Ollama

Ollama runs open-weight models on your own machine. It is seeded in ivx/ai Chat, so it is already listed under Settings → Providers without being added.

Base URL

http://localhost:11434

No API key is needed.

Ollama Cloud

Ollama also offers a paid subscription, Ollama Cloud, which runs models on their servers. Point a provider at:

https://ollama.com

It requires an API key from your Ollama account.

Allowing browser requests

Ollama refuses cross-origin requests by default. Start it with your origin allowed:

OLLAMA_ORIGINS='http://localhost:5173' ollama serve

The desktop app includes a local proxy and needs no such configuration.

Choosing a model

Open the model picker and use Fetch models. Endpoints with no /models route are fine — choose Type a model name instead, and hand-typed names are remembered per provider.