README

🧠 Brain - The Adorably Simple OpenAI Wrapper

Just a small and simple wrapper around the OpenAI SDK.

📦 Installation

composer require helgesverre/brain

php artisan vendor:publish --provider="OpenAI\Laravel\ServiceProvider"

OPENAI_API_KEY="your-key-here"
OPENAI_REQUEST_TIMEOUT=60

🛠 Usage

🔧 Basic Setup

After installation, set up the Brain Facade in config/app.php:

'aliases' => [
    'Brain' => HelgeSverre\Brain\Facades\Brain::class,
],

📝 Example: Generating Blog Titles

Generate a list of blog titles about a given subject:

use HelgeSverre\Brain\Facades\Brain;

$subject = 'Technology';
$count = 5;

$titles = Brain::list("Suggest $count blog titles about '$subject'", 200);

foreach ($titles as $title) {
    echo $title . PHP_EOL;
}

📄 Example: Generating a Structured Blog Post

Generate a structured blog post in JSON format:

use HelgeSverre\Brain\Facades\Brain;

$title = 'The Future of Technology';
$style = 'informative';
$minWords = 500;

$response = Brain::slow()->json(<<<PROMPT
Create an $style blog post with the title '$title'. 
Write over $minWords words.
PROMPT
);

echo "Title: " . $response['title'] . PHP_EOL;
echo "Body: " . $response['body'] . PHP_EOL;

📄 Example: Text Classification with Arrays or Enums

Brain::classify simplifies the categorization of text. You can classify text using either an array of options or an Enum class.

Array Classification

Pass a list of categories as an array to classify your text.

use HelgeSverre\Brain\Facades\Brain;

$input = 'banana';
$categories = ["bread", "animal", "car", "plane"];

$result = Brain::fast()->classify($input, $categories);

This method evaluates 'banana' and categorizes it as one of the provided options.

Enum Classification

For structured categorization, use an Enum class.

use HelgeSverre\Brain\Facades\Brain;

enum Category: string {
    case Fruit = 'fruit';
    case Animal = 'animal';
    case Car = 'car';
}

$input = 'banana';

$result = Brain::fast()->classify($input, Category::class);

Here, 'banana' is classified into the most fitting Enum category.

📄 Example: Generating a Vector Embedding

This method returns a vector embedding of the input, or a list of vector embeddings if you pass an array or a collection.

use HelgeSverre\Brain\Facades\Brain;

// Single embedding
$result = Brain::embedding('banana');
// Returns ['0.123', '0.456', '0.789' ....]

// Or, for multiple inputs:
$result = Brain::embedding(['banana', 'apple', 'orange']);
// Returns [['0.123', '0.456', '0.789' ....], ['0.123', '0.456', '0.789' ....], ['0.123', '0.456', '0.789' ....]]

// Or, for a collection of inputs:
$result = Brain::embedding(collect(['banana', 'apple', 'orange']));
// Returns [['0.123', '0.456', '0.789' ....], ['0.123', '0.456', '0.789' ....], ['0.123', '0.456', '0.789' ....]]

Changing the Base URL to use Together.AI, Mistral.AI, Perplexity.AI or other compatible API

You can change the base URL to use other compatible APIs by using the usingTogetherAI, usingMistralAI or usingPerplexity methods.

use HelgeSverre\Brain\Facades\Brain;

Brain::usingTogetherAI()->text('Hello, world!');
Brain::usingMistralAI()->text('Hello, world!');
Brain::usingPerplexity()->text('Hello, world!');

Or you can set the base URL directly, note that the API must be compatible with OpenAI's API, specifically the Chat completion endpoints.

Also note that JSON Mode (as used by Brain::json, Brain::classify, Brain::embedding and Brain::list) is not supported by all APIs.

use HelgeSverre\Brain\Facades\Brain;

Brain::baseUrl('api.example.com');

📖 Available Methods

Method	Parameters	Description
`Brain::maxTokens()`	`int $maxTokens`	Sets the maximum number of tokens (words) the AI response can contain.
`Brain::temperature()`	`float $temperature`	Sets the 'temperature' for the AI responses, influencing the randomness of the output.
`Brain::fast()`	None	Sets the AI model to 'gpt-3.5-turbo-1106' for faster responses.
`Brain::slow()`	None	Sets the AI model to 'gpt-4-1106-preview' for more detailed responses.
`Brain::text()`	`$prompt, ?int $max = null`	Sends a text prompt to the AI and returns a text response. Optionally set a custom maximum token limit for this request.
`Brain::json()`	`$prompt, ?int $max = null`	Sends a prompt to the AI and returns a response in JSON format. Optionally set a custom maximum token limit for this request.
`Brain::list()`	`$prompt, ?int $max = null`	Sends a prompt to the AI and returns a list of items in an array, useful for generating multiple suggestions or ideas. Optionally set a custom maximum token limit for this request.
`Brain::classify()`	`$input, array\|StringBackedEnum $categories`	Classifies the given input text into one of the provided categories. Categories can be an array of strings or an Enum class.
`Brain::embedding()`	`$input`	Uses the `text-embedding-ada-002` model to generate an embedding vector, returns a single vector for string input, an array of vectors when passed an array or collection.
`Brain::toText()`	`CreateResponse $response, $fallback = null`	Converts an OpenAI `CreateResponse` object to a text string. Includes an optional fallback value.
`Brain::toJson()`	`CreateResponse $response, $fallback = null`	Converts an OpenAI `CreateResponse` object to a JSON object. Includes an optional fallback value.
`Brain::usingTogetherAI()`	None	Uses the Together.AI API instead of OpenAI
`Brain::usingMistralAI()`	None	Uses the Mistral.AI API instead of OpenAI
`Brain::usingPerplexity()`	None	Uses the Perplexity.AI API instead of OpenAI

📜 License

This package is licensed under the MIT License. For more details, refer to the License File.

helgesverre / brain

Maintainers

Details