VoidDrop is a notes app which removes the need for organisation by using AI embeddings.
demo.mp4
Download the latest build for your operating system on the Releases page.
When you open the app for the first time, it'll take a minute to download the models. After that, the app will load much quicker.
Note for macOS Users
Because VoidDrop is an independent project without an official Apple Developer Certificate, macOS will flag the app as "damaged" and refuse to open it, despite it being safe.
After dragging the app into your Applications folder, run this command in the terminal:
xattr -d com.apple.quarantine /Applications/VoidDrop.appNote: For macOS, VoidDrop is compiled exclusively for Apple Silicon (M1 chips and newer).
Legacy Intel Macs are not supported – the underlying ONNX Runtime no longer provides prebuiltx86_64-apple-darwinbinaries, and Rust itself has demoted the target to Tier 2.
- Backend: Rust + Tauri
- Frontend: React + TypeScript + TailwindCSS + Framer Motion + Three.js
- Storage: SQLite with sqlite-vec for vector storage
- Models: Local sentence embedding models - BGE (384d) + CLIP (512d) for images
- A central Black Hole to drop files and type out notes
- Efficient Semantic Search to find fragments based on meaning
- An Archive screen containing every fragment saved
- A visualisation of the Latent Space, showing every fragment as a star in a galaxy
- No internet connection required after launching the app for the first time
The Latent Space visualiser uses 2 hardcoded vectors to show a 2D "snapshot" of the embeddings
I'm the kind of person who dumps everything into Notes, reminder apps, or onto the desktop and hope I can find it later.
So I made VoidDrop – a place where I can do exactly that, except I can now retrieve things in seconds instead of digging through chaos.
This project is licensed under the MIT License
