Set up
Local models with Ollama
Run open models on your own server. Prompts go to your hardware instead of a model provider, and there is no key to add.
What you need
An Ollama server that Auprimus can reach over the internet. Auprimus calls it from the cloud, so an address that only works inside your network will not connect. For a laptop, a Cloudflare Tunnel works; for a company server, put it behind a proxy with a public address.
Ollama has no login of its ownIf you expose it, put an authenticating proxy in front of it and add the proxy's header in the connection form.
Connect your server
Open Settings → API Keys → Local models
Enter the Server URL. If a proxy guards it, add the Auth header name and value, for example
Authorization.Test and connect
Auprimus lists the server's models before saving anything. When it answers, you see how many models are installed and the connection is saved.
In the model menu
| What you see | What selecting it does |
|---|---|
| Not connected | Opens Settings so you can connect your server. |
| Not installed | Pulls the model onto your server, with progress. |
| Installed | Selects it for the run. |
Models installed on your server but missing from the catalogue are added to the menu automatically.
NoteFor safety, Auprimus only connects to public addresses, never to private networks, and does not follow redirects. To change or remove the server, use Disconnect in the same section.
Something wrong or missing? Write to [email protected] about “Local models with Ollama”.Updated 26 Sep 2026