Model Keys for Ollama

Model Keys puts your local Ollama models on your Stream Deck.

Loaded Model key

Shows the model Ollama is holding right now, and how much memory it takes.

Turns red when a model has spilled out of VRAM onto the CPU (the usual reason replies suddenly get slow).

Shows a "+1" badge when more than one model is loaded.

Press to unload everything and give the GPU back, for example before you start a game or a render.

Model Toggle key

One key per model, picked from the models you have installed.

Press to load it ahead of time so the first reply is fast. Press again to unload it.

Choose how long it stays loaded: 5 minutes, 30 minutes, 2 hours, or until you unload it.

Details

Works with Ollama on the same computer (by default, this computer on port 11434) or on another machine on your network.

Refreshes every 2–60 seconds, only while a key is on screen. Ten keys pointed at one server share a single check.

Mac and Windows.

Privacy Runs entirely on your computer. The plugin only talks to the Ollama address you set. No account, no analytics, no data collected.

Requires Ollama (free, ollama.com) and Stream Deck 7.1 or later. Ollama is a trademark of its owner; this plugin is not affiliated with or endorsed by Ollama.

First release: Loaded Model key (model, memory, GPU/CPU split, press to unload all) and Model Toggle key (load/unload one model, keep-alive choice).