Ollama
The Ollama nav item (route /models) manages local model runtimes across the Home. It is visible to admins and to users with the inference role.
Picking a Device
Section titled “Picking a Device”The left rail lists Devices with their Ollama reach state and pulled-model counts. Selecting one scopes the inspect pane.
Status and models
Section titled “Status and models”The right pane shows:
- A status chip with a Recheck button to re-detect the runtime now
- The Completions endpoint for this Home (
/v1/chat/completions) with Copy - Pull by name as a fallback when you already know the model slug
- Filter catalog to match names already pulled on the Home
- The model table on that Device (Model · Size · Device · State) with per-row Start; starting a model that lives on several Devices offers a Device picker limited to Devices that have it
For installing Ollama itself, pulling models, and how completions route through the Home, read the Ollama guide.
Completions
Section titled “Completions”The inspect pane shows the Home completions URL (/v1/chat/completions). Inference keys live under Settings → API Keys.
Export
Section titled “Export”The More menu holds Export JSON, which dumps the current ollama ps output. It is hidden when no Device is reachable.