Unstructured.io

Unstructured.io is a library and API for turning messy real-world documents (PDFs, HTML, images, scans) into clean, structured pieces ready for embedding and chunking. It handles the ugly parts — tables, headers, multi-column layouts, OCR for images — and outputs partitioned elements you can feed straight into a RAG pipeline. Useful the moment your source data is anything messier than plain markdown.