feat: 添加 PaddleX OCR 支持及示例更新

- 在 example_usage.py 中更新分析文档的示例，使用新的 PaddleX OCR 服务。
- 在 data_router.py 中优化文件处理反馈信息，增加处理失败文件的统计。
- 在 indexing.py 中新增对 PaddleX OCR 的支持。
- 在 _ocr.py 中添加 process_pdf_paddlex 方法，集成 PaddleX OCR 处理逻辑。
- 新增 paddlex.py 文件，封装 PaddleX 版面解析服务的 API 调用。
- 更新前端 DataBaseInfoView.vue，增加 PaddleX OCR 选项。
- 新增测试文件 PixPin_2025-06-19_23-42-17.png 以支持 OCR 测试。

scripts/paddlestructure/example_usage.py

@@ -1,11 +1,10 @@
 #!/usr/bin/env python3
-# -*- coding: utf-8 -*-
 """
 PaddleX 文档分析使用示例
 演示如何使用 analyze_document 函数分析文档
 """
 
-from paddlex_layout_parser import analyze_document
+from src.plugins.paddlex import analyze_document
 import json
 
 
@@ -14,7 +13,7 @@ def analyze_custom_file(file_path: str):
 
     print(f"\n🔍 分析自定义文件: {file_path}")
 
-    result = analyze_document(file_path=file_path)
+    result = analyze_document(file_path=file_path, base_url="http://172.19.13.5:8080")
 
     if result["success"]:
         return result
@@ -27,6 +26,7 @@ if __name__ == "__main__":
     # main()
 
     # 如果您想分析其他文件，可以取消注释下面的代码
-    custom_file = "test/struct_pdf/P020241226617572090546.pdf"
+    custom_file = "test/data/ocr_test/1706.03762v7_扫描版.pdf"
-    custom_file = "test/data/PixPin_2025-06-19_23-42-17.png"
+    custom_file = "test/data/ocr_test/PixPin_2025-06-19_23-42-17.png"
-    print(analyze_custom_file(custom_file))
+
+    print(analyze_custom_file(custom_file))

server/routers/data_router.py

@@ -56,7 +56,9 @@ async def file_to_chunk(db_id: str = Body(...), files: list[str] = Body(...), pa
     logger.debug(f"File to chunk for db_id {db_id}: {files} {params=}")
     try:
         processed_files = await knowledge_base.save_files_for_pending_indexing(db_id, files, params)
-        return {"message": "Files processed and pending indexing", "files": processed_files, "status": "success"}
+        processed_failed_count = len([_p['status'] == 'failed' for _p in processed_files])
+        processed_info = f"Processed {len(processed_files)} files for pending indexing, {processed_failed_count} files failed"
+        return {"message": processed_info, "files": processed_files, "status": "success"}
     except Exception as e:
         logger.error(f"Failed to process files for pending indexing: {e}, {traceback.format_exc()}")
         return {"message": f"Failed to process files for pending indexing: {e}", "status": "failed"}

src/core/indexing.py

@@ -133,6 +133,10 @@ def parse_pdf(file, params=None):
         from src.plugins import ocr
         return ocr.process_pdf_mineru(file)
 
+    elif opt_ocr == "paddlex_ocr":
+        from src.plugins import ocr
+        return ocr.process_pdf_paddlex(file)
+
     else:
         return pdfreader(file, params=params)
 

src/plugins/_ocr.py

@@ -159,11 +159,10 @@ class OCRPlugin:
         :param pdf_path: PDF文件路径
         :return: 提取的文本
         """
+        import requests
+        from .mineru import parse_doc
         mineru_ocr_uri = os.getenv("MINERU_OCR_URI", "http://localhost:30000")
         mineru_ocr_uri_health = f"{mineru_ocr_uri}/health"
-        import requests
-        import json
-        from .mineru import parse_doc
 
         health_check_response = requests.get(mineru_ocr_uri_health, timeout=5)
         if health_check_response.status_code != 200:
@@ -180,6 +179,28 @@ class OCRPlugin:
         logger.debug(f"Mineru OCR result: {pdf_text[:50]}(...) total {len(pdf_text)} characters.")
         return pdf_text
 
+    def process_pdf_paddlex(self, pdf_path):
+        """
+        使用Paddlex OCR处理PDF文件
+        :param pdf_path: PDF文件路径
+        :return: 提取的文本
+        """
+        from .paddlex import analyze_document, check_paddlex_health
+
+        paddlex_uri = os.getenv("PADDLEX_URI", "http://localhost:8080")
+        health_check_response = check_paddlex_health(paddlex_uri)
+        if not health_check_response.ok:
+            logger.error(f"Paddlex OCR service health check failed with {paddlex_uri}: {health_check_response.json()}")
+            raise RuntimeError("Paddlex OCR service health check failed. Please check the log use `docker logs paddlex`")
+
+        result = analyze_document(pdf_path, base_url=paddlex_uri)
+
+        if not result["success"]:
+            logger.error(f"Paddlex OCR failed: {result['error']}")
+            return ""
+
+        return result["full_text"]
+
 def get_state(task_id):
     return GOLBAL_STATE.get(task_id, {})
 

src/plugins/paddlex.py

@@ -4,8 +4,8 @@ import json
 import base64
 import os
 import time
-from typing import Optional, Dict, Any
+from typing import Optional, Any
-
+from src.utils import logger
 
 
 
@@ -24,43 +24,43 @@ class PaddleXLayoutParser:
     def _process_file_input(self, file_input: str) -> str:
         # 检查是否为本地文件路径
         if os.path.exists(file_input):
-            print(f"📁 检测到本地文件: {file_input}")
+            logger.info(f"📁 检测到本地文件: {file_input}")
-            print(f"📏 文件大小: {os.path.getsize(file_input) / 1024 / 1024:.2f} MB")
+            logger.info(f"📏 文件大小: {os.path.getsize(file_input) / 1024 / 1024:.2f} MB")
 
             try:
                 # 将本地文件编码为Base64
                 encoded_content = self.encode_file_to_base64(file_input)
-                print(f"✅ 文件已编码为Base64，长度: {len(encoded_content)} 字符")
+                logger.info(f"✅ 文件已编码为Base64，长度: {len(encoded_content)} 字符")
                 return encoded_content
             except Exception as e:
-                print(f"❌ 文件编码失败: {e}")
+                logger.error(f"❌ 文件编码失败: {e}")
                 raise
 
         # 检查是否为URL
         elif file_input.startswith(('http://', 'https://')):
-            print(f"🌐 检测到URL: {file_input}")
+            logger.info(f"🌐 检测到URL: {file_input}")
             return file_input
 
         # 否则假设为Base64编码内容
         else:
-            print(f"📝 假设为Base64编码内容，长度: {len(file_input)} 字符")
+            logger.info(f"📝 假设为Base64编码内容，长度: {len(file_input)} 字符")
             return file_input
 
     def layout_parsing(self,
             file_input: str,
-            file_type: Optional[int] = None,
+            file_type: int | None = None,
-            use_textline_orientation: Optional[bool] = None,
+            use_textline_orientation: bool | None = None,
-            use_seal_recognition: Optional[bool] = None,
+            use_seal_recognition: bool | None = None,
-            use_table_recognition: Optional[bool] = None,
+            use_table_recognition: bool | None = None,
-            use_formula_recognition: Optional[bool] = None,
+            use_formula_recognition: bool | None = None,
-            use_chart_recognition: Optional[bool] = None,
+            use_chart_recognition: bool | None = None,
-            use_region_detection: Optional[bool] = None,
+            use_region_detection: bool | None = None,
-            layout_threshold: Optional[float] = None,
+            layout_threshold: float | None = None,
-            layout_nms: Optional[bool] = None,
+            layout_nms: bool | None = None,
-            use_doc_orientation_classify: Optional[bool] = True,
+            use_doc_orientation_classify: bool = True,
-            use_doc_unwarping: Optional[bool] = False,
+            use_doc_unwarping: bool | None = False,
-            use_wired_table_cells_trans_to_html: Optional[bool] = True, # 是否启用无有线表单元格检测结果直转HTML，默认False，启用则直接基于有线表单元格检测结果的几何关系构建HTML。
+            use_wired_table_cells_trans_to_html: bool = True, # 启用则直接基于有线表单元格检测结果的几何关系构建HTML。
-            **kwargs) -> Dict[str, Any]:
+            **kwargs) -> dict[str, Any]:
         """
         调用版面解析API：https://paddlepaddle.github.io/PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/PP-StructureV3.html#22-python
         """
@@ -104,28 +104,30 @@ class PaddleXLayoutParser:
 
             if response.status_code == 200:
                 result = response.json()
-                print("✅ 请求成功!")
+                logger.info("✅ 请求成功!")
                 return result
             else:
-                print("❌ 请求失败!")
+                logger.error("❌ 请求失败!")
                 try:
                     error_result = response.json()
-                    print(f"错误信息: {json.dumps(error_result, indent=2, ensure_ascii=False)}")
+                    logger.error(f"错误信息: {json.dumps(error_result, indent=2, ensure_ascii=False)}")
                     return error_result
-                except:
+                except Exception as e:
-                    print(f"响应内容: {response.text}")
+                    logger.error(f"响应内容: {response.text}")
-                    return {"error": response.text, "status_code": response.status_code}
+                    return {"error": f"{e}: {response.text}", "status_code": response.status_code}
 
         except requests.exceptions.RequestException as e:
-            print(f"❌ 网络请求异常: {e}")
+            health_check_response = requests.get(f"{self.base_url}/health", timeout=5)
+            logger.error(f"❌ 网络请求异常: {e}: {health_check_response.json()}")
             return {"error": str(e)}
+
         except Exception as e:
-            print(f"❌ 其他异常: {e}")
+            logger.error(f"❌ 其他异常: {e}")
             return {"error": str(e)}
 
 
 
-def _parse_recognition_result(api_result: Dict[str, Any], file_path: str) -> Dict[str, Any]:
+def _parse_recognition_result(api_result: dict[str, Any], file_path: str) -> dict[str, Any]:
     # 基本信息
     parsed_result = {
         "success": True,
@@ -238,7 +240,7 @@ def _parse_recognition_result(api_result: Dict[str, Any], file_path: str) -> Dic
     return parsed_result
 
 
-def analyze_document(file_path: str) -> Dict[str, Any]:
+def analyze_document(file_path: str, base_url: str = "http://localhost:8080") -> dict[str, Any]:
 
     # 检查文件是否存在
     if not os.path.exists(file_path):
@@ -249,7 +251,7 @@ def analyze_document(file_path: str) -> Dict[str, Any]:
         }
 
     # 初始化客户端
-    client = PaddleXLayoutParser()
+    client = PaddleXLayoutParser(base_url=base_url)
 
     # 判断文件类型
     file_ext = os.path.splitext(file_path)[1].lower()
@@ -264,9 +266,9 @@ def analyze_document(file_path: str) -> Dict[str, Any]:
             "file_path": file_path
         }
 
-    print(f"📄 开始分析文档: {os.path.basename(file_path)}")
+    logger.info(f"📄 开始分析文档: {os.path.basename(file_path)}")
-    print(f"📏 文件大小: {os.path.getsize(file_path) / 1024 / 1024:.2f} MB")
+    logger.info(f"📏 文件大小: {os.path.getsize(file_path) / 1024 / 1024:.2f} MB")
-    print(f"📋 文件类型: {'PDF' if file_type == 0 else '图片'}")
+    logger.info(f"📋 文件类型: {'PDF' if file_type == 0 else '图片'}")
 
     try:
         # 调用API进行识别
@@ -291,3 +293,7 @@ def analyze_document(file_path: str) -> Dict[str, Any]:
             "error": f"处理异常: {str(e)}",
             "file_path": file_path
         }
+
+
+def check_paddlex_health(base_url: str = "http://localhost:8080") -> bool:
+    return requests.get(f"{base_url}/health", timeout=5)

web/src/views/DataBaseInfoView.vue

@@ -445,6 +445,7 @@ const enable_ocr_options = ref([
   { value: 'disable', payload: { title: '不启用' } },
   { value: 'onnx_rapid_ocr', payload: { title: 'ONNX with RapidOCR' } },
   { value: 'mineru_ocr', payload: { title: 'MinerU OCR' } },
+  { value: 'paddlex_ocr', payload: { title: 'Paddlex OCR' } },
 ])
 
 const use_rewrite_queryOptions = ref([
@@ -811,7 +812,7 @@ const chunkFiles = () => {
   .then(data => {
     console.log('文件处理结果:', data)
     if (data.status === 'success') {
-      message.success(data.message || '文件已提交处理，请稍后在列表刷新查看状态');
+      message.info(data.message || '文件已提交处理，请稍后在列表刷新查看状态');
       fileList.value = []; // 清空已上传文件列表
       addFilesModalVisible.value = false; // 关闭弹窗
       getDatabaseInfo(); // 刷新数据库信息以显示新文件及其状态