优化招标文件判断&新增支持采购附件参数&优化招标文件验证切分正则

flask_app/general/merge_pdfs.py

@@ -1,8 +1,50 @@
 import os
 import fitz
-from PyPDF2 import PdfReader, PdfWriter
+import tempfile
+import logging
+from PyPDF2 import PdfReader, PdfWriter, PdfMerger
 from flask_app.general.读取文件.clean_pdf import create_get_text_function
 
+
+def merge_pdf_to(original_pdf, to_pdf, logger) -> str:
+    """ merge original pdf to exists pdf """
+    merger = None
+    temp_filename = None
+    try:
+        # 初始化PdfMerger
+        merger = PdfMerger()
+        
+        # 读取目标PDF的内容
+        with open(to_pdf, 'rb') as f:
+            merger.append(f)
+        
+        # 读取原始PDF的内容
+        with open(original_pdf, 'rb') as f:
+            merger.append(f)
+
+        # 创建临时文件
+        with tempfile.NamedTemporaryFile(dir=os.path.dirname(to_pdf), suffix='.pdf', delete=False) as temp_file:
+            temp_filename = temp_file.name
+            merger.write(temp_file)
+        
+        # 用临时文件替换原文件
+        os.replace(temp_filename, to_pdf)
+        return to_pdf
+    except Exception as e:
+        logger.error(f"merge_pdf_to 遇到错误: {str(e)}")
+        # 发生错误时删除临时文件
+        if temp_filename and os.path.exists(temp_filename):
+            try:
+                os.remove(temp_filename)
+            except:
+                pass
+        return to_pdf
+    finally:
+        # 确保merger和临时文件被正确关闭/清理
+        if merger is not None:
+            merger.close()
+
+
 #合并PDF
 # 主合并函数，尝试使用 PyPDF2，若失败则使用 fitz
 def merge_pdfs(paths, output_path):

flask_app/general/截取pdf通用函数.py

@@ -319,8 +319,11 @@ def get_invalid_file(pdf_path, output_folder, common_header, begin_page):
                         if regex.search(exclusion_pat, text):
                             continue
                         if regex.search(pattern, text):
+                            end_exclusion_pat = regex.compile(r'第[一二三四五六七八九十百千]+(?:章|部分)\s*“[^“”]*合同[^“”]*”', regex.DOTALL)
+                            if regex.search(end_exclusion_pat, text):
+                                continue
                             # print(f"结束模式在第 {i + 1} 页匹配: {pattern.pattern}")
-                            return i
+                            return i 
             else:
                 # 当页数较少时，从后向前查找（至少保留前30页）
                 for pattern in end_patterns:

flask_app/routes/upload.py

@@ -15,20 +15,20 @@ from flask_app.ConnectionLimiter import require_connection_limit, require_execut
 
 upload_bp = Blueprint('upload', __name__)
 
-def _child_target(main_func, queue, output_folder, file_path, file_type, unique_id):
+def _child_target(main_func, queue, output_folder, file_path, file_type, require_downloaded_filepath, require_file_type, unique_id):
     """
     子进程中调用 `main_func`（它是一个生成器函数），
     将其 yield 出的数据逐条放进队列，最后放一个 None 表示结束。
     """
     try:
-        for data in main_func(output_folder, file_path, file_type, unique_id):
+        for data in main_func(output_folder, file_path, file_type, require_downloaded_filepath, require_file_type, unique_id):
             queue.put(data)
     except Exception as e:
         # 如果要把异常也传给父进程，以便父进程可感知
         queue.put(json.dumps({'error': str(e)}, ensure_ascii=False))
     finally:
         queue.put(None)
-def run_in_subprocess(main_func, output_folder, file_path, file_type, unique_id):
+def run_in_subprocess(main_func, output_folder, file_path, file_type, require_downloaded_filepath, require_file_type, unique_id):
     """
     启动子进程调用 `main_func(...)`，并在父进程流式获取其输出（通过 Queue）。
     子进程结束时，操作系统回收其内存；父进程则保持实时输出。
@@ -36,7 +36,7 @@ def run_in_subprocess(main_func, output_folder, file_path, file_type, unique_id)
     queue = multiprocessing.Queue()
     p = multiprocessing.Process(
         target=_child_target,
-        args=(main_func, queue, output_folder, file_path, file_type, unique_id)
+        args=(main_func, queue, output_folder, file_path, file_type, require_downloaded_filepath, require_file_type, unique_id)
     )
     p.start()
 
@@ -59,11 +59,12 @@ def zbparse():                               #大解析
         logger.info("Received JSON data: " + str(received_data))
         file_url = g.file_url
         zb_type = g.zb_type
+        require_file_url = g.require_file_url
         file_name = urllib.parse.unquote(os.path.basename(file_url.split('?')[0]))
         logger.info(f"Starting parsing file: {file_name}")
         try:
             logger.info("starting parsing url:" + file_url)
-            return process_and_stream(file_url, zb_type)     #主要执行函数
+            return process_and_stream(file_url, zb_type, require_file_url)     #主要执行函数
         except Exception as e:
             logger.error('Exception occurred: ' + str(e))
             if hasattr(g, 'unique_id'):
@@ -84,7 +85,7 @@ def zbparse():                               #大解析
             data='Internal server error'
         )
         return jsonify(error_response)
-def process_and_stream(file_url, zb_type):
+def process_and_stream(file_url, zb_type, require_file_url):
     """
     下载文件并进行处理，支持工程标和货物标的处理。
     """
@@ -92,11 +93,14 @@ def process_and_stream(file_url, zb_type):
     unique_id = g.unique_id
     output_folder = g.output_folder
     filename = "ztbfile"
+    require_filename = "ztbfile_require"
     downloaded_filename = os.path.join(output_folder, filename)
+    require_downloaded_filename = os.path.join(output_folder, require_filename)
     start_time = time.time()
     file_name = urllib.parse.unquote(os.path.basename(file_url.split('?')[0]))
     try:
         downloaded_filepath, file_type = download_file(file_url, downloaded_filename, True)
+        require_downloaded_filepath, require_file_type = ("", -1) if not require_file_url else download_file(require_file_url, require_downloaded_filename, True)
         if not downloaded_filepath or file_type == 4:
             logger.error("下载文件失败或不支持的文件类型")
             log_error_unique_id(unique_id, 1, file_name=file_name)
@@ -119,7 +123,7 @@ def process_and_stream(file_url, zb_type):
         }
         processing_func = processing_functions.get(zb_type, goods_bid_main)
 
-        for data in run_in_subprocess(processing_func, output_folder, downloaded_filepath, file_type, unique_id):    #逐一接收货物标 工程标解析内容，为前端网页展示服务
+        for data in run_in_subprocess(processing_func, output_folder, downloaded_filepath, file_type, require_downloaded_filepath, require_file_type, unique_id):    #逐一接收货物标 工程标解析内容，为前端网页展示服务
             if not data.strip():
                 logger.error("Received empty data, skipping JSON parsing.")
                 continue

flask_app/routes/utils.py

@@ -16,13 +16,14 @@ def validate_request():
         return jsonify({'error': 'Missing JSON in request'}), 400
     file_url = request.json.get('file_url')
     zb_type = request.json.get('zb_type', 2)    #zb_type:默认按货物标解析
+    require_file_url = request.json.get('require_file_url', "")
     if not file_url:
         return jsonify({'error': 'No file URL provided'}), 400
     try:
         zb_type = int(zb_type)
     except (ValueError, TypeError):
         return jsonify({'error': 'Invalid zb_type provided'}), 400
-    return file_url, zb_type
+    return file_url, zb_type, require_file_url
 def generate_deviation_response(tech_deviation, tech_star_deviation, business_deviation, business_star_deviation,
                                 zigefuhe_deviation,proof_materials, logger):
     logger.info(f"技术偏离表: {json.dumps(tech_deviation, ensure_ascii=False, indent=4)}")
@@ -105,7 +106,7 @@ def validate_and_setup_logger(f):
         # 进行请求验证
         validation_result = validate_request()
         if isinstance(validation_result, tuple):
-            file_url, zb_type = validation_result
+            file_url, zb_type, require_file_url = validation_result
 
             # 根据蓝图确定子文件夹
             blueprint = request.blueprint
@@ -125,6 +126,7 @@ def validate_and_setup_logger(f):
             # 将验证后的数据存储在 g 对象中
             g.file_url = file_url
             g.zb_type = zb_type
+            g.require_file_url = require_file_url
             return f(*args, **kwargs)
         else:
             # 验证失败，返回错误响应

flask_app/routes/判断是否是招标文件.py

@@ -2,30 +2,76 @@ import time
 from PyPDF2 import PdfReader  # 确保已安装 PyPDF2: pip install PyPDF2
 from docx import Document
 import fitz
+from flask_app.general.llm.qianwen_plus import qianwen_plus
 from flask_app.general.llm.通义千问long import upload_file, qianwen_long
 
-def judge_zbfile_exec(file_path):
+
+def pre_judge(few_pages_content: str):
+    if few_pages_content.strip() == "":
+        return "未知"
+
+    # 使用大模型进行判断
+    user_query = f"""
+    一, 任务描述: 
+        我会给定一段从文件中摘取出来的文本, 你需要根据该文本的内容来判断文件是否属于招标文件。
+    
+    二, 判断准则: 
+        以下是常见的招标文件类型：
+            公开招标文件、邀请招标文件、竞争性谈判文件、竞争性磋商文件、询价文件、问询文件、货物类招标文件、工程类招标文件、施工类招标文件、服务类招标文件、比选文件。
+        若有未涵盖的类型，但其内容明确表达了项目需求、采购或招标信息，且包含指导投标人参与的关键要素，则可视为招标文件。
+        排除情况:
+            1. 请注意区分招标文件和投标文件,若文件仅有投标文件格式要求部分,或是投标、响应性文件，则不视为招标文件。
+            2. 若文件内容为乱码,无有效信息,请直接返回"否"。
+
+    三, 注意事项: 你并不是完全依赖于判断准则来给出判断, 而是根据自己对招标文件的理解来判断, 给定判断准则只是帮助你进行理解。
+    
+    四, 回答格式: 你的回答只能为"是","否","未知"三种情况之一, 不要在回答中进行任何的解释或输出其他不属于"是","否","未知"的字符。
+        
+    五, 给定从文件中摘取出来的文本:
+        {few_pages_content}
+            
+    提问: 该内容是否属于招标文件? 如果属于返回"是"; 如果不属于则返回"否"; 如果你无法判断则返回"未知", 不要进行假设性判断, 不理解就返回"未知"。
+    """
+    start_time = time.time()
+    model_res = qianwen_plus(user_query)
+    end_time = time.time()
+    print(f"pre_judge 判断是否属于招标文件：{model_res} 实际耗时：{end_time - start_time:.2f} 秒")
+
+    if "是" in model_res:
+        return "是"
+    elif "否" in model_res:
+        return "否"
+    else:
+        return "未知"
+
+
+def judge_zbfile_exec(file_path, end_page=5, max_len=3000):
     """
     判断文件是否属于招标文件，并返回结果。
     """
     try:
         start_time = time.time()
+        few_pages_content = ""
         # 检查文件是否为PDF格式
         if file_path.lower().endswith('.pdf'):
             try:
                 with open(file_path, 'rb') as f:
                     reader = PdfReader(f)
                     num_pages = len(reader.pages)
+                    if num_pages > end_page:
+                        few_pages_content = "".join([reader.pages[i].extract_text() for i in range(end_page)])
             except Exception:
                 try:
                     doc = fitz.open(file_path)
                     num_pages = len(doc)
+                    if num_pages > end_page:
+                        few_pages_content = "".join([doc.load_page(i).get_text() for i in range(end_page)])
                 except Exception:
                     print("PDF 文件读取失败")
                     return False  # 两种解析方式都失败，直接返回 False
-
             if num_pages <= 5:
                 return False  # 小于等于 5 页的 PDF 直接判定为非招标文件
+            few_pages_content = few_pages_content[:max_len]
         elif file_path.lower().endswith('.docx'):
             try:
                 doc = Document(file_path)
@@ -37,13 +83,20 @@ def judge_zbfile_exec(file_path):
                     chunk = paragraphs[i:i + chunk_size]
                     for para in chunk:
                         accumulated_text += para.text
-                    if len(accumulated_text) >= 1000:
+                    if len(accumulated_text) >= max_len:
                         break  # 读取超过1000字后即可停止
                 if len(accumulated_text) < 1000:
                     return False  # 若累计内容不足1000字，则直接返回 False
             except Exception:
                 print("DOCX 文件读取失败，可能为乱码文件")
                 return False  # 解析失败直接返回 False
+            few_pages_content = accumulated_text[:max_len]
+    
+        # pre_judge -> 是, 否, 未知
+        pre_res = pre_judge(few_pages_content)
+        if pre_res != "未知":
+            return '否' not in pre_res
+        
         pre_endtime=time.time()
         print(f"judge_zbfile_exec预处理耗时：{pre_endtime - start_time:.2f} 秒")
         # 使用大模型进行判断
@@ -60,7 +113,7 @@ def judge_zbfile_exec(file_path):
         model_res = qianwen_long(file_id, user_query)
         end_time = time.time()
         print(f"judge_zbfile_exec实际耗时：{end_time - start_time:.2f} 秒")
-        print(f"判断是否属于招标文件：{model_res} 实际耗时：{end_time - start_time:.2f} 秒")
+        print(f"qianwen_long 判断是否属于招标文件：{model_res} 实际耗时：{end_time - start_time:.2f} 秒")
 
         return '否' not in model_res
 

flask_app/routes/工程标解析main.py

@@ -33,7 +33,7 @@ def engineering_bid_main_process(output_folder, file_path, file_type, unique_id)
     return result
 
 
-def preprocess_files(output_folder, file_path, file_type, logger):
+def preprocess_files(output_folder, file_path, file_type, require_file_path, require_file_type, logger):
     logger.info("starting 文件预处理...")
     start_time = time.time()
     is_pure_image_flag = False
@@ -239,10 +239,10 @@ def fetch_bid_opening(invalid_deleted_docx, merged_baseinfo_path_more, clause_pa
     return result
 
 
-def engineering_bid_main(output_folder, file_path, file_type, unique_id):
+def engineering_bid_main(output_folder, file_path, file_type, require_file_path, require_file_type, unique_id):
     logger = get_global_logger(unique_id)
     # 预处理文件，获取处理后的数据
-    processed_data = preprocess_files(output_folder, file_path, file_type, logger)
+    processed_data = preprocess_files(output_folder, file_path, file_type, require_file_path, require_file_type, logger)
     if not processed_data:
         error_response = {
             'error': '文件预处理失败。请检查文件类型并重试。'

flask_app/routes/货物标解析main.py

@@ -15,17 +15,18 @@ from flask_app.general.无效标和废标公共代码 import combine_find_invali
 from flask_app.货物标.资格审查main import combine_qualification_review
 from flask_app.general.商务技术评分提取 import combine_evaluation_standards
 from concurrent.futures import ThreadPoolExecutor
+from flask_app.general.merge_pdfs import merge_pdf_to, merge_pdfs
+
+
+def verify_file(file_path, file_type, logger):
+    pdf_path = ""
+    is_pure_image_flag = False  # 判断是否为纯图片类型的docx
 
-def preprocess_files(output_folder, file_path, file_type,logger):
-    logger.info("starting 文件预处理...")
-    start_time = time.time()
-    is_pure_image_flag=False        #判断是否为纯图片类型的docx
-    pdf_path=""
     # 根据文件类型处理文件路径
     if file_type == 1:  # docx
         # docx_path = file_path
         if is_pure_image(file_path):
-            is_pure_image_flag=True
+            is_pure_image_flag = True
         else:
             pdf_path = docx2pdf(file_path)  # 将docx转换为pdf以供后续处理
     elif file_type == 2:  # pdf
@@ -36,7 +37,37 @@ def preprocess_files(output_folder, file_path, file_type,logger):
         pdf_path = docx2pdf(file_path)  # 将docx转换为pdf以供后续处理
     else:
         logger.error("Unsupported file type provided. Preprocessing halted.")
+        return None, None
+
+    return pdf_path, is_pure_image_flag
+
+
+def preprocess_files(output_folder, file_path, file_type, require_file_path, require_file_type, logger):
+    logger.info("starting 文件预处理...")
+    start_time = time.time()
+    # is_pure_image_flag=False        #判断是否为纯图片类型的docx
+    # pdf_path=""
+    # # 根据文件类型处理文件路径
+    # if file_type == 1:  # docx
+    #     # docx_path = file_path
+    #     if is_pure_image(file_path):
+    #         is_pure_image_flag=True
+    #     else:
+    #         pdf_path = docx2pdf(file_path)  # 将docx转换为pdf以供后续处理
+    # elif file_type == 2:  # pdf
+    #     pdf_path = file_path
+    #     # docx_path = pdf2docx(pdf_path)
+    # elif file_type == 3:  # doc
+    #     # docx_path = doc2docx(file_path)
+    #     pdf_path = docx2pdf(file_path)  # 将docx转换为pdf以供后续处理
+    # else:
+    #     logger.error("Unsupported file type provided. Preprocessing halted.")
+    #     return None
+    pdf_path, is_pure_image_flag = verify_file(file_path, file_type, logger)
+    require_pdf_path, _ = verify_file(require_file_path, require_file_type, logger)
+    if pdf_path is None:
         return None
+    
     if not is_pure_image_flag:     #大多数情况 不是纯图片doc/docx
         # 调用截取PDF多次
         truncate_files = truncate_pdf_multiple(pdf_path, output_folder,logger,'goods')
@@ -44,7 +75,9 @@ def preprocess_files(output_folder, file_path, file_type,logger):
         truncate_files=['','','','','','',file_path,'']  #纯图片，无需切片
     # print("切割出的文件："+str(truncate_files))
     # 处理各个部分
-    procurement_path = truncate_files[5]  # 采购需求
+    # procurement_path = truncate_files[5]  # 采购需求
+    procurement_path = truncate_files[5] if not require_pdf_path else merge_pdf_to(require_pdf_path, truncate_files[5], logger)   # 采购需求
+    # procurement_path = truncate_files[5] if not require_pdf_path else merge_pdfs([truncate_files[5], require_pdf_path], truncate_files[5])
     evaluation_method_path = truncate_files[1]  # 评标办法
     qualification_path = truncate_files[2]  # 资格审查
     tobidders_notice_path = truncate_files[4]  # 投标人须知正文
@@ -223,10 +256,10 @@ def post_process_baseinfo(base_info,logger):
         logger.error(f"Error in post_process_baseinfo: {e}")
         return base_info, []  # 返回空列表
 
-def goods_bid_main(output_folder, file_path, file_type, unique_id):
+def goods_bid_main(output_folder, file_path, file_type, require_file_path, require_file_type, unique_id):
     logger = get_global_logger(unique_id)
     # 预处理文件，获取处理后的数据
-    processed_data = preprocess_files(output_folder, file_path, file_type,logger)
+    processed_data = preprocess_files(output_folder, file_path, file_type, require_file_path, require_file_type, logger)
     if not processed_data:
         error_response = {
             'error': '文件预处理失败。请检查文件类型并重试。'