Skip to main content
This section covers how to create requests to LLM providers using the Polyglot library. It includes examples of basic requests, handling multiple messages, and using different message formats.

​
Creating Requests

The with() method is the main way to set the parameters of the requests to LLM providers. It accepts several parameters: 
<?php
use Cognesy\Polyglot\Inference\Enums\OutputMode;

$inference = $inference->with(
    messages: 'Summarize this text',
    model: 'gpt-4o-mini',
    tools: [],
    toolChoice: [],
    responseFormat: [],
    options: [],
    mode: OutputMode::Text
);
// @doctest id="624a"

​
Basic Request Example

Here’s a simple example of creating a request: 
<?php
use Cognesy\Polyglot\Inference\Inference;

$inference = new Inference();
$response = $inference
    ->with(
        messages: 'What is the capital of France?'
    )
    ->create() // create pending inference
    ->get();   // get the data - here it executes the request

echo "Response: $response";
// @doctest id="97d1"

​
Request with Multiple Messages

For chat-based interactions, you can pass an array of messages: 
<?php
use Cognesy\Polyglot\Inference\Inference;

$inference = new Inference();
$response = $inference
    ->withMessages([
        ['role' => 'system', 'content' => 'You are a helpful assistant who provides concise answers.'],
        ['role' => 'user', 'content' => 'What is the capital of France?'],
        ['role' => 'assistant', 'content' => 'Paris.'],
        ['role' => 'user', 'content' => 'And what about Germany?']
    ])
    ->get();

echo "Response: $response";
// @doctest id="dd60"

​
Using Messages Class

You can also use the Messages class to create message sequences more conveniently: 
<?php
use Cognesy\Messages\Messages;
use Cognesy\Polyglot\Inference\Inference;

$messages = (new Messages)
    ->asSystem('You are a senior PHP8 backend developer.')
    ->asDeveloper('Be concise and use modern PHP8.3+ features.') // OpenAI developer role is supported and normalized for other providers
    ->asUser([
        'What is the best way to handle errors in PHP8?',
        'Provide a code example.',
    ]); // you can pass array of strings to create multiple content parts

$response = Inference::using('openai')
    ->withModel('gpt-4o')
    ->withMessages($messages)
    ->get();
// @doctest id="c7ab"

​
Message Formats

Polyglot supports different message formats depending on the provider:
  • String: A simple string will be converted to a user message
  • Array of messages: Each message should have a role and content field
  • Multimodal content: Some providers support images in messages
Example with image (for providers that support it): 
<?php
use Cognesy\Polyglot\Inference\Inference;

$imageData = base64_encode(file_get_contents('image.jpg'));
$messages = [
    [
        'role' => 'user',
        'content' => [
            [
                'type' => 'text',
                'text' => 'What\'s in this image?'
            ],
            [
                'type' => 'image_url',
                'image_url' => [
                    'url' => "data:image/jpeg;base64,$imageData"
                ]
            ]
        ]
    ]
];

$response = Inference::using('openai')
    ->withModel('gpt-4o') // use multimodal model
    ->with(messages: $messages)
    ->get();
// @doctest id="bf75"
Instructor library offers Cognesy\Messages\Utils\Image class for easier conversion of image files to the message format.