Update: Refactor to absolute imports, fix QPS limits, and enhance Gitea sync with SQLite support

app/services/ocr_service.py

@@ -7,9 +7,9 @@ OCR服务模块
 from typing import Dict, List, Optional, Tuple, Union, Any, Callable
 import os
 
-from ..config.settings import ConfigManager
-from ..core.utils.log_utils import get_logger
-from ..core.ocr.table_ocr import OCRProcessor
+from app.config.settings import ConfigManager
+from app.core.utils.log_utils import get_logger
+from app.core.ocr.table_ocr import OCRProcessor
 
 logger = get_logger(__name__)
 
@@ -62,16 +62,10 @@ class OCRService:
             if not self._is_valid_image(image_path):
                 logger.error(f"不支持的文件类型: {image_path}")
                 return None
-                
-            # 检查是否已处理
-            excel_file = self._get_excel_path(image_path)
-            if os.path.exists(excel_file):
-                logger.info(f"文件已处理过，跳过OCR识别: {image_path}")
-                # 即使 xlsx 已存在，仍补写 meta.json（首次 OCR 后可能未生成）
-                meta_path = str(excel_file).replace('.xlsx', '.meta.json')
-                if not os.path.exists(meta_path):
-                    self._write_meta_json(str(excel_file), {}, image_path)
-                return excel_file
+
+            # 不再做 xlsx 已存在就跳过的判断（xlsx 可能被业务层重命名过），
+            # 跳过逻辑交给 OCRProcessor 内部的 record_manager.is_processed(image_path)
+            # —— 由它读 processed_files.json（业务层会同步更新 xlsx 路径）
                 
             # 执行OCR识别
             result = self.ocr_processor.process_image(image_path)
@@ -212,24 +206,36 @@ class OCRService:
             base = Path(excel_path)
             meta_path = base.with_suffix('.meta.json')
 
-            # 0) 优先：调百度通用文字识别（/accurate），覆盖全图文字（含手写抬头/日期）
+            # 0) 优先：调百度通用票务识别（/general_ocr），精准捕获表头（含手写抬头/日期）
             general_text = ''
             general_lines = []
             try:
-                # OCRService.ocr_processor = core.ocr.table_ocr.OCRProcessor
-                # 其 .ocr_client = BaiduOCRClient (process_image 时初始化)
                 client = getattr(self.ocr_processor, 'ocr_client', None)
                 if client is None:
-                    # 显式触发一次 process_image 准备流程 (不会重复 OCR)
                     try:
                         self.ocr_processor._ensure_ocr_client()
                     except Exception:
                         pass
                     client = getattr(self.ocr_processor, 'ocr_client', None)
-                if client and hasattr(client, 'recognize_general') and image_path and os.path.exists(str(image_path)):
+                
+                # 关键：在进行第二次 OCR (票务识别) 前稍微等待，避免触发 QPS 限制
+                import time
+                time.sleep(0.5)
+
+                # 优先使用票务识别 (Ticket OCR)
+                if client and hasattr(client, 'recognize_ticket') and image_path and os.path.exists(str(image_path)):
+                    logger.info("使用通用票务识别捕获表头元数据...")
+                    words = client.recognize_ticket(str(image_path))
+                    if words:
+                        general_lines = [(w.get('words') or '').strip() for w in words if (w.get('words') or '').strip()]
+                        general_text = '\n'.join(general_lines)
+                        logger.info(f"票务识别获取 {len(general_lines)} 行文字")
+                
+                # 如果票务识别无结果或不可用，则兜底使用高精度文字识别
+                if not general_lines and client and hasattr(client, 'recognize_general') and image_path and os.path.exists(str(image_path)):
+                    logger.info("尝试高精度文字识别作为兜底...")
                     words = client.recognize_general(str(image_path))
                     if words:
-                        # 按 location.top 排序（顶部先），方便后续提取供应商/日期
                         def _top(w):
                             loc = w.get('location') or {}
                             try:
@@ -239,9 +245,9 @@ class OCRService:
                         words_sorted = sorted(words, key=_top)
                         general_lines = [(w.get('words') or '').strip() for w in words_sorted if (w.get('words') or '').strip()]
                         general_text = '\n'.join(general_lines)
-                        logger.info(f"通用识别获取 {len(general_lines)} 行文字")
+                        logger.info(f"通用高精度识别获取 {len(general_lines)} 行文字")
             except Exception as e:
-                logger.warning(f"通用识别失败（不影响主流程）: {e}")
+                logger.warning(f"获取元数据识别失败（不影响主流程）: {e}")
 
             # 1) 表格识别 header/body 拼表内文字（已在前面逻辑处理）
             ocr_text = ''