Chris wrote:
Short answer: yes. That Mac Mini counts as a normal Linux server — SparkBox installs on standard Ubuntu/Debian Linux hosts (Ubuntu server is an explicitly supported family), and your 16GB RAM + 500GB drive clear the requirements comfortably. Nothing about it being a 2012 Mac matters once you have a plain Ubuntu server on it. Two different things sit behind “the AI”, and the Mini handles them differently: 1. Tom AI (the assistant built into your dashboard) — the heavy lifting happens on our side; the box just relays requests. Any supported install runs it without fuss. 2. Local AI (the Local AI module — Ollama + Open WebUI) — runs a model on the box itself. The module’s own guidance: 3B models are comfortable on 8GB of RAM, 7B on 16GB, 13B+ wants 32GB+. Your Mini’s 16GB puts it in the small-to-7B range, and the default model is a 3B, so you’d be fine. The honest caveat: a 2012 Mini has no Nvidia GPU, so inference is CPU-only — a 3B–7B model works, but answers come at a few tokens per second rather than GPU speed. The 500GB HDD carries plenty of models (they’re roughly 2–8GB each); an SSD would mainly just speed up model loading. On the license: correct, this box would use one of your activations — that’s your call. If you mainly want Tom AI plus small local models, the Mini is a reasonable “AI-only” box. If you ever want the large models (13B+) or near-instant replies, that’s the case for more RAM or a GPU box instead.