Models
The free local option, swapping models, and bringing your own API key.
How does the free local option work?
Donna downloads a pinned release of the Ollama runtime into her own data folder, runs it herself, and pulls a default model (qwen2.5:3b, about 2 GB) with a progress bar. No admin rights, no separate install, and everything runs on your hardware; nothing ever leaves your machine.
If you already run your own Ollama, Donna detects and uses it instead of installing anything.
Can I use a different local model?
Yes. Any model your Ollama has pulled shows up in Settings → Model. Pull new ones with ollama pull <model> (or ask Donna once model management lands in Settings). Bigger models are smarter and slower; 3B–8B models are the sweet spot for most laptops.
How do cloud models work, and what do they cost?
Paste an API key from OpenAI, Anthropic, or Google in onboarding or Settings → Provider. You pay the provider directly for what you use; Donna adds no markup and has no subscription. Voice features (push-to-talk transcription and spoken replies) currently require an OpenAI key.
Where are my API keys stored?
In the brain’s credentials store: a file inside Donna’s data folder with owner-only permissions (0600), never in the repo, never sent anywhere except to the provider you chose. Google OAuth client credentials additionally use your OS keychain on the desktop side.