Ship Onde from Flutter.

The onde_inference plugin gives Flutter apps the same on-device engine story: one Dart API, one package dependency, and a model that stays where the user is.

View sourceAll SDKs
pub.dev package

`onde_inference` is published on pub.dev. Add one dependency and you are in business.

Rust under the hood

The Dart layer talks to the compiled Onde engine through `dart:ffi`. The Flutter app stays pleasant. Inference stays on-device.

One mental model

Load a model, send a message, stream tokens, manage history. If you know the Swift SDK, this already feels familiar.

Broader reach

The package metadata already covers iOS, macOS, Android, Linux, and Windows, which is the practical upside of starting from a Rust core.

pub.devcrates.ioSwift SDK
ondeinference on GitHub

Quick start

Add one plugin.

Add the package with the Flutter CLI or pin it in `pubspec.yaml`. If you want to verify model downloads or GGUF output before you wire up the Flutter UI, use Onde CLI. If you want the product-level overview first, read What Onde Is.

# via CLI
flutter pub add onde_inference

# or in pubspec.yaml
dependencies:
  onde_inference: ^1.1.1

Coverage

One Flutter layer, wider reach.

Apple still feels like the natural starting point, but the Dart package is wider than that now. For the privacy argument behind local inference, read Privacy as a Feature, Not a Compliance Check.

PlatformABI / sliceStatus
iOSarm64 device, arm64 simulatorReady
macOSarm64Ready
Androidarm64Ready
Linuxx86_64 / arm64Ready
Windowsx64Ready