An enterprise-grade AI skill that transforms unstructured documents into structured JSON data using advanced Vision Language Models.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install laiye-adp-doc-processing
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install laiye-adp-doc-processing using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Laiye Agentic Document Processing (ADP)?
Laiye Agentic Document Processing (ADP) is a high-performance extraction engine designed for Openclaw Skills users who need to digitize complex business documents. By combining Vision Language Models (VLM) and Large Language Models (LLM), this skill goes beyond simple OCR to understand the context and structure of over 10 different file formats, including PDFs, images, and Microsoft Office files.
This skill is specifically optimized for enterprise environments where accuracy is paramount. It provides field-level confidence scores and detailed metadata, allowing developers to build robust automation pipelines that handle everything from simple receipts to multi-page financial reports with minimal configuration.
Laiye Agentic Document Processing (ADP) Use Cases
- Automating accounts payable by extracting data from invoices and receipts.
- Digitizing purchase orders and sales contracts for ERP integration.
- Processing logistics and shipping documents to track supply chain line items.
- Converting unstructured financial statements into clean, analyzable data tables.
- Implementing high-confidence document verification for compliance workflows.
How Laiye Agentic Document Processing (ADP) Works
- The skill receives a document input via a secure file URL or a Base64 encoded string.
- It authenticates with the enterprise API using Openclaw Skills environment credentials.
- The document is processed through a VLM to identify layout, tables, and text elements.
- An LLM agent performs structured extraction based on the specific document type and business rules.
- The system returns a comprehensive JSON object containing extracted fields, data types, and confidence scores.
- For large documents, the skill utilizes an asynchronous task system to poll for results once processing is complete.
Laiye Agentic Document Processing (ADP) Setup
To integrate this skill into your Openclaw Skills environment, first secure your API credentials from the Laiye provider. It is highly recommended to use environment variables for credential management:
export ADP_ACCESS_KEY="your_access_key_here"
export ADP_APP_KEY="your_app_key_here"
export ADP_APP_SECRET="your_app_secret_here"
Configure your skill entry as follows:
{
"skills": {
"entries": {
"adp-doc-extraction": {
"enabled": true
}
}
}
}
Laiye Agentic Document Processing (ADP) Data Schema & Taxonomy
The extraction results are returned in a structured format designed for easy downstream consumption:
| Property | Type | Description |
|---|---|---|
field_key |
String | The technical name of the extracted field (e.g., total_amount) |
field_value |
String | The actual data extracted from the document |
field_type |
String | The data format: text, number, date, or table |
confidence |
Float | Accuracy score between 0.0 and 1.0 |
source_pages |
Array | The specific pages where the field was located |
Additionally, a metadata block provides details on total_pages, processing_time, and the model_used.
Laiye Agentic Document Processing (ADP) Advanced Features
- Support for asynchronous processing to handle high-volume or multi-page documents without timeouts.
- Adjustable
scaleparameters to improve extraction quality on low-resolution scans. - Version-specific configuration IDs to ensure consistent extraction logic across document updates.
- Recognition-only mode to retrieve layout and VLM data without full field extraction.
- Comprehensive error handling for document conversion and authentication failures.
SKILL.md
Loading
Related Openclaw Skills
Apple Watch Health Sync
A local health data pipeline that syncs Apple Watch metrics directly to your PC via a Python-based server and iOS automation.
ValueSider Superinvestor Data
A specialized tool for fetching and parsing real-time 13F portfolio holdings and trading activity of elite superinvestors from ValueSider.
Polymarket Trading
A professional-grade toolset for executing trades, managing orders, and analyzing liquidity on the Polymarket prediction platform.
Kalshi Trading Skill
A professional-grade interface for trading CFTC-regulated prediction markets on Kalshi via the kalshi-cli command-line tool.
LoveTago AI Dating Skill
A specialized skill that allows AI agents to autonomously register, match, and engage in meaningful conversations on the LoveTago dating platform.
LoveTago AI Dating Skill
A social integration skill that allows AI agents to autonomously participate in the LoveTago dating platform through registration, matching, and personality-driven conversation.
METADATA
Featured*
