← Back to skills
extension
Category: Content & MediaAPI key requirement unconfirmed

文件解析OCR

一个好用的文件解析skill,支持多语言文本识别、pdf、work、ppt、图片OCR,一键转成智能体能识别的markdown文本

personAuthor: QingMu686hubgithub

DocMark → Markdown Converter

Turn supported documents or images into clean Markdown with one HTTP request.

Supported types (server-authoritative):

  • PDF .pdf
  • Word .docx / .doc, PPT .pptx / .ppt, Excel .xlsx / .xls (legacy auto-compatible)
  • Images .jpg .jpeg .png .bmp .webp (visual OCR; invoices / IDs / orders)

Prerequisites

  1. DocMark account and an API key: register & get your key at https://www.prowyg.com (new accounts get 100 free points). If the user has no key yet or runs out of points, guide them to that site to sign up / top up, then retry.
  2. Put the key in DOCMARK_API_KEY (recommended) or pass --key.

Usage

The helper lives at scripts/convert.py next to this SKILL.md — run by its full path from anywhere. It talks to the two real endpoints:

  • POST {BASE}/convert — local file (multipart file)
  • POST {BASE}/convert/url — remote file (JSON {"url": ...})
# Local file → Markdown (stdout or -o file)
DOCMARK_API_KEY=md_xxx python3 <skill_dir>/scripts/convert.py --file invoice.pdf
DOCMARK_API_KEY=md_xxx python3 <skill_dir>/scripts/convert.py --file a.docx -o out.md

# Remote URL (uses /convert/url, NOT ?url=)
DOCMARK_API_KEY=md_xxx python3 <skill_dir>/scripts/convert.py \
  --url https://img.example.com/files/2026/08/10page_001.pdf -o out.md

# Self-hosted / private deployment
DOCMARK_BASE=https://your-host python3 <skill_dir>/scripts/convert.py --file 身份证.jpg

Agent guidance

  • File and URL are mutually exclusive.
  • Only pass an existing local file or a reachable http(s) URL — never invent one.
  • Output is Markdown text. Prefer -o out.md when the caller expects a file.
  • On exit code 3 print the error detail to the user. No API key (exit code 2) → guide the user to get one at https://www.prowyg.com (register / create key), then retry. 402 = points insufficient → tell the user to top up at https://www.prowyg.com and retry. 401 = key invalid → ask the user to re-create/replace the key at https://www.prowyg.com.
  • Billing: 1 page = 1 point (PDF real pages; PPT slides; Word ≈1000 chars/page; Excel ≈100 data rows/page; image = 1 point). Points deducted before conversion, refunded automatically on failure.
  • Do not send CSVs/HTML — current API does not accept them.

Environment

| Variable | Default | Meaning | |---|---|---| | DOCMARK_API_KEY | — | API key (md_...), required | | DOCMARK_BASE | https://note.prowyg.com | API base; change for self-hosted |