linuxrebel/DocuBrowser

What it solves

DocuBrowse solves the problem of managing and searching through large, disorganized collections of local documents. It allows users to find specific information across various file formats using both literal keywords and conceptual meaning, without requiring the data to leave the local machine.

How it works

The application indexes local files (PDFs, Word docs, ebooks, etc.) into a SQLite FTS5 database for keyword search and uses Ollama with local AI models (nomic-embed-text for embeddings and dolphin3 for summaries) to enable semantic search and automated synopsis generation. It features a "Deep Links" system that allows users to jump directly to the matching passage within a document, highlighting the relevant text by page, line, or section.

Who it’s for

It is designed for users who have extensive local document libraries and prioritize privacy, as it runs entirely offline with no API keys, accounts, or internet connection required.

Highlights

• Hybrid Search: Combines fast keyword matching with AI-powered semantic similarity to find documents based on meaning.
• Deep Links: Pinpoints and highlights the exact passage inside a document that matches the search query.
• Local AI: Uses Ollama for on-device embeddings and synopsis generation, ensuring data privacy and zero per-query costs.
• Broad Format Support: Indexes a wide array of files including PDF, DOCX, PPTX, XLSX, ODT, EPUB, MOBI, Visio, draw.io, LaTeX, and various markup languages.
• PII Protection: Includes a post-ingest scanner that detects and blacklists documents containing sensitive patterns like SSNs or credit card numbers.