TFLite Plugin

Run on-device TensorFlow Lite model inference from NativePHP Mobile. Load .tflite models, perform real-time audio inference, and receive detection events in PHP.

Platform: Android only. Adds android.permission.RECORD_AUDIO.

Features

• Load TFLite models from base64 or app storage
• List and manage stored models on-device
• Real-time inference on device microphone audio
• Configurable sample rate, threshold, and sensitivity
• Detection events dispatched to PHP via NativePHP event system

Installation

composer require 1338/nativephp-tflite
php artisan vendor:publish --tag=nativephp-plugins-provider
php artisan native:plugin:register 1338/nativephp-tflite

Usage

use OneThreeThreeEight\NativephpTflite\Tflite;

// Upload a model (base64)
$base64 = base64_encode(file_get_contents('/path/to/model.tflite'));
Tflite::addModel('my_model.tflite', $base64);

// List stored models
$models = Tflite::listModels();

// Load and start listening
Tflite::loadModelFromFile('my_model.tflite');
Tflite::startListening([
    'sampleRate'  => 16000,
    'threshold'   => 0.7,
    'sensitivity' => 0.5,
]);

// Stop inference
Tflite::stopListening();

Events

• DetectionEvent — fired when the model detects a match above the configured threshold; includes inference results