Skip to main content

Getting Started

byjg/llm-api-objects provides a strongly-typed PHP model layer for building OpenAI Chat Completions request payloads. All models serialize to array via toApi(), ready to pass directly to openai-php/client.

It works with both OpenAI and Ollama (via the /v1 OpenAI-compatible endpoint).

Installation​

composer require byjg/llm-api-objects

Requirements​

  • PHP 8.3+

Connecting to OpenAI​

<?php
require 'vendor/autoload.php';

$client = OpenAI::factory()
    ->withApiKey($_ENV['OPENAI_API_KEY'])
    ->make();

Connecting to Ollama​

Ollama exposes an OpenAI-compatible /v1 endpoint. Pass any string as the API key (Ollama does not validate it) and set the base URI:

$client = OpenAI::factory()
    ->withApiKey('ollama')
    ->withBaseUri('http://localhost:11434/v1')
    ->make();

Your First Request​

<?php
use ByJG\LlmApiObjects\Enum\Role;
use ByJG\LlmApiObjects\Model\Chat;
use ByJG\LlmApiObjects\Model\Message;

$chat = new Chat(
    model: 'gpt-4o',
    messages: [
        new Message(role: Role::user, message: 'What is the capital of France?'),
    ],
    system: 'You are a helpful assistant.',
);

$result = $client->chat()->create($chat->toApi());

echo $result->choices[0]->message->content;
// Paris
tip

toApi() returns a plain PHP array. You can inspect it with print_r($chat->toApi()) before sending to debug the exact payload.

Multi-turn Conversation​

Pass multiple Message objects to build a conversation history:

use ByJG\LlmApiObjects\Enum\Role;
use ByJG\LlmApiObjects\Model\Chat;
use ByJG\LlmApiObjects\Model\Message;

$chat = new Chat(
    model: 'gpt-4o',
    messages: [
        new Message(role: Role::user,      message: 'My name is Alice.'),
        new Message(role: Role::assistant, message: 'Nice to meet you, Alice!'),
        new Message(role: Role::user,      message: 'What is my name?'),
    ],
);

$result = $client->chat()->create($chat->toApi());
echo $result->choices[0]->message->content;
// Your name is Alice.

Next Steps​