​
Overview

Instructor PHP provides a way to debug HTTP calls made to LLM APIs via withDebug() method call on Inference object. When debug mode is turned on all HTTP requests and responses are dumped to the console.

​
Example

<?php
require 'examples/boot.php';

use Cognesy\Polyglot\Inference\Inference;

$response = (new Inference)
    ->withDebugPreset('on') // Enable debug mode
    ->with(
        messages: [['role' => 'user', 'content' => 'What is the capital of Brasil']],
        options: ['max_tokens' => 128]
    )
    ->get();

echo "USER: What is capital of Brasil\n";
echo "ASSISTANT: $response\n";
?>