unstructured

Why we included this project

Most RAG pipelines break on the ingestion step long before the model ever sees a query. Unstructured sits between raw files and your vector store, turning PDFs, Word documents, HTML pages, and a long tail of other formats into structured JSON or CSV elements. The partitioning logic reads document structure instead of flattening everything to plain text, which matters when your knowledge base is built from contracts, manuals, or internal wikis. It also covers the unglamorous parts of ingestion, like OCR for scanned pages and layout-aware extraction, so the text you embed is actually usable. That saves you from writing and maintaining a separate parser for every file type your users upload.