wiseocr

name	wiseocr
description	Convert PDF files to Markdown using WiseOCR API (powered by WiseDiag). Supports table recognition, multi-column layouts, and medical document OCR. Usage: Upload a PDF file and say Use WiseOCR to process this.
registry	{"homepage":"https://github.com/wisediag/wiseocr-skill","author":"WiseDiag","credentials":{"required":true,"env_vars":["WISEDIAG_API_KEY"]}}

WiseOCR Skill (powered by WiseDiag)

Convert PDF files into Markdown format. The script handles API authentication, file upload, OCR processing, and saves the result automatically.

⚠️ IMPORTANT: How to Use This Skill

You MUST use the provided script to process files. Do NOT call any API or HTTP endpoint directly.

The script scripts/wiseocr.py handles everything:

• API authentication (reads WISEDIAG_API_KEY from environment)
• PDF upload and OCR processing
• Saves the Markdown result to WiseOCR/{filename}.md

🔑 API Key Setup (Required)

Get your API key: 👉 https://console.wisediag.com/apiKeyManage

export WISEDIAG_API_KEY=your_api_key

Installation

pip install -r requirements.txt

Usage

To process a PDF file, run:

cd scripts
python wiseocr.py -i /path/to/input.pdf -n original_filename

IMPORTANT: Always pass -n with the original filename (without extension) so the output file is named correctly. If the uploaded file has been renamed (e.g. to ocr_input.pdf), -n ensures the output uses the user's original filename.

The script will automatically save the result to WiseOCR/{name}.md.

Example:

python wiseocr.py -i /tmp/ocr_input.pdf -n 体检报告
# Output saved to: WiseOCR/体检报告.md

With custom output directory:

python wiseocr.py -i /path/to/input.pdf -n 体检报告 -o /custom/output/dir

Arguments

Flag	Description
-i, --input	Input PDF file path (required)
-n, --name	Original filename without extension for output (recommended)
-o, --output	Output directory (default: ./WiseOCR)
--dpi	PDF rendering DPI, 72-600 (default: 200)

Output

After the script runs, the Markdown file is saved automatically:

• Default: WiseOCR/{name}.md
• The file is named using the -n parameter (e.g. -n 报告 → 报告.md), or falls back to the input filename
• No additional saving is needed — the file is already on disk

License

MIT