Unstructured

Unstructured is an open-source-first platform that turns raw documents into clean, chunked data for vector stores and LLMs.

Overview

Unstructured provides the ingestion layer for AI stacks. Its open-source library parses PDFs, HTML, Word docs, images, and more into normalized elements, while its hosted platform adds managed partitioning, enrichment, and connectors that push clean data into vector databases and data warehouses. With tens of millions of downloads and broad framework support, it is one of the default ways teams get unstructured content into retrieval pipelines.

The library’s element-level output — titles, narrative text, tables, images — gives downstream chunking strategies real structure to work with. The hosted platform adds enrichment steps like entity extraction and embedding-ready partitioning, letting teams standardize ingestion across many sources.

Key Features

• Open-source parsing library for 30+ file types
• Managed partitioning and enrichment APIs
• Pre-built connectors for vector DBs and warehouses
• Chunking strategies tuned for embeddings and RAG
• OCR and table extraction for scans
• Enterprise deployment options including VPC

The community scale is meaningful: millions of downloads and broad framework support make Unstructured a safe, boring choice for the ingestion layer — exactly what production AI systems need. The hosted platform’s connectors standardize the path from file share to vector store.

Pricing

The open-source library is free. Hosted API plans are usage-based with a free developer tier and enterprise agreements for scale.

Who Is It For

AI engineers and data teams building retrieval pipelines who want a battle-tested ingestion layer they can self-host or run managed.

FAQ

Is Unstructured free to use?

Yes, the core library is free and open source; the hosted platform charges by usage beyond the free tier.

Does Unstructured offer an API?

Yes, the Unstructured API provides managed partitioning and enrichment with SDKs and connectors.

How does Unstructured compare to alternatives?

Its open-source foundation and connector ecosystem differentiate it from purely hosted parsers like Reducto.

Can Unstructured run in my own cloud?

Yes, the open-source library self-hosts anywhere, and the platform offers enterprise deployment options.

Does it handle images and scans?

Yes, OCR and image extraction are supported alongside native document formats.

Which vector databases are supported?

Pre-built connectors cover major vector stores and data warehouses.

Can I self-host everything?

Yes, the open-source library runs entirely within your own infrastructure.

Is Unstructured free for commercial use?

The open-source library permits broad use; hosted services are billed by usage.

How do I get started with Unstructured?

Install the open-source library for local parsing or sign up for the hosted API and follow the quickstart for your pipeline.