projects
A couple of tools I build and use daily. Both came out of running local language models on Apple Silicon and wanting the machine to be more legible than it was out of the box. Source is on GitHub.
llamactl
A command-line process manager for local language models. It shows what is loaded, what is serving traffic, and what is sitting idle, and it starts, stops, and pins models without a pile of open terminals.
brew install gregmundy/tap/llamactlWhy I built it: Running four or five models by hand on one machine got old fast. I wanted the ps and systemctl I already had in my head, but for models.
llamavm
Lightweight VM tooling that gives each local model its own isolated sandbox, so a runaway process or a bad config cannot take the whole machine down with it.
brew install gregmundy/tap/llamavmWhy I built it: llamactl made it easy to run many models at once. llamavm is what keeps one of them misbehaving from ruining the rest of my afternoon.