AI-Driven DTO Hydrator & Context Mapper

An innovative, lightweight, and framework-agnostic PHP 8+ library designed to bridge the gap between unstructured AI responses and PHP strict typing.

Using PHP's native Reflection API, this package automatically generates a type-safe schema from any Data Transfer Object (DTO) or class, prompts an AI provider to extract data from raw text, and safely hydrates the object while enforcing native PHP types (int, float, bool, string, DateTime).

? Why is it Innovative?

1. Strict Type Enforcement: Most AI packages only return raw strings or loose JSON. This library uses reflection to inspect your target class properties and dynamically forces the AI to output schema-matching data types.
2. Native Type Coercion & Safety: It shields your application from `TypeError` exceptions. The engine sanitizes and casts data before injecting it into private or protected properties.
3. Zero Dependencies: It does not rely on heavy frameworks like Laravel or Symfony. It can be dropped into any legacy or modern PHP project.
4. Provider Agnostic: Easily switch between OpenAI, Gemini, Claude, or local LLMs (like Ollama) by implementing a simple interface.

? Installation & Structure

Clone the repository or download the package structure:

ai-hydrator-package/
??? src/
?   ??? AIHydrator.php
?   ??? AIProviderInterface.php
?   ??? OpenAIProvider.php
??? tests/
?   ??? HydratorTest.php
??? composer.json
??? README.md

If using Composer, ensure your autoloader is updated:

composer dump-autoload

? Quick Start / Usage Example

1. Define your Target Class (DTO)

Create any standard PHP class using native type-hinting:

class OrderDTO
{
    private string $customerName;
    private int $quantity;
    private float $totalPrice;
    private \DateTime $purchaseDate;
    private bool $isPaid;

    // Getters
    public function getCustomerName(): string { return $this->customerName; }
    public function getQuantity(): int { return $this->quantity; }
    public function getTotalPrice(): float { return $this->totalPrice; }
    public function getPurchaseDate(): \DateTime { return $this->purchaseDate; }
    public function isPaid(): bool { return $this->isPaid; }
}

2. Hydrate the Object from Unstructured Text

use Innovation\AIHydrator\AIHydrator;
use Innovation\AIHydrator\OpenAIProvider;

// 1. Initialize the AI Provider (or any custom provider)
$provider = new OpenAIProvider('your-api-key-here');
$hydrator = new AIHydrator($provider);

// 2. Raw unstructured text (e.g., from an email or chat webhook)
$rawEmailText = "Hey team, I'm Robson. I just transferred 249.90 via Pix yesterday for those 5 t-shirts. Please confirm order.";

// 3. Magic happens here: Turn text into a strictly-typed PHP Object
/ @var OrderDTO $order */
$order = $hydrator->hydrate($rawEmailText, OrderDTO::class);

// 4. Use your object safely with full IDE autocomplete support
echo $order->getCustomerName();       // Output: Robson
echo $order->getQuantity();           // Output: 5 (as an integer)
echo $order->getPurchaseDate()->format('Y-m-d'); // Output: 2026-06-15 (as DateTime)

?? Testing

The package includes a built-in test suite that simulates an AI response using a manual Mock, allowing you to verify the reflection engine and type coercion without spending API credits.

To run the test localy, execute:

php -d assert.active=1 tests/HydratorTest.php

? License

This project is open-source and licensed under the MIT License. Feel free to use, modify, and share.

Submitted to the PHP Innovation Award on PHPClasses.org.