DocMind document parsing skill (alibabacloud-docmind-parse)

Updated at:

alibabacloud-docmind-parse is an Agent Skill for DocMind, Alibaba Cloud's intelligent document parsing service. This skill converts various document types, including PDF, Word, PPT, Excel, and images, into structured Markdown, JSON, or HTML formats.

Introduction

alibabacloud-docmind-parse is an Agent Skill for DocMind, Alibaba Cloud's intelligent document parsing service. Once installed, you can use the skill to parse various documents, converting files including PDF, Word, PPT, Excel, and images into structured Markdown, JSON, or HTML.

Core capabilities

Capability

Description

Multi-format support

PDF, Word, PPT, Excel, images, Markdown, HTML, EPUB, TXT, and more.

Structured output

Preserves layout information, such as heading levels, tables, images, and paragraphs.

Multiple enhanced modes

VLM (visual large model), LLM, DIGITAL (digital document), OCR, and AUTO.

Dual-channel access

Combines the free V2 API Direct Connection and Alibaba Cloud POP, automatically routing requests to the optimal channel.


Installation

npx skills add https://github.com/aliyun/alibabacloud-aiops-skills --skill alibabacloud-docmind-parse --full-depth

After installation, the V2 free mode works by default and requires no configuration.

To use the Alibaba Cloud POP mode (which provides a free quota of 3,000 pages per month), you must configure your Alibaba Cloud credentials. If you are not using environment variables, see this reference:

export ALIBABA_CLOUD_ACCESS_KEY_ID=your_id
export ALIBABA_CLOUD_ACCESS_KEY_SECRET=your_sk # Visit the [DocMind console](https://docmind.console.aliyun.com/doc-overview) to activate the service.

Usage

After installing the skill, instruct the Agent to parse a document using natural language.

Skill commands

/alibabacloud-docmind-parse parse ./test.pdf and summarize the content
/alibabacloud-docmind-parse parse ./test.pdf and convert it to Markdown

Basic examples

Parse this document: https://example.com/report.pdf
Convert ./contract.docx to Markdown
Parse the first 5 pages of this PDF

Parsing strategies

Use VLM mode to parse this scanned file: ./scanned.pdf
Parse report.pdf and output in JSON format.
Parse this document and save it to ./result.md: https://example.com/doc.pdf

Complex document processing

Parse this report and output the tables in Markdown format
Parse this PDF and extract the content of the header and footer.
Parse the digital document in DIGITAL mode, processing only pages 3–8.

Two Access Modes

Mode

Description

Free quota

File size limit

V2 API Direct Connection (Default)

Use without an Alibaba Cloud account.

Limited daily free quota

5 MB

Alibaba Cloud POP

Uses Alibaba Cloud credentials.

3,000 pages/month

150 MB

The V2 API Direct Connection provides a daily 100,000-credit free quota. The consumption standards for various resources are as follows, and the format types are described in the text below:

File type

Cost

Documents, images, others

Per page 50 credits (or per file)

Video

Per second 5 credits

Audio

Per second 1 credit

The Agent automatically selects the optimal channel:

  • Alibaba Cloud credentials detected → Uses POP mode.

  • No credentials found → Uses the V2 free mode.

  • POP failure → Automatically downgrades to the V2 free mode.

  • File exceeds the V2 size limit → Automatically upgrades to POP mode.


Enhanced mode description

Mode

Applicable Scenarios

AUTO

Automatic detection. Recommended for most scenarios.

VLM

For scanned documents, complex layouts, and handwriting. Enhanced using a visual large model.

LLM

For semantic understanding and content enhancement.

DIGITAL

For digital documents, such as PDFs with existing text layers and Word documents.

OCR

For text recognition and extraction from images.


Supported File Formats

  • Documents: PDF, Word (doc/docx), PPT (ppt/pptx), Excel (xls/xlsx/xlsm)

  • Images: JPG, JPEG, PNG, BMP, GIF

  • Other: Markdown, HTML, EPUB, MOBI, RTF, TXT

  • Audio and video (POP mode only): MP4, MKV, AVI, MOV, WMV, MP3, WAV, AAC


Output format

Format

Description

Markdown

The default format. It preserves structures such as headings, paragraphs, and tables.

JSON

Raw layout data that includes detailed information such as type, coordinates, and page numbers.

HTML

Suitable for embedding in web pages.


Precautions

  • The file upload limit for the V2 free API mode is 5 MB. For large files, use POP mode (150 MB).

  • Parsing audio and video files is supported only in POP mode.

  • Private URLs (resources that require authentication) cannot be parsed directly. Download the resource locally before parsing.

  • Used up your free quota? Visit the DocMind console to activate the service.