- please check limitations before continuing.
You can run vicinus AI by just running one command!
brew install mathobsession/tap/vicinus-aionce installed, you can update it using:
brew reinstall vicinus-aito run the app use:
vicinus-aiOn first launch it copies itself to /Applications/VicinusAI.app so it lives
there permanently — double-click it from Finder anytime afterwards, or keep
running vicinus-ai to stay up to date. The window shows a Start servers
button; once the stack is up, the console UI renders inside the window via
WebKit — locked to 127.0.0.1:5001, external links open in your default
browser. Quitting the window stops all servers.
vicinus-ai-cliRuns the same stack without the native window — use this in scripts or terminal-only workflows.
vicinus-ai-cli --devIn dev mode the web backend owns the inference server, which unlocks Reload model and Save & reload runtime flags buttons in the settings panel (toggle Dev mode in the sidebar). Reloading restarts TurboFieldfareServer with the selected runtime flags (context window, expert-cache slots, prefill, rdadvise). Sampling parameters — temperature, top-k, top-p, seed — always apply per message, no reload needed.
Vicinus AI is a homebrew tap of the repository turbo-fieldfare. It is made to simplify the use of the repository's built in server and wrap it around a calming, good looking UI.
Though it is still in beta 0.1.1 the product is relatively stable. We have a scheduled build release timeline of about a new update every two days in the first two weeks and then eventually shift to a monthly release.
A new repository for Vicinus BETA will be created by the end of August 2026 for the latest updates.
To learn more about the architecture check turbo-fieldfare.
This repository supports only Apple Silicon MacBooks and installing it on intel based Macs might cause issues.