app_settings_ai 0.1.0

On-device AI settings proposals for the app_settings plugin. Apple Foundation Models on iOS, Gemini Nano on Android. Free and MIT licensed.

0.1.0

• Initial release. Companion package to app_settings that turns a natural-language request into a ranked list of settings screens to open, using an on-device model. Free and MIT licensed.
• AppSettingsAI.availability(), AppSettingsAI.prepareModel() and AppSettingsAI.proposeSettings().
• iOS 26+: Apple Foundation Models with guided generation, so the response structure is guaranteed. FoundationModels is weak-linked, so the plugin still loads on systems without it, which report unsupportedOsVersion rather than failing to launch. The effective floor is iOS 15, the minimum deployment target the iOS 27 SDK in Xcode 26 accepts; Package.swift declares .iOS("13.0") for older toolchains. Requires Xcode 26+ to build the real implementation.
• Android: ML Kit GenAI Prompt API (Gemini Nano). genai-prompt requires API 26, but this plugin keeps minSdk 16 and reports unsupportedOsVersion below that, so apps with a lower minSdk can adopt it and degrade gracefully.
• proposeSettings() never throws: an ineligible device yields an unavailable result and a failed generation an error one, including when the plugin is not registered. Both native handlers check capability before generating, so Dart does not spend a round trip pre-checking it.
• Both platforms receive identical instructions and prompt, built in Dart. All validation — unknown and duplicate setting keys, confidence clamping, ordering and the proposal cap — happens once in Dart, so the two platforms are held to the same rules.