extend-hq/ui

Extend UI is a collection of open-source building blocks for apps that work heavily with documents. If you are building a product where users need to view PDFs, Word files, or Excel spreadsheets, upload and manage files, sign documents, or review data extracted from documents, these components give you a big head start. Instead of wiring together a basic file viewer from scratch, you get polished, production-tested pieces. The library offers 14 components covering things like PDF/DOCX/XLSX viewers, e-signature, file thumbnails, and a Finder-style file browser. A standout feature is the bounding box citation tool, which lets you show users exactly where in a document a specific piece of data was extracted from, useful for building human review workflows where someone checks an AI's work against the source file. There are also layout blocks for visualizing OCR (text recognition) output overlaid on the original document. To use these components, you install them into your project using the shadcn CLI, a popular tool for adding pre-built UI pieces to React apps. The key thing is that components are copied in as actual source code rather than installed as hidden dependencies. That means once a component is in your project, you can open it up and customize it however you want. The components assume your app already has basic UI primitives like buttons and dialogs, so they plug into your existing design system rather than imposing their own. The project is maintained by Extend, a company that processes millions of document pages in production. They built these components for their own internal use after finding existing libraries insufficient for document-heavy products, and they continue to maintain them as real production code. This makes Extend UI a good fit for founders or product teams building document processing tools, internal review dashboards, or AI-powered document agents where you need more than a generic file viewer. Because components ship as editable source code using React and Tailwind, they also work well with AI coding assistants that need to read and modify component code directly.