Where your local Ollama is running. Default is correct for most installs.

Any model you have pulled via ollama pull <name>

How much conversation the model holds in memory at once. 8k is right for most hardware.

When set, cloud models become available as an upgrade. Ollama remains the default.