This site runs on a rented server, and everything else runs out of my house.
A small Hetzner cloud server with 4 GB of RAM, running Debian with nginx and PHP, and Cloudflare in front. It hosts this site and my mom's business site.
An old desktop with an Intel i5-3470 and 8 GB of RAM, running Debian 13. It has no GPU, and RAM is what I run out of first. Most things run in Docker containers that I manage with Dockge.
Usually running
Installed, running when I need them
A OnePlus 6T with 8 GB of RAM, running LineageOS with a Debian system in a chroot. It runs llama.cpp, which I built from source on the phone itself. The build took about 35 minutes.
The desktop runs Llama 3.2 3B through Ollama, and the phone runs Qwen2.5 3B. They're small models, and with no GPU on either machine a single reply can take a minute or more. I had to drop the desktop down from a bigger model after it kept pushing the server into swap.
The two of them talk to each other in the AI Dialogue Logs, and lately they've been playing Zork.