DocuBrowser v0.9.0 introduces a local, AI-powered search engine for documents, aiming to organize and make searchable diverse file types including PDFs, ebooks, and Word documents. The tool operates entirely offline, requiring no internet connection, accounts, or API keys, and emphasizes user data privacy. It combines traditional keyword search with AI-driven semantic similarity and synopsis generation, utilizing local models like Ollama, nomic-embed-text, and dolphin3.

Its hybrid search mode blends 70% semantic and 30% keyword relevance, re-ranking results for improved accuracy. Users can generate on-demand AI synopses for documents, which are then cached. DocuBrowser supports numerous file formats, employing libraries like `pdfplumber` and `python-docx` for extraction. It also includes a PII scanner to detect and blacklist sensitive information.

Packaged for Linux, Windows, and macOS, DocuBrowser prioritizes a stable interface. Performance metrics cite typical search latency under 150ms. The tool’s architecture uses SQLite FTS5 for keyword indexing and Ollama for embeddings and synopsis generation, ensuring all processing happens client-side. This approach offers a cost-effective and secure alternative to cloud-based document management solutions.