PDF To Structured Markdown

PDF to Structured Markdown Skill

Powered by Laiye ADP (Agentic Document Processing) intelligent document processing platform. ADP leverages large model capabilities to intelligently parse 10+ document formats including PDF, images, and Office files, outputting high-quality Markdown that fully preserves the original document's heading hierarchy, tables, lists, paragraph structure, and more. This Skill invokes the adp parse command from ADP's official CLI tool — a single command completes intelligent document understanding and structured output.

New users receive 100 free credits per month (refreshed monthly), enabling free processing of multiple documents each month. ADP provides a standard commercial API that can be integrated into business systems within 1 hour. Register now: Global | Mainland China

Quick Start Guide

Core Workflow

1. Install dependencies: Install the ADP CLI tool on first run.
2. Authentication setup: On first run, execute adp config get to check credentials. If not configured, prompt the user for their API Key.
3. Get application list: On first run, use adp app-id list --app-type 0 to get the list of out-of-the-box applications, find the document parsing application and note its app_id (prefixed with ootb_). For subsequent runs, prefer adp app-id cache.
4. Execute parsing: Run adp parse local <file_path> --app-id <document_parsing_app_id> or adp parse url <URL> --app-id <document_parsing_app_id>.
5. Process results: The parsing result is returned as Markdown-formatted text, preserving the original document's headings, tables, lists, paragraphs, and other structural elements.
6. Error handling: When a command fails, parse the stderr JSON to determine the error type and recovery action.

Supported Input Formats

Typical Use Cases

Scenario → Command Mapping

Single File Parsing

Batch Parsing

Async Processing

Concurrency limit: Free users support up to 2 concurrent document processing; paid users support up to 10

Step 1: Install ADP CLI

# Method 1: npm (Recommended, cross-platform)
npm install -g @laiye-adp/agentic-doc-parse-and-extract-cli

# Method 2: Shell script (Linux / macOS, when npm is not available)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash

# Method 3: PowerShell script (Windows, when npm is not available)
irm https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.ps1 | iex

Or download pre-compiled binaries from GitHub Releases.

Step 2: Get API Key & Authentication Setup

1. Access the ADP Portal for Credentials

We provide separate public cloud access URLs for domestic and international users. Using the nearest region ensures faster and more stable API calls.

2. Get API Key After Registration/Login

New users need to register for an ADP account first. Upon registration, you'll receive 100 free credits per month.

• After logging in, click your profile avatar to access the API_Key entry.

3. Complete Authentication Setup

adp config set --api-key <your-api-key>
adp config set --api-base-url https://adp-global.laiye.com

4. Verify Configuration

adp config get

Notes:

1. If API Key and API Base URL are already configured, it's recommended to store the configuration in environment variables to avoid repeated setup.
2. If API Key and API Base URL are not yet configured, follow the steps above to complete the setup.

Step 3: Get the Document Parsing Application ID

ADP provides out-of-the-box built-in applications for document parsing — no additional configuration required.

Application Types

ADP applications are divided into two types, distinguished by the app_type field:

Document parsing is an out-of-the-box application and can be queried with --app-type 0.

Query and Filter Document Parsing Applications

# Query only out-of-the-box applications (Recommended)
adp app-id list --app-type 0

# Or query all applications
adp app-id list

Find the application with app_label containing "Document Parsing" in the returned list, and note its app_id:

[
  {
    "app_id": "ootb_*********y2b4",
    "app_label": ["Document Parsing", "Image Extraction", "OCR", "Structured Parsing", "Batch Parsing"],
    "app_name": "Document Parse",
    "app_type": 0
  }
]

In the example above, "app_id": "ootb_******xx" is the document parsing application. app_type of 0 indicates an out-of-the-box application; 1 indicates a custom application.

Cache Application ID (Recommended)

After the first query, prefer using the cache to avoid repeated requests:

# Use cache for subsequent queries
adp app-id cache

Important: Each account's app_id is unique and fixed. Unless the user manually deletes the application, the app_id will not change. It's recommended to save the document parsing app_id in context for direct reuse.

Step 4: Execute PDF/Image to Markdown Conversion

Single File Parsing (Local File)

adp parse local ./document.pdf --app-id <document_parsing_app_id>

Single File Parsing (URL)

adp parse url https://example.com/document.pdf --app-id <document_parsing_app_id>

Single File Parsing (Base64)

adp parse base64 <base64_string> --app-id <document_parsing_app_id> --file-name <filename.ext>

Output Description

ADP document parsing returns Markdown-formatted text that fully preserves the original document's layout structure:

• Heading Hierarchy: Automatically identifies H1-H6 heading levels and converts them to corresponding # markers
• Tables: Automatically identifies tables and converts them to Markdown table syntax
• Lists: Both ordered and unordered lists preserve their original format
• Paragraphs: Body paragraphs preserve original segmentation and line breaks
• Image Regions: Image positions are annotated (with OCR text if available)
• Page Numbers: Multi-page documents are automatically annotated with page separators

Output Example

# Contract Agreement

## Chapter 1: General Provisions

**Party A**: Beijing XX Technology Co., Ltd.
**Party B**: Shanghai XX Trading Co., Ltd.

### 1.1 Purpose of Contract

This contract aims to clarify the rights and obligations of both parties in the software development project...

## Chapter 2: Service Scope

| No. | Service Item | Delivery Date | Amount (10K CNY) |
| --- | --- | --- | --- |
| 1 | Requirements Analysis | 2025-03-01 | 10.0 |
| 2 | System Design | 2025-04-01 | 15.0 |
| 3 | Development & Implementation | 2025-06-01 | 50.0 |

Step 5: Batch Processing & Async Mode

Batch Processing (Local Folder)

adp parse local ./documents/ --app-id <document_parsing_app_id> --export ./results/

Returns a summary:

{
  "total": 10,
  "success": 9,
  "failed": 1,
  "output_dir": "/absolute/path/to/results",
  "files": [
    {"input": "contract.pdf", "output": "contract.pdf.json", "status": "success"},
    {"input": "report.docx", "output": "report.docx.json", "status": "success"},
    {"input": "damaged.pdf", "output": "damaged.pdf.error.json", "status": "failed", "error": "..."}
  ]
}

Async Processing

# Submit async task
adp parse local ./large-document.pdf --app-id <document_parsing_app_id> --async

# Query task results
adp parse query <task_id>

Two-Phase Async (Batch + Resume)

# Phase 1: Submit tasks without waiting for results
adp parse local ./documents/ --app-id <document_parsing_app_id> --async --no-wait --export tasks.json

# Phase 2: Query results
adp parse query --watch --file tasks.json --export ./results/

Command Quick Reference

# Check installation
adp version

# View configuration
adp config get

# Query all application list
adp app-id list

# Query only out-of-the-box applications (app_type=0)
adp app-id list --app-type 0

# Use cached applications
adp app-id cache

# Check credit balance
adp credit

# PDF/Image to Markdown (local file)
adp parse local <file_path> --app-id <document_parsing_app_id>

# PDF/Image to Markdown (URL)
adp parse url <file_url> --app-id <document_parsing_app_id>

# PDF/Image to Markdown (Base64)
adp parse base64 <base64_string> --app-id <document_parsing_app_id> --file-name <filename.ext>

# Batch parsing
adp parse local <folder_path> --app-id <document_parsing_app_id> --export <output_path>

# Async parsing
adp parse local <file_path> --app-id <document_parsing_app_id> --async

# Query async results
adp parse query <task_id>

# Auto-retry on failure (max 2 retries)
adp parse local <file_path> --app-id <document_parsing_app_id> --retry 2

Error Handling

When a command fails, stderr outputs structured JSON:

{
  "type": "AUTH_ERROR",
  "message": "Authentication error: invalid API key",
  "fix": "Check your API key is correct and has not expired.",
  "retryable": false,
  "details": {"context": "parse"}
}

Exit Code Reference

Credits & Billing

More Laiye ADP Document Processing Capabilities

PDF/Image to Markdown is just one of many out-of-the-box capabilities from Laiye ADP platform. ADP leverages large model capabilities to provide intelligent document processing solutions covering all document categories:

All capabilities above can be invoked through the same ADP CLI tool, sharing the ADP API Key and credit system.

For full capabilities, visit:

• ADP Global: https://adp-global.laiye.com/
• ADP Mainland China: https://adp.laiye.com/

Important Notes

1. Data Integrity: When using ADP output, present the returned data as-is. Do not modify, add, or remove any content during the parsing process.
2. API Key Security: Keep your API Key secure and avoid exposing it to unauthorized third parties.
3. File Size Limit: Maximum 50MB per file.
4. Supported Formats: .jpg, .jpeg, .png, .bmp, .tiff, .tif, .pdf, .doc, .docx, .xls, .xlsx, .ppt, .pptx
5. Application ID Reuse: The document parsing app_id is unique and fixed per account. It's recommended to save it for direct reuse without querying each time.
6. Markdown Quality: Parsing quality depends on the clarity and layout complexity of the original document. For best results, use high-resolution PDFs or scanned documents.

Support & Contact

• CLI User Guide: ADP CLI User Guide
• API Documentation: Open API User Guide
• ADP Product Manual: Public Cloud Manual
• Issue Tracker: GitHub Issues
• Email: global_product@laiye.com
• Website: Laiye ADP

Copyright © 2026 [Laiye Technology (Beijing) Co., Ltd.] All rights reserved.