import { pipeline, env, RawImage } from '@huggingface/transformers';
import sharp from 'sharp';
import { readFileSync } from 'fs';

env.localModelPath = './'; // Path to your ONNX model
env.allowRemoteModels = false; // Disable remote models

// Load the ONNX model
const imageClassifier = await pipeline('image-classification', 'saved-model/bk');

// Load and preprocess the image
const imageBuffer = readFileSync('./training_images/shirt/00e745c9-97d9-429d-8c3f-d3db7a2d2991.jpg');
const image = await sharp(imageBuffer).resize(128, 128).raw().toBuffer();

// Run inference
const results = await imageClassifier(image);
console.log(results);


// import { pipeline, env, RawImage } from '@huggingface/transformers';
// import sharp from 'sharp';

// // Configure environment
// env.localModelPath = './'; // Path to your ONNX model
// env.allowRemoteModels = false; // Disable remote models

// async function preprocessImage(imagePath) {
//     const imageBuffer = await sharp(imagePath)
//         .resize(128, 128) // Resize to model's expected dimensions
//         .raw() // Get raw pixel data
//         .toBuffer();

//     const array = new Float32Array(imageBuffer.length).map((_, i) => imageBuffer[i] / 255.0); // Normalize to [0, 1]

//     // RawImage expects data as Uint8ClampedArray, convert and reshape accordingly
//     return new RawImage(
//         Uint8ClampedArray.from(array.map(v => v * 255)), // Rescale back for RawImage
//         128,
//         128,
//         3 // Channels
//     );
// }

// async function classifyImage(imagePath) {
//     const classifier = await pipeline('image-classification', 'saved-model/');

//     // Preprocess the image
//     const preprocessedImage = await preprocessImage(imagePath);

//     // Run the model inference
//     const results = await classifier(preprocessedImage);

//     console.log(results);
//     return results[0]?.label || 'Unknown';
// }

// // Example usage
// classifyImage('./training_images/shirt/00e745c9-97d9-429d-8c3f-d3db7a2d2991.jpg')
//     .then(partNumber => {
//         console.log(`Predicted Part Number: ${partNumber}`);
//     })
//     .catch(error => console.error(error));