Glimt

Capture ideas the moment they happen. Entirely on your device.

Glimt is a lightweight desktop app for capturing ideas in seconds. Press a global hotkey, type or speak your thought, and it's saved instantly to a local database. There are no accounts, no cloud sync, and no external services. Everything stays on your machine, including voice transcription and search.

Features

• Instant capture. Press a hotkey from any app to open a small floating editor. Type your idea and hit Enter. Done in under two seconds.
• Voice input. Speak instead of typing. Transcription runs locally and supports 99 languages. Start recording with a hotkey without even opening the capture window.
• Semantic search. Find ideas by meaning, not just exact words. Search "marketplace for freelancers" and find a note from last month about "Upwork takes too big a cut, there's room for something leaner."
• AI-generated titles. Short, descriptive titles are generated for each idea in the background, entirely on-device.
• Markdown export. Auto-export ideas as .md files with YAML frontmatter. Works with Obsidian, Logseq, or any markdown-based tool.
• Timeline view. Browse ideas grouped by day with inline editing, archive, and delete.
• Command palette. Quickly navigate and act with Ctrl+K.
• System tray. Minimizes to tray on close. Stays out of your way until you need it.
• Customizable shortcuts. Change the capture and recording hotkeys in settings.

Download

Grab the latest release for your platform from GitHub Releases:

• Windows .exe installer
• Linux .deb or .AppImage
• macOS .dmg (Apple Silicon)

Build from source

Prerequisites

• Bun
• Rust
• Tauri v2 prerequisites for your platform

Steps

git clone https://github.com/varsan-g/glimt.git
cd glimt
bun install
bun run tauri dev

The first launch downloads default AI models (~40 MB). After that, startup is instant.

To produce distributable installers:

bun run tauri build

Dev commands

Command	Description
bun run tauri dev	Run the full desktop app
bun run dev	Vite dev server only (no Tauri)
bun run lint	ESLint
bun run format	Prettier
bun run test	Vitest
cargo clippy --manifest-path src-tauri/Cargo.toml	Lint Rust
cargo fmt --manifest-path src-tauri/Cargo.toml	Format Rust

Tech overview

Glimt is built with Tauri v2 (Rust backend, webview frontend), React 19, TypeScript, and Vite. The UI uses shadcn/ui on Tailwind CSS v4, with TipTap as the rich text editor and SQLite for local storage.

All AI inference runs on-device through Transformers.js in Web Workers, keeping the UI responsive. No data is sent to external services.

Models (all ONNX, all local):

Purpose	Model	Details
Speech-to-text	Whisper (multiple sizes)	~40 MB default, larger options for higher accuracy
Embeddings	Multilingual E5 Small	384 dimensions, 100+ languages
Title generation	SmolLM2-360M-Instruct	Lightweight on-device text generation

Contributing

Contributions are welcome. Please ensure lint and format checks pass before opening a pull request:

bun run lint
bun run format:check
cargo clippy --manifest-path src-tauri/Cargo.toml
cargo fmt --manifest-path src-tauri/Cargo.toml --check

If you find a bug or have a feature request, open an issue.