Introduction - Aryn Documentation

Aryn is an agentic Document Processing and Deep Analytics system, enabling you to query your document collections at scale with natural language. It can improve the accuracy (up to 3x) and efficiency (up to 10x) of manual workflows like processing, extraction, and getting insights from complex documents such as financial reports, legal memos, contracts, technical manuals, and RFPs with technical diagrams. Aryn serves as an Agentic Unstructured Data Warehouse, enabling search and complex analytics with natural language over complex documents at scale. Aryn does this using:

A reasoning model for “Deep Analytics” — complex, multi-step tasks to extract patterns and insights across large document sets (up to millions of documents).
State-of-the-art models for document parsing and ETL, including OCR, metadata and table extraction, and summarization on 30+ document formats (e.g. PDF, Word, PPT). This is also available stand-alone with DocParse.
Powerful Workspaces UI to navigate and query document sets for data scientists and analysts. Includes queryable Bookmarks for intermediate results, editing and inspecting query plans, and organizing results.
Customizable document ETL with Sycamore integration and APIs for building custom applications and workflows for end users.

You can also directly use DocParse, a building block of Aryn, to process your documents into structured JSON or Markdown. You can use it to prepare complex, unstructured data for retrieval-augmented generation (RAG) applications, document processing workflows, extracting content from documents (like tables), and semantic search systems. Also, it integrates with Sycamore for custom document ETL pipelines in Python.