Skip to content
All work
AI & LLM2025

ULTRON3 — Offline Desktop Assistant

A GTK3 desktop assistant wired straight to llama.cpp, so inference happens locally and nothing leaves the machine.

PythonGTK3llama.cppLinux
Source

The problem

Every desktop assistant worth using wanted an API key, a network connection, and a per-token bill. None of that should be necessary to talk to a model on your own machine.

How it works

A native GTK3 interface talks directly to llama.cpp, running quantised models on local hardware with no cloud round-trip in the loop.

Because inference is local, the assistant keeps working offline and no conversation ever leaves the machine.

Outcomes

  • Fully offline local inference in a native Linux desktop app.
  • No API keys, no network dependency, no per-token cost.