README

An AI-powered PHP library to describe and analyze clothing items from images, designed specifically to assist visually impaired users in making informed fashion choices.

This tool provides not just a neutral description of a product, but also a personalized style analysis, outfit suggestions, and occasion recommendations, making online shopping more accessible and intuitive.

Features

Multi-Image Analysis: Gathers details from multiple product images for a comprehensive description.
Personalized Style Matching: Compares an item against a user's personal style profile.
Structured Data Output: Returns a clean, easy-to-use PHP object (StyleAnalysisResult) with distinct data points.
Outfit & Occasion Suggestions: Provides actionable advice on how to wear an item and where.
Provider-Agnostic: Can be configured to work with any OpenAI-compatible API, including commercial services like OpenAI and free, local models via Ollama.
Multi-language Support: Delivers analysis in the user's preferred language.

Installation

You can install the package via Composer:

composer require dotxdd/image-stylist

How to Use

The library is designed to be straightforward. Here is a basic example of how to use it with the OpenAI API.

1. Initialization

First, initialize the ImageStylistService. You need to provide your API key, the API endpoint, and the model name.

<?php

require 'vendor/autoload.php';

use Dotxdd\ImageStylist\ImageStylistService;
use Dotxdd\ImageStylist\StyleAnalysisResult;

// --- Configuration for OpenAI ---
$config = [
    'api_key' => 'YOUR_OPENAI_API_KEY',
    'endpoint' => 'https://api.openai.com/v1/chat/completions',
    'model' => 'gpt-4o',
];

$service = new ImageStylistService(
    apiKey: $config['api_key'],
    apiEndpoint: $config['endpoint'],
    model: $config['model']
);

2. Performing the Analysis

Prepare your data and call the getStyleAnalysis method.

try {
    // A description of the user's personal style
    $userStyleProfile = "I prefer an elegant and business-casual style. I mainly wear muted colors like navy, grey, and white. I like well-tailored blazers and simple trousers.";

    // An array of one or more image URLs of the product
    $imageUrls = [
        'https://.../image-of-a-blazer-front.jpg',
        'https://.../image-of-a-blazer-detail.jpg',
    ];

    // Call the service and get a structured result object
    /** @var StyleAnalysisResult $result */
    $result = $service->getStyleAnalysis($imageUrls, $userStyleProfile, 'en'); // 'en' for English

    // --- 3. Using the Results ---
    
    if ($result->isStyleMatch) {
        echo "✅ This item is a likely match for your style.\n";
    } else {
        echo "⚠️ This item is likely not a match for your style.\n";
    }

    echo "Description: " . $result->objectiveDescription . "\n";
    echo "Style Analysis: " . $result->styleAnalysis . "\n";
    echo "Outfit Suggestion: " . ($result->outfitSuggestion ?? 'None provided.') . "\n";
    echo "Occasion: " . $result->occasionAnalysis . "\n";

} catch (\Exception $e) {
    echo "An error occurred: " . $e->getMessage();
}

Using with Ollama (Free, Local)

This library works seamlessly with local models via Ollama. Simply change the configuration when initializing the service:

// --- Configuration for a local Ollama server ---
$config = [
    'api_key' => 'ollama', // Not required, can be any string
    'endpoint' => 'http://localhost:11434/api/generate', // Use the /api/generate endpoint
    'model' => 'llava', // The name of your local multimodal model
];

$service = new ImageStylistService(
    apiKey: $config['api_key'],
    apiEndpoint: $config['endpoint'],
    model: $config['model']
);

Jak Używać (Polish Guide)

Ta biblioteka została stworzona, aby pomagać osobom niewidomym i słabowidzącym w dokonywaniu świadomych wyborów modowych podczas zakupów online.

Instalacja

Zainstaluj paczkę za pomocą Composer:

composer require dotxdd/image-stylist

Przykład Użycia

Poniższy przykład pokazuje, jak uzyskać analizę produktu, używając API OpenAI.

<?php

require 'vendor/autoload.php';

use Dotxdd\ImageStylist\ImageStylistService;
use Dotxdd\ImageStylist\StyleAnalysisResult;

// 1. Konfiguracja serwisu
$service = new ImageStylistService(
    apiKey: 'TWOJ_KLUCZ_API_OPENAI',
    apiEndpoint: 'https://api.openai.com/v1/chat/completions',
    model: 'gpt-4o'
);

// 2. Przygotowanie danych
$profilStyluUzytkownika = "Preferuję styl elegancki i biznesowy. Noszę głównie stonowane kolory, takie jak granat, szarość i biel.";
$adresyUrlZdjec = [ 'https://.../zdjecie-produktu.jpg' ];

// 3. Wywołanie analizy (w języku polskim)
$wynik = $service->getStyleAnalysis($adresyUrlZdjec, $profilStyluUzytkownika, 'pl');

// 4. Użycie wyników
echo "Rekomendacja: " . ($wynik->isStyleMatch ? 'Pasuje' : 'Nie pasuje') . "\n";
echo "Opis Obiektywny: " . $wynik->objectiveDescription . "\n";
echo "Sugestia Stroju: " . ($wynik->outfitSuggestion ?? 'Brak') . "\n";

License

This project is licensed under the MIT License. See the LICENSE file for details.

dotxdd / image-stylist

Maintainers

Details