技能备份 - 2026-04-15 (40个技能)

skills/ocr-python/SKILL.md

@@ -0,0 +1,45 @@
+---
+name: ocr
+description: Optical Character Recognition (OCR) tool, supports Chinese and English text extraction from PDFs and images. Use cases: (1) extract text from scanned PDFs, (2) recognize text from images, (3) extract text content from invoices, contracts, and other documents
+---
+
+# OCR Text Recognition
+
+This skill uses PaddleOCR for text recognition, supporting both Chinese and English.
+
+## Quick Start
+
+### Basic Usage
+
+Perform OCR recognition directly on image or PDF files:
+
+```python
+from paddleocr import PaddleOCR
+
+ocr = PaddleOCR(lang='ch')
+result = ocr.predict("file_path.jpg")
+```
+
+## Dependency Installation
+
+Install dependencies before first use:
+
+```bash
+pip3 install paddlepaddle paddleocr
+```
+
+## Output Format
+
+Recognition results return JSON containing:
+- `rec_texts`: List of recognized text
+- `rec_scores`: Confidence score for each text
+
+## Typical Use Cases
+
+1. **PDF Scans**: Use PyMuPDF to extract images first, then OCR
+2. **Image Text Recognition**: Perform OCR directly on images
+3. **Multi-page PDFs**: Process page by page
+
+## Scripts
+
+Common scripts are located in the `scripts/` directory.

skills/ocr-python/SKILL.zh-CN.md

@@ -0,0 +1,45 @@
+---
+name: ocr
+description: 光学字符识别 (OCR) 工具，支持从 PDF 和图片中提取中英文文本。适用场景：(1) 从扫描版 PDF 提取文字，(2) 识别图片中的文字，(3) 提取发票、合同等文档的文字内容
+---
+
+# OCR 文字识别
+
+本技能使用 PaddleOCR 进行文字识别，支持中文和英文。
+
+## 快速开始
+
+### 基础用法
+
+直接对图片或 PDF 文件进行 OCR 识别：
+
+```python
+from paddleocr import PaddleOCR
+
+ocr = PaddleOCR(lang='ch')
+result = ocr.predict("file_path.jpg")
+```
+
+## 依赖安装
+
+首次使用前请安装依赖：
+
+```bash
+pip3 install paddlepaddle paddleocr
+```
+
+## 输出格式
+
+识别结果返回 JSON 格式，包含：
+- `rec_texts`: 识别出的文字列表
+- `rec_scores`: 每段文字的置信度分数
+
+## 典型使用场景
+
+1. **PDF 扫描件**：先用 PyMuPDF 提取图片，再进行 OCR
+2. **图片文字识别**：直接对图片进行 OCR
+3. **多页 PDF**：逐页处理
+
+## 脚本
+
+常用脚本位于 `scripts/` 目录下。

skills/ocr-python/_meta.json

@@ -0,0 +1,6 @@
+{
+  "ownerId": "kn7cpqbnatvf5664znnnamfg6s81jzv4",
+  "slug": "ocr-python",
+  "version": "1.0.0",
+  "publishedAt": 1771704541442
+}

skills/ocr-python/scripts/ocr.py

@@ -0,0 +1,111 @@
+#!/usr/bin/env python3
+"""
+OCR Text Recognition Script
+Usage: python3 ocr.py <file_path> [--output <output_file>]
+Supported: .pdf, .jpg, .jpeg, .png
+"""
+
+import sys
+import os
+import json
+import argparse
+
+def extract_images_from_pdf(pdf_path):
+    """Extract images from PDF"""
+    import fitz
+    
+    images = []
+    doc = fitz.open(pdf_path)
+    
+    for page_num in range(len(doc)):
+        page = doc[page_num]
+        page_images = page.get_images()
+        
+        for img_index, img in enumerate(page_images):
+            xref = img[0]
+            base_image = doc.extract_image(xref)
+            image_bytes = base_image["image"]
+            image_ext = base_image["ext"]
+            
+            output_path = f"/tmp/pdf_page{page_num+1}_img{img_index}.{image_ext}"
+            with open(output_path, "wb") as f:
+                f.write(image_bytes)
+            images.append(output_path)
+    
+    doc.close()
+    return images
+
+
+def ocr_file(file_path, output_path=None):
+    """Perform OCR recognition on file"""
+    from paddleocr import PaddleOCR
+    
+    # Initialize OCR
+    ocr = PaddleOCR(lang='ch', use_angle_cls=True)
+    
+    ext = os.path.splitext(file_path)[1].lower()
+    
+    if ext == '.pdf':
+        # PDF: extract images first, then recognize
+        images = extract_images_from_pdf(file_path)
+        all_texts = []
+        
+        for img_path in images:
+            print(f"Recognizing: {img_path}")
+            result = ocr.predict(img_path)
+            if result:
+                texts = result[0].get('rec_texts', [])
+                all_texts.extend(texts)
+        
+        # Clean up temporary images
+        for img_path in images:
+            try:
+                os.remove(img_path)
+            except:
+                pass
+        
+        final_texts = all_texts
+    else:
+        # Image: recognize directly
+        print(f"Recognizing: {file_path}")
+        result = ocr.predict(file_path)
+        
+        if result and len(result) > 0:
+            final_texts = result[0].get('rec_texts', [])
+        else:
+            final_texts = []
+    
+    # Output results
+    if output_path:
+        with open(output_path, 'w', encoding='utf-8') as f:
+            for text in final_texts:
+                f.write(text + '\n')
+        print(f"Results saved to: {output_path}")
+    else:
+        print("\n=== OCR Recognition Results ===\n")
+        for text in final_texts:
+            print(text)
+    
+    return final_texts
+
+
+def main():
+    parser = argparse.ArgumentParser(description='OCR Text Recognition Tool')
+    parser.add_argument('file', help='File to recognize (PDF or image)')
+    parser.add_argument('--output', '-o', help='Output file path (optional)')
+    
+    args = parser.parse_args()
+    
+    if not os.path.exists(args.file):
+        print(f"Error: File not found: {args.file}")
+        sys.exit(1)
+    
+    try:
+        ocr_file(args.file, args.output)
+    except Exception as e:
+        print(f"Error: {e}")
+        sys.exit(1)
+
+
+if __name__ == '__main__':
+    main()