KMPLiteRT documentation
KMPLiteRT runs LiteRT models from Kotlin Multiplatform applications. Use Core to load a model and run inference; add Data Processing only when you need image or audio preprocessing.
-
Get started
Add Core, prepare a model, and run your first inference.
-
Core runtime
Learn about model lifecycle, tensor buffers, and hardware acceleration.
-
Data processing
Decode, transform, and normalize image or audio inputs.
-
Platform guides
Configure Android, JVM, Apple, Native, JavaScript, or WasmJS.
Two modules, separate responsibilities
| Module | What it does | Use it when |
|---|---|---|
| Core | Loads LiteRT models, manages tensor buffers, and runs inference. | Your app already prepares model-compatible tensors. |
Data Processing (dataproces) |
Decodes and transforms image or audio data into model-ready values. | Your app needs preprocessing before inference. |
Core can be used on its own and does not depend on Data Processing. Data Processing is optional; it depends on Core only to offer direct TFBuffer integration.