open-webui/desktop: Open WebUI Desktop 🌐

Open WebUI Desktop 🌐

Summary

Open WebUI Desktop is a cross-platform native app, written in Svelte, that wraps the Open WebUI chat interface. It can either connect to a remote Open WebUI server or run models locally through a built-in llama.cpp engine. The project is marked Early Alpha in the README and is released under the AGPL-3.0 license.

What it is useful for

Open WebUI Desktop targets engineers who want the Open WebUI experience without a browser tab or a Docker setup. The README documents these uses:

• Connecting to one or more Open WebUI servers and switching between them from the sidebar.
• Running models locally. The README states the built-in llama.cpp engine handles model downloads and execution, and that nothing leaves the machine.
• Working with AI across the operating system. A floating chat bar called Spotlight is triggered with Shift+Cmd+I on macOS or Shift+Ctrl+I on Windows and Linux, and you can drag to capture a screenshot of any screen region.
• System-wide voice input through a push-to-talk shortcut that records, transcribes, and sends the text to the active chat.

Documented system requirements split into two modes. For local models the README lists 5 GB or more of disk and 16 GB or more of RAM, noting that a 7B model needs roughly 8 GB and a 13B model needs roughly 16 GB. Remote-only use needs about 500 MB of disk and 4 GB of RAM. Supported operating systems are macOS 12 or newer, Windows 10 or newer, and modern Linux with glibc 2.28 or newer.

How engineers can use it

The README ships installers for macOS (Apple Silicon and Intel .dmg), Windows (x64 and ARM64 .exe), and Linux (x64 .AppImage, .deb, .snap, .flatpak, plus ARM64 .AppImage and .deb). Internet is required on first launch; after that, the app runs offline. The flow is to download the installer, launch the app, and either point it at a server or enable local inference from settings. Auto-updates are handled in the background.

For contribution, the README documents the standard Node setup:

npm install
npm run dev

Two caveats from the README are worth noting. First, the project is tagged Early Alpha, so behavior may break and changes are frequent. Second, local inference carries the RAM and disk costs above, and machines that fall short should stay on the remote-connection mode.