This project provides an offline, local document search engine for Windows and macOS, enabling users to find information within files by content rather than just filename. It targets power users and professionals who manage large local document repositories and require fast, private search capabilities, offering significant time savings and enhanced data retrieval.

How It Works

Docufinder employs a hybrid indexing and search approach. It locally parses and indexes content from various file types, including Korean-specific formats like HWP, using SQLite FTS5 for keyword search and a vector database (usearch HNSW) for semantic understanding. An optional AI layer integrates with Gemini or OpenAI compatible APIs for natural language querying and summarization, sending only relevant text chunks for processing, ensuring data privacy.

Quick Start & Requirements

• Windows: Download the x64-setup.exe installer. If Microsoft Edge WebView2 runtime is not present (common in corporate/LTSC environments), install MicrosoftEdgeWebView2RuntimeInstallerX64.exe first. Requires 8GB RAM (16GB recommended) and 1GB disk space. Initial setup requires internet for model downloads (~420MB). Users must manually bypass security warnings during installation.
• macOS (Apple Silicon): Download the .dmg file from the Releases page. Requires macOS 11+, Apple Silicon (M1/M2/M3), 8GB RAM, and 1GB disk space. Intel Macs are not supported. Users must bypass Gatekeeper security warnings using xattr -dr com.apple.quarantine. Automatic updates are not supported.
• Links: Downloads are available via the project's GitHub Releases page.

Highlighted Details

• Extensive Format Support: Natively parses HWP (.hwpx, .hwp) without Hancom Office, PDFs (with OCR), Office documents (.docx, .pptx, .xlsx/.xls), images (.jpg, .png, etc. with OCR), text, and emails (.eml).
• Hybrid Search Capabilities: Offers keyword, filename (Everything-style), and semantic search modes, supporting advanced operators for precise queries (e.g., exact phrases, exclusions, path filtering, date ranges).
• Optional AI Features: Enables natural language Q&A and document summarization, displaying source document and page numbers. These features can be fully disabled for a pure offline experience.
• Advanced Document Handling: Includes original layout previews for HWPX, PDF, and HWP files, full-screen zoom, and side-by-side document comparison.

Maintenance & Community

Bug reports and feature suggestions are managed through the project's GitHub Issues page. No other community channels (like Discord or Slack) or specific roadmap details are mentioned in the README.

Licensing & Compatibility

The project is licensed under the Business Source License 1.1 (BSL 1.1). It is free for non-production use (development, testing, learning). Production or commercial use requires a separate license. The license automatically converts to Apache License 2.0 on April 15, 2030. Commercial license inquiries can be directed to ryuseungin@gmail.com.

Limitations & Caveats

Production and commercial use are restricted under the BSL 1.1 until April 2030, necessitating a separate license. macOS support is limited exclusively to Apple Silicon hardware. The application uses ad-hoc code signing, requiring users to manually bypass security warnings during installation and execution on both Windows and macOS. Automatic updates are not available for the macOS version.