vision_engine.js

import 'dotenv/config';
import axios from 'axios';
import { readFileSync } from 'fs';

// CapSolver API configuration
const CAPSOLVER_API_KEY = process.env.CAPSOLVER_API_KEY;
const CREATE_TASK_URL = 'https://api.capsolver.com/createTask';

// Vision Engine modules
const MODULES = {
    slider_1: 'Slider captcha - requires image and imageBackground',
    rotate_1: 'Rotation detection - requires image and imageBackground',
    rotate_2: 'Standalone rotation - requires only image',
    shein: 'Multi-object selection with coordinates - requires image and question'
};

/**
 * Solve slider captcha.
 * @param {string} imageBase64 - Base64 encoded slider piece image
 * @param {string} backgroundBase64 - Base64 encoded background image
 * @returns {Promise<Object>} Solution containing distance to slide
 */
async function solveSlider(imageBase64, backgroundBase64) {
    const payload = {
        clientKey: CAPSOLVER_API_KEY,
        task: {
            type: 'VisionEngine',
            module: 'slider_1',
            image: imageBase64,
            imageBackground: backgroundBase64
        }
    };

    const response = await axios.post(CREATE_TASK_URL, payload);
    const result = response.data;

    if (result.errorId !== 0) {
        throw new Error(`Failed to solve: ${result.errorDescription}`);
    }

    return result.solution || {};
}

/**
 * Solve rotation captcha.
 * @param {string} imageBase64 - Base64 encoded image to rotate
 * @param {string} backgroundBase64 - Base64 encoded background (required for rotate_1)
 * @param {string} module - "rotate_1" (with background) or "rotate_2" (standalone)
 * @returns {Promise<Object>} Solution containing angle to rotate
 */
async function solveRotate(imageBase64, backgroundBase64 = null, module = 'rotate_1') {
    const payload = {
        clientKey: CAPSOLVER_API_KEY,
        task: {
            type: 'VisionEngine',
            module: module,
            image: imageBase64
        }
    };

    if (backgroundBase64 && module === 'rotate_1') {
        payload.task.imageBackground = backgroundBase64;
    }

    const response = await axios.post(CREATE_TASK_URL, payload);
    const result = response.data;

    if (result.errorId !== 0) {
        throw new Error(`Failed to solve: ${result.errorDescription}`);
    }

    return result.solution || {};
}

/**
 * Solve Shein-style multi-object selection captcha.
 * @param {string} imageBase64 - Base64 encoded image
 * @param {string} question - The object type to identify
 * @returns {Promise<Object>} Solution containing rects (coordinate arrays)
 */
async function solveShein(imageBase64, question) {
    const payload = {
        clientKey: CAPSOLVER_API_KEY,
        task: {
            type: 'VisionEngine',
            module: 'shein',
            image: imageBase64,
            question: question
        }
    };

    const response = await axios.post(CREATE_TASK_URL, payload);
    const result = response.data;

    if (result.errorId !== 0) {
        throw new Error(`Failed to solve: ${result.errorDescription}`);
    }

    return result.solution || {};
}

/**
 * Helper to load image as base64.
 * @param {string} imagePath - File path
 * @returns {string} Base64 encoded image
 */
function loadImageAsBase64(imagePath) {
    const imageBuffer = readFileSync(imagePath);
    return imageBuffer.toString('base64');
}

async function main() {
    if (!CAPSOLVER_API_KEY) {
        console.log('Error: CAPSOLVER_API_KEY not found in .env file');
        console.log('Please create a .env file with your API key:');
        console.log('CAPSOLVER_API_KEY=your_api_key_here');
        return;
    }

    console.log('Vision Engine - Slider/Rotate Captcha Solver');

    console.log('\nAvailable modules:');
    for (const [module, desc] of Object.entries(MODULES)) {
        console.log(`  ${module}: ${desc}`);
    }

    console.log('\nExample usage:');
    console.log('\n  // Slider captcha:');
    console.log('  const sliderImg = loadImageAsBase64("slider.png");');
    console.log('  const bgImg = loadImageAsBase64("background.png");');
    console.log('  const solution = await solveSlider(sliderImg, bgImg);');
    console.log('  // Returns: { distance: 150 }  // pixels to slide');

    console.log('\n  // Rotation captcha (with background):');
    console.log("  const solution = await solveRotate(imageBase64, backgroundBase64, 'rotate_1');");
    console.log('  // Returns: { angle: 45 }  // degrees to rotate');

    console.log('\n  // Rotation captcha (standalone):');
    console.log("  const solution = await solveRotate(imageBase64, null, 'rotate_2');");
    console.log('  // Returns: { angle: 90 }');

    console.log('\n  // Shein multi-object selection:');
    console.log('  const solution = await solveShein(imageBase64, "Select all shoes");');
    console.log('  // Returns: { rects: [[x1,y1,x2,y2], [x1,y1,x2,y2]] }  // object coordinates');
}

main();

export { solveSlider, solveRotate, solveShein, loadImageAsBase64, MODULES };