API Reference

Complete API documentation for react-litert. This section provides detailed information about all components, hooks, types, and configuration options.

Overview

react-litert provides a simple, React-friendly API for running TensorFlow Lite models in the browser. The library consists of:

• 1 Provider Component - <LiteRtProvider> for global configuration
• 1 Primary Hook - useModel for loading and running models
• 1 Runtime Hook - useLiteRtRuntime for runtime information
• Multiple Configuration Options - To customize runtime behavior

Provider Component

<LiteRtProvider>

The root provider component that initializes the LiteRT runtime and makes it available to all child components.

<LiteRtProvider config={{ wasmRoot: '/litert-wasm/' }}>
  <App />
</LiteRtProvider>

When to use: Wrap your entire application (or the part using models) with this provider.

Read full documentation →

Model Hooks

useModel

The primary hook for loading and running models. Supports both TensorFlow.js and raw LiteRT tensors with automatic type inference.

const { status, run, error } = useModel({
  modelUrl: '/models/mobilenet_v2.tflite',
  runtime: 'tfjs', // or 'litert'
});

When to use:

• For all model loading (replaces useLiteRtTfjsModel and useLiteRtModel)
• Use runtime: 'tfjs' for TensorFlow.js tensors
• Use runtime: 'litert' for raw LiteRT tensors

Read full documentation →

---

useLiteRtRuntime

Hook for checking runtime status and capabilities.

const { status, supportsWebGpu, supportsWasm } = useLiteRtRuntime();

When to use:

• Display runtime capabilities to users
• Conditionally render based on available accelerators
• Debug runtime initialization issues

Read full documentation →

---

Deprecated Hooks

The following hooks are deprecated and will be removed in v1.0.0. Please migrate to useModel.

useLiteRtTfjsModel (deprecated)

Replaced by useModel({ runtime: 'tfjs' }).

Migration guide →

useLiteRtModel (deprecated)

Replaced by useModel({ runtime: 'litert' }).

Migration guide →

Common Types

LiteRtModelStatus

type LiteRtModelStatus =
  | 'idle' // Not started
  | 'initializing-runtime' // Setting up WebGPU/WASM
  | 'compiling' // Compiling the model
  | 'ready' // Ready to use
  | 'error'; // Error occurred

LiteRtConfig

interface LiteRtConfig {
  wasmRoot?: string;
  preferAccelerators?: ('webgpu' | 'wasm')[];
  tfBackend?: 'webgpu' | 'wasm' | 'cpu';
  autoShareWebGpuWithTfjs?: boolean;
  onRuntimeError?: (error: Error) => void;
}

LiteRtTensorInfo

interface LiteRtTensorInfo {
  name: string;
  index: number;
  shape: number[];
  dtype: 'float32' | 'int32';
}

Quick Reference

API	Purpose	Import From
<LiteRtProvider>	Configure runtime	react-litert
useModel	Load and run models	react-litert
useLiteRtRuntime	Check runtime status	react-litert

Related Documentation

• Getting Started - Quick start guide
• Installation - Setup instructions
• Basic Usage - Common usage patterns
• Advanced Usage - Advanced techniques
• Examples - Complete example applications