feat: OCR 服务 + 会议材料模块

ry-ocr/ (新)
  本地发票识别微服务 (PaddleOCR 3.x + FastAPI, 8801)
  - QR 优先: 扫到二维码即取开票时间/发票号/金额; 没扫到/格式不合法直接判非发票, 不跑 OCR
  - 配置 QR_FULL_OCR 控制快路径(false, 0.2s)还是全字段(true, 4.5s)
  - /recognize/invoice (multipart) + /recognize/invoice/by-path (本地路径, 白名单) + /recognize/text
  - is_invoice / from_qr / qr_raw / qr_error / error_code 字段
  - 12 字段发票抽取 (regex + 启发式, 左右主体识别)
  - 超时保护 (15s 单页 / 60s 总流程) + PaddleOCR 单例 + ThreadPoolExecutor

ry-api/ruoyi-business/
  - pom.xml: 加 hutool-http/json/core 5.8.27, lombok 1.18.30 (OcrClient @Slf4j 所需)
  - ocr/: OcrClient + InvoiceResult/Fields/Line + ZipExtractor + InvoiceOcrScheduler
  - oss/: OssUploader + OssConfMeta (OCR 识别后重传 OSS)
  - config/: OcrConfig + OcrExecutorConfig (后台线程池)
  - service/impl/InvoiceOcrService: 后台提交 OCR, ZIP 路径解压识别, 替换场景先清旧
  - 会议材料 CRUD 全套 (BizMeetingAuditLog/Executor/Invoice/Material/Supervisor):
    controller + service + mapper + domain + xml

ry-vue3/
  - MeetingDetail.vue (新建): 会议详情页 (含评分维度章节, 改只读)
  - Meetings.vue / OssFileUploader.vue / router / Login.vue: 适配新字段

ry-api/ruoyi-admin/
  - RuoYiApplication.java + application.yml: 启用 @Async 异步支持

_self/
  - manager_meetings.md / manager_meeting_detail.md: 文档

ry-ocr/app/core/image_processor.py

@@ -0,0 +1,47 @@
+"""图像预处理：自动旋转、放大、去噪"""
+from __future__ import annotations
+
+from pathlib import Path
+
+import cv2
+import numpy as np
+
+
+def auto_rotate(img_path: str | Path) -> Path:
+    """基于方向检测的简易旋转（> 阈值倾斜就转 90°）
+
+    说明：发票多为横向，此处只处理 0/90/180/270 四方向，
+    复杂倾斜交给 PaddleOCR 自带的 textline orientation。
+    """
+    p = Path(img_path)
+    img = cv2.imread(str(p))
+    if img is None:
+        return p
+
+    h, w = img.shape[:2]
+    # 横图宽 > 高 * 1.2，认为方向正确；否则旋转
+    if h > w * 1.2:
+        rotated = cv2.rotate(img, cv2.ROTATE_90_CLOCKWISE)
+        out = p.with_name(f"{p.stem}_rot.png")
+        cv2.imwrite(str(out), rotated)
+        return out
+    return p
+
+
+def enhance(img_path: str | Path) -> Path:
+    """轻度增强：灰度 + 自适应二值化（对手机拍的发票有帮助）"""
+    p = Path(img_path)
+    img = cv2.imread(str(p))
+    if img is None:
+        return p
+
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    # 弱增强：只对低对比度图做二值化
+    if gray.std() < 50:
+        binary = cv2.adaptiveThreshold(
+            gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 31, 10
+        )
+        out = p.with_name(f"{p.stem}_enh.png")
+        cv2.imwrite(str(out), binary)
+        return out
+    return p