调整文档相关性判断逻辑，增强Excel数据验证功能，改进日志记录，

rag2_0/demo/intent_recognition_example.py

@@ -16,7 +16,7 @@ from tqdm import tqdm
 import time
 import sys
 import argparse
-from typing import List, Dict, Any, Optional
+from typing import List, Dict, Any
 from langchain.output_parsers import PydanticOutputParser
 from pydantic import BaseModel, Field
 sys.path.append(os.getcwd())
@@ -28,7 +28,7 @@ from rag2_0.tool.ModelTool import OpenAiLLM
 load_dotenv()
 
 # 示例查询
-examples_query = """ PE2211PK0801是什么软件"""
+examples_query = """T1软件中，配件和材料有什么区别"""
 conversation_context=""
 chat_history=[
         {
@@ -102,41 +102,30 @@ class QueryRewriteProcessor:
         
         doc_text_list = json.dumps(retrieved_doc, ensure_ascii=False, indent=2)
         class TempModel(BaseModel):
-            can_solve_problem: bool = Field(description="是否能解决用户问题")
-            relevance_score: int = Field(description="相关性评分，0-100分")
+            can_solve_problem: bool = Field(description="是否能解答用户问题")
+            relevance_score: int = Field(description="置信度评分，0-100分")
             explanation: str = Field(description="解释文档是否能解决(回答)提问")
 
         class all_relevant_document(BaseModel):
-            most_relevant_document: list[TempModel] = Field(description="最相关的文档的判断结果")
+            document_list: list[TempModel] = Field(description="每个文档的判断结果")
 
         parser = PydanticOutputParser(pydantic_object=all_relevant_document)
         # 构建提示词
-        prompt = f"""请判断以下检索文档列表中是否与用户提问相关,能够解决用户的问题，并给出相关性评分（0-100分）。输出最相关的文档的判断结果。
+        prompt = f"""请判断以下检索文档列表中是否解答用户提问,能够解决用户的问题,能够基于检索文档给出回答，并给出置信度评分（0-100分）。输出每个文档的判断结果。
+        用户提问: {query}
 
-用户提问: {query}
+        检索文档列表:
+        {doc_text_list}
 
-检索文档列表:
-{doc_text_list}
-
-请按照以下JSON格式返回结果:
-json```
-{{
-    "most_relevant_document":[{{
-        "can_solve_problem": true,
-        "relevance_score": 60,
-        "explanation":"xxxx"
-    }}]
-}}
-```
-
-"""
-        
+        请按照以下JSON格式返回结果:
+        {parser.get_format_instructions()}
+        """   
         try:
             # 初始化LLM并调用
-            llm = OpenAiLLM(api_key=self.api_key, base_url=self.base_url, model="deepseek-ai/DeepSeek-R1", response_format={"type": "json_object"})
+            llm = OpenAiLLM(api_key=self.api_key, base_url=self.base_url, model="deepseek-ai/DeepSeek-R1")
             response = llm.invoke(prompt)
             
-            result_list = parser.parse(response.content).most_relevant_document
+            result_list = parser.parse(response.content).document_list
             
             # 如果列表为空，返回默认的不相关结果
             if not result_list:
@@ -145,9 +134,11 @@ json```
                     "explanation": "无法解析文档相关性结果",
                     "relevance_score": 0.0
                 }
-            
+            true_document_list=[cur for cur in result_list if cur.can_solve_problem]
+            if len(true_document_list)==0:
+                true_document_list = result_list
             # 找出分数最高的文档
-            max_score_doc = max(result_list, key=lambda x: x.relevance_score)
+            max_score_doc = max(true_document_list, key=lambda x: x.relevance_score)
             
             return {
                 "is_relevant": max_score_doc.can_solve_problem,
@@ -155,12 +146,7 @@ json```
                 "explanation": max_score_doc.explanation
             }
         except Exception as e:
-            logging.error(f"判断文档相关性时出错: {str(e)}", exc_info=True)
-            return {
-                "is_relevant": False,
-                "explanation": f"判断过程出错: {str(e)}",
-                "relevance_score": 0.0
-            }
+            raise e
     
     def load_questions_from_excel(self, file_path=None):
         """
@@ -254,7 +240,7 @@ json```
                     "槽位信息": slot_filling_str,
                     "检索的文档": "\n".join(retrieved_doc_titles),
                     "检索的内容":  json.dumps(retrieved_doc, ensure_ascii=False, indent=2) if retrieved_doc else "",
-                    "文档是否相关": "相关" if relevance_result["is_relevant"] else "不相关",
+                    "文档能否解决问题": "能" if relevance_result["is_relevant"] else "不能",
                     "文档相关性解释": relevance_result["explanation"]
                 }
             except Exception as e:

rag2_0/demo/validate_excel_data_batch.py

@@ -555,8 +555,8 @@ def main():
     parser.add_argument("--input", "-i", type=str, help="输入Excel文件路径", default=input_excel)
     parser.add_argument("--output", "-o", type=str, help="输出结果Excel文件路径", default=output_excel)
     parser.add_argument("--workers", "-w", type=int, default=20, help="并行工作线程数")
-    logging.info(f"输入文件路径: {args.input}, 输出文件路径: {args.output}, 并行工作线程数: {args.workers}")
     args = parser.parse_args()
+    logging.info(f"输入文件路径: {args.input}, 输出文件路径: {args.output}, 并行工作线程数: {args.workers}")
     is_debug = hasattr(sys, 'gettrace') and sys.gettrace() is not None
     
     # 创建验证器实例并执行验证