Skip to content

README.md

Latest commit

 

History

History
86 lines (62 loc) · 3.98 KB

README.md

File metadata and controls

86 lines (62 loc) · 3.98 KB

DocFinder

CI CodeQL License: AGPL v3 Python Release Downloads

DocFinder Logo

Local-first semantic search for your documents.
Supports PDF, Word (.docx), Markdown, and plain text files.
Everything runs on your machine — no cloud, no accounts, complete privacy.

DocFinder Demo

Features

  • Semantic search — find documents by meaning, not just keywords (PDF, DOCX, Markdown, TXT)
  • AI chat — ask questions about any document and get precise answers, powered by local Qwen models (automatically selects the best model for your hardware)
  • 100% local — your files never leave your machine
  • GPU accelerated — auto-detects Apple Silicon (Metal), NVIDIA (CUDA), AMD (ROCm)
  • Cross-platform — native apps for macOS, Windows, and Linux
  • Global shortcut (Experimental) — bring DocFinder to front from anywhere with a configurable hotkey

DocFinder Chat

Download

Platform Installer
macOS DocFinder-macOS.dmg
Windows DocFinder-Windows-Setup.exe
Linux DocFinder-Linux-x86_64.AppImage

macOS — open the DMG, drag DocFinder to Applications, then right-click → Open on first launch (Gatekeeper warning — normal for unsigned open-source apps).

Windows — run the installer; if SmartScreen appears choose More info → Run anyway.

Linux

chmod +x DocFinder-Linux-x86_64.AppImage && ./DocFinder-Linux-x86_64.AppImage

Run from Source

Requires Python 3.10+ and make.

git clone https://github.com/filippostanghellini/DocFinder.git
cd DocFinder
make setup   # create .venv and install all dependencies
make run     # desktop GUI
make run-web # web interface at http://127.0.0.1:8000

Runtime acceleration (auto)

DocFinder automatically selects the best available runtime on your machine:

  • NVIDIA: ONNX CUDA provider when available, otherwise PyTorch CUDA
  • AMD: ONNX ROCm provider when available, otherwise PyTorch fallback
  • Apple Silicon: optimized ONNX path
  • CPU-only hosts: ONNX or PyTorch CPU fallback

Indexing uses a balanced parallel parser strategy by default, selected automatically based on your machine resources.

Contributing

Contributions are welcome, feel free to open an issue or submit a pull request.

License

Licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).

DocFinder was originally released under the MIT License. Starting from version 1.1.1 the license was changed to AGPL-3.0 to comply with the PyMuPDF licensing requirements, as PyMuPDF itself is AGPL-3.0 licensed.