03 / On-device AI
Saventia AI
Room to think. On your device.
A focused AI app using Apple’s on-device Foundation Model. Conversations and preferences stay local.
iPhone & iPad
Product flow
Where a request goes
- Chat prompt → Foundation Models on device
- Chat history → SwiftData on device
- Weather lookup → Apple WeatherKit over a network
Local conversation first
I am building Saventia around Apple’s Foundation Models framework. When the device supports it, prompts and chat responses run on the device, while SwiftData keeps conversation history and preferences local.
Availability is a product state
The app checks whether a model is available and gives the interface a useful response when it is not. That makes hardware and Apple Intelligence requirements visible instead of pretending every device can run the same model.
Weather has a different boundary
Saventia’s optional weather tool uses Apple’s native WeatherKit framework to request a forecast for a location. It is intentionally separate from local chat, so “on device” describes the model and conversation storage, not every tool.
A small interface for a big constraint
I keep the conversation focused, with tool settings that make optional features clear. Local persistence lets people return to their conversations, while availability checks explain when the device cannot generate a response.