A typed Python chauffeur for LandingAI’s document API

What it does

landingai-ade is essentially polished glue code for LandingAI’s Agentic Document Extraction REST API. It handles authentication, file uploads, retries, and response parsing so you can turn PDFs and images into chunked text, classified splits, or structured data without writing raw HTTP requests. Both synchronous and asynchronous clients are available, along with background job support for oversized documents.

The interesting bit

Instead of plain dictionaries, the SDK returns Pydantic models with .to_json() and .to_dict() helpers, and it lets you define extraction schemas as Pydantic classes. It also bundles an MCP server, meaning Cursor or other AI assistants can explore endpoints and make test requests on your behalf. If you need more concurrency, you can swap the default httpx backend for aiohttp.

Key highlights

• Fully typed SDK using Pydantic response models and TypedDict request parameters.
• Sync and async clients with built-in retries, exponential backoff, and swappable HTTP backends (httpx or aiohttp).
• Async job queue for large document processing.
• Schema-based extraction via Pydantic models.
• MCP server included for AI assistant integration.

Caveats

• Requires a LandingAI API key; the library is strictly a client for their hosted service.
• Default request timeout is eight minutes, suggesting the upstream API can be slow with large documents.

Verdict

Worth importing if you are already paying for LandingAI’s document extraction and want Pydantic types out of the box. Skip it if you need a self-hosted parser—this is a vendor-specific client, not a standalone engine.