README

laravel-gpt

This package provides a type-safe interface for making requests to the GPT-3 API.

use Capevace\GPT\Facades\GPT;

$response = GPT::generate(
    'Name a thing that is blue.',
    model: 'text-davinci-003',
    maxTokens: 400,
    frequencyPenalty: 1.0,
);

echo $response->first(); // "The sky"

Installation

You can install the package via composer:

composer require capevace/laravel-gpt

Configuration

You will need an API key for the OpenAI GPT-3 API. Once you have obtained an API key, you can configure it in your .env file by adding the following line:

OPENAI_API_KEY=your-api-key-here

You could also publish the config file directly, but this really probably isn't necessary:

php artisan vendor:publish --tag="laravel-gpt-config"

Usage

The Capevace\GPT\GPTService class provides methods for making requests to the GPT-3 API. You can inject it into controllers or use the Facade to access the container.

# Access via injection

use Capevace\GPT\GPTService;

class MyController extends Controller {
    protected GPTService $gpt;

    public function __construct(GPTService $gpt) {
        $this->gpt = $gpt;
    }

    public function index() {
        $this->gpt->generate(
            // ..
        );
    }

}

# Access via Facade

use Capevace\GPT\Facades\GPT;

GPT::generate(/* .. */);

GPT::generate(<prompt>, [...options])

The generate method returns a GPTResponse object that contains the response from the GPT-3 API. If no text is returned (empty string), the method will throw an error.

generate takes the following arguments:

• prompt (required): the prompt to send to the GPT-3 API
• model: the GPT-3 model to use (defaults to text-davinci-003)
• temperature: a value between 0 and 1 that determines how "creative" the response will be (defaults to 0.83)
• maxTokens: the maximum number of tokens (i.e., words) to return in the response (defaults to 1200)
• stop: a string that, when encountered in the response, will cause the response to end (defaults to null)
• frequencyPenalty: a value between 0 and 1 that determines how much the model will penalize frequent words (defaults to 0.11)
• presencePenalty: a value between 0 and 1 that determines how much the model will penalize words that don't appear in the prompt (defaults to 0.03)

Example

use Capevace\GPT\Facades\GPT;

$response = GPT::generate(
    'Generate a list of things that are blue.',
    model: 'text-davinci-003',
    maxTokens: 400,
    frequencyPenalty: 1.0,
);

Handling responses

The generate method returns a GPTResponse object that contains the response from the GPT-3 API.

It has two methods:

• $response->first() (string): returns the first text suggested by GPT-3
• $response->all() (array): returns a list of all the text choices suggested by GPT-3

Example

use Capevace\GPT\Facades\GPT;

$response = GPT::generate(
    'Name a thing that is blue.',
    model: 'text-davinci-003',
    maxTokens: 400,
    frequencyPenalty: 1.0,
);

$firstChoice = $response->first(); // "the sky"

$allChoices = $response->all(); // ["the sky", "the ocean" ...]

Error handling

If an error occurs while making a request to the GPT-3 API, the generate() method will throw a Capevace\GPT\Support\GPTException exception.

laravel-gpt will also throw an error, if a response does not contain any text (empty string).

Example

use Capevace\GPT\Facades\GPT;
use Capevace\GPT\Support\GPTException;
use Exception;

try {
    $response = GPT::generate('Do nothing.');
} catch (GPTException $e) {
    // Exception will be thrown, as the response text is ""
} catch (Exception $e) {
    // Catch connectivity errors etc.
}

Changelog

Please see CHANGELOG for more information on what has changed recently.

License

The MIT License (MIT). Please see License File for more information.