Day 4 – Matching GPT Suggestions to Real Products in Laravel #LaravelGPT #ProductMatching #AIRecommendations #EcommerceAI #PersonalizedUX

Today, we’ll take the text-based suggestions from GPT (like “Gaming Chair” or “Monitor Arm”) and match them against real products in your database to return full product details.

🧩 Step 1: Update ProductRecommendationService to match suggestions

In recommend() method, replace the final return logic with:

use App\Models\Product;

$matchedProducts = Product::query()
    ->where(function ($query) use ($decoded) {
        foreach ($decoded['suggestions'] ?? [] as $suggestion) {
            $query->orWhere('name', 'LIKE', '%' . $suggestion . '%');
        }
    })
    ->limit(5)
    ->get();

return $matchedProducts;

Your updated recommend() method now returns a Collection of Product models.

🧠 Optional: Add fallback matching (description/category)

You can enhance it:

->orWhere('description', 'LIKE', '%' . $suggestion . '%')
->orWhere('category', 'LIKE', '%' . $suggestion . '%')

🧪 Step 2: Test route with JSON product output

In web.php:

Route::get('/recommendations', function (ProductRecommendationService $service) {
    $user = \App\Models\User::first(); // demo
    $products = $service->recommend($user);

    return response()->json($products);
});

Expected output:

[
  {
    "id": 7,
    "name": "Gaming Chair",
    "description": "Comfortable chair for long gaming sessions",
    "price": "450.00",
    "category": "Furniture"
  },
  ...
]

🎯 Bonus: Add a score system (optional)

To return better matches, you can use Laravel Scout or similarity algorithms like Levenshtein distance or full-text search in future days.

✅ Summary

✅ Today you:

• Took GPT output (plain text)
• Matched it to real products in the DB
• Returned full product info ready for UI display

✅ Up next (Day 5): We’ll build a Blade view to display recommendations in a card-based layout, complete with images, titles, and “View More” links.