One engine, four ways to ship it
Onde’s SDKs are thin wrappers around the same onde Rust crate — pick the
layer that matches your app.
| SDK | Distribution | Notes |
|---|---|---|
| Rust | crates.io , docs.rs | The source of truth. Direct access to ChatEngine, model configs, streaming, and the full async API. |
| Swift | Swift Package Index | Installs with Swift Package Manager; already ships across iPhone, Mac, Apple TV, Vision Pro, and Apple Watch. |
| Kotlin Multiplatform | Maven Central | |
| Flutter (Dart) | pub.dev | One plugin, one Dart API, local inference on iPhone and Mac. |
| React Native | npm | JavaScript on top, native engine underneath. |
Supported platforms
macOS, iOS, tvOS, visionOS, watchOS, Android, Linux, and Windows all build
against onde today.
Current scope
The product is strongest in chat today: model loading, history, platform defaults, one-shot generation, and streaming. That’s enough to build real apps, not just toy demos — see Siti AI for a complete, open-source (Apache-2.0 ) reference app built on Tauri across macOS, iOS, and Android.
For install steps and platform notes per SDK, see
ondeinference.com/sdk . Once your app has a
model assigned through GresIQ, load_assigned_model is what fetches it — see
GresIQ & deployment.