linuxrebel/docubrowser

DocuBrowse is a local search engine for your personal documents. You point it at folders full of PDFs, ebooks, Word files, spreadsheets, and notes, and it builds an index that understands meaning, not just exact words. You can search for something like a contract about a lease renewal and find it even if those exact words never appear in the file, then click a result to get an instant AI generated summary before opening the actual document. Everything runs entirely on your own computer using local AI models through a tool called Ollama. There is no internet connection required, no account to create, no API keys to manage, and no per query costs, since nothing is sent to an outside service. Under the hood it combines a fast keyword search built on SQLite with AI powered semantic similarity and summary generation. DocuBrowse can read an unusually wide range of file types: PDFs, Word documents, PowerPoint presentations, Excel spreadsheets, OpenDocument files, Visio diagrams, draw.io diagrams, PlantUML and Mermaid diagram files, XML and DocBook markup, reStructuredText, AsciiDoc, LaTeX, emails, RTF files, CSV and TSV tables, plain config style text, and ebooks in EPUB, MOBI, and AZW3 formats. Each format has its own extraction method, and scanned PDFs that are really just images get flagged separately since they cannot be searched as text. It also includes a privacy protection feature that scans newly added documents for sensitive personal information such as social security numbers, credit card numbers, bank account numbers, dates of birth, and passport numbers. Any document matching these patterns is removed from the index and blocked from being added again. The interface offers a dark and light theme, paginated results with an alphabetic index for quick navigation, a tag cloud for filtering by topic, and relevance scores shown as a percentage on every result. From a result you can open the file directly in its default application, copy its file path, or delete it from both disk and the index. The full README is longer than what was shown.