更新环境配置，添加gevent和gunicorn依赖；新增chat_dify_by_workorder.py文件以处理工单对话逻辑；优化PgSql类中的异常处理，确保连接失败时抛出异常；改进意图识别API，使用单例模式管理意图识别器实例，增强线程安全性；新增workflow_chat.py文件以支持新工作流对话功能。

2025-06-05 10:52:31 +08:00
parent 01dc1c3c91
commit c715fca5ef
8 changed files with 784 additions and 42 deletions
@@ -0,0 +1,151 @@
+from rag2_0.dify.workflow_chat import NewWorkflowChat
+import pandas as pd
+from concurrent.futures import ThreadPoolExecutor
+from tqdm import tqdm
+import concurrent.futures
+
+
+class ChatDifyByWorkorder:
+
+    def __init__(self, api_key=None, base_url="https://api.dify.ai/v1") -> None:
+        """
+        初始化ChatDifyByWorkorder类
+        
+        Args:
+            api_key: Dify API密钥，默认为None
+            base_url: Dify API的基础URL，默认为"https://api.dify.ai/v1"
+        """
+        baseurl = "http://172.20.0.145/v1" 
+        new_workflow_api_key = "app-qxsSybCs7ABiKlC1JabTYVn6"
+        self.new_chat = NewWorkflowChat(api_key=new_workflow_api_key, base_url=baseurl)
+        self.new_chat_answer = NewWorkflowChat(api_key=new_workflow_api_key, base_url=baseurl)
+        
+
+    def get_soft_name(self, row) -> str:
+        if "博微配网计价通D3" in row["产品线"]:
+            return "博微配网计价通D3"
+        elif "博微电力建设计价通软件" in row["产品线"]:
+            return "电力建设计价通软件"
+        elif "新能源系列" in row["产品线"] and "博微新型储能电站建设计价通C1软件" in row["产品名称"]:
+            return "储能C1软件"
+        elif "博微西藏计价通Z1" in row["产品线"]:
+            return "西藏计价通Z1"
+        elif "博微技改检修计价通T1软件" in row["产品线"] and "技改检修计价通T1软件-概预算" in row["产品名称"]:
+            return "技改检修工程计价通T1"
+        elif "博微技改检修计价通T1软件" in row["产品线"] and "技改检修计价通T1软件-清单" in row["产品名称"]:
+            return "检修清单计价通T1"
+        return ""
+
+    def process_query(self, q:str) -> dict:
+        """
+        发送问题并获取回答及相关工作流信息
+        
+        Args:
+            q: 用户问题
+            
+        Returns:
+            dict: 包含问题、回答和工作流信息的字典
+        """
+        retry_count = 0
+        max_retries = 2
+        
+        while retry_count <= max_retries:
+            try:
+                # 发送问题获取回答和消息ID
+                result = self.new_chat.process_question(q)
+                return result
+            except Exception as e:
+                retry_count += 1
+                if retry_count <= max_retries:
+                    continue
+                else:
+                    raise e
+
+    def process_answer(self, q:str) -> dict:
+        """
+        发送问题并获取回答及相关工作流信息
+        
+        Args:
+            q: 用户问题
+            
+        Returns:
+            dict: 包含问题、回答和工作流信息的字典
+        """
+        retry_count = 0
+        max_retries = 2
+        
+        while retry_count <= max_retries:
+            try:
+                # 发送问题获取回答和消息ID
+                result = self.new_chat_answer.process_question(q)
+                return result
+            except Exception as e:
+                retry_count += 1
+                if retry_count <= max_retries:
+                    continue
+                else:
+                    raise
+
+    def process_row(self, row):
+        """处理单行数据"""
+        soft_name = self.get_soft_name(row=row)
+        if soft_name == "":
+            return None
+            
+        # 使用线程池并发执行查询
+        with ThreadPoolExecutor() as executor:
+            try:
+                # 提交两个任务并获取Future对象
+                query_future = executor.submit(self.process_query, q=f"{soft_name},{row['客户问题']}")
+                answer_future = executor.submit(self.process_answer, q=f"{soft_name},{row['解决方案']}")
+                
+                # 获取结果
+                query_result = query_future.result()
+                answer_result = answer_future.result()
+            except Exception as e:
+                print(f"处理工单 {row.get('工单编号', '未知')} 时发生错误: {str(e)}")
+                return None
+        
+        worker_id = str(row["工单编号"])
+        if query_result is None or answer_result is None:
+            print("处理对话出现错误")
+            return None
+
+        worker_order_info = {
+            "工单编号": worker_id,
+            "用户问题": row['客户问题'],
+            "解决方案": row['解决方案'],
+            "AI回答": query_result["新流程答案"],
+            "用户问题检索到的词条": query_result["新检索词条"],
+            "解决方案检索到的词条": answer_result["新检索词条"],
+        }
+        return worker_order_info
+
+    def run(self, excel_path:str):
+        df_data = pd.read_excel(excel_path)
+        list_worker_order_info = []
+        
+        # 创建进度条
+        with tqdm(total=len(df_data), desc="处理工单") as pbar:
+            # 创建线程池，最大并发数可以根据需要调整
+            with ThreadPoolExecutor(max_workers=5) as executor:
+                # 提交所有任务
+                future_to_row = {executor.submit(self.process_row, row): idx for idx, row in df_data.iterrows()}
+                
+                # 处理完成的任务
+                for future in concurrent.futures.as_completed(future_to_row):
+                    result = future.result()
+                    if result is not None:
+                        list_worker_order_info.append(result)
+                    pbar.update(1)
+                    
+        return list_worker_order_info
+
+
+
+if __name__=="__main__":
+    worker_chat = ChatDifyByWorkorder()
+    result = worker_chat.run(excel_path="data/excel/工单记录_均衡提取2000条.xlsx")
+    # 可以选择保存结果到Excel
+    if result:
+        pd.DataFrame(result).to_excel("data/excel/工单处理结果.xlsx", index=False)