README

This is an API client written in PHP to interact with the GPT@EC API.

Important notes

The GPT@EC API is very similar to the Open AI API.
Therefore, in order to ease the development, this library reuses the openai-client/php where possible, or extends it if needed.
Since some parts of the openai-client/php are marked as internal, this library might become incompatible with future versions of it. Consider locking the openai-client/php version, and update it to the latest version reported as compatible in the test builds.

Install

Use Composer to require the library:

composer require openeuropa/gpt-at-ec-php-client

Make sure you have a PSR-18 client installed, or install one manually, e.g.:

composer require guzzlehttp/guzzle

The php-http/discovery plugin can also help installing and discovering a compatible PSR-18 client.

Usage

You can create a new API client using the factory class. All you need is a GPT@EC API key:

$api_key = getenv('KEY_AI_GPT_AT_EC');
$factory = new \Openeuropa\GptAtEcPhpClient\Factory();
$client = $factory->withApiKey($key)->make();

Now the client can be used to interact with the 3 available endpoints.

Chat

The Chat Completions API endpoint will generate a model response from a list of messages comprising a conversation.

create method

Create a chat completion. The full model response will be returned at once by the API.

$response = $client->chat()->create([
    'model' => 'gpt-4o',
    'messages' => [
    [
        'role' => 'system',
        'content' => 'You are a helpful assistant.',
    ],
    [
        'role' => 'user',
        'content' => 'Hello!',
    ],
]);

echo $response->id; // "chatcmpl-0123456789abcdef0123456789ab"
echo $result->object; // "chat.completion"
echo $result->created; // 1755523764
echo $result->model; // "gpt-4o-2000-01-01"

echo $response->choices[0]->message->role; // "assistant"
echo $response->choices[0]->message->content; // "Hello! How can I assist you today?"
echo $response->choices[0]->index; // 0
echo $response->choices[0]->finishReason; // "stop"

echo $response->usage->promptTokens; // 9
echo $response->usage->completionTokens; // 10
echo $response->usage->totalTokens; // 19

createStreamed method

Create a chat completion. The response will be streamed back in parts as the model generates it.

$stream = $client->chat()->createStreamed([
    'model' => 'gpt-4o',
    'messages' => [
    [
        'role' => 'system',
        'content' => 'You are a helpful assistant.',
    ],
    [
        'role' => 'user',
        'content' => 'Hello!',
    ],
]);

foreach ($stream as $chunk) {
    echo $chunk->id; // "chatcmpl-0123456789abcdef0123456789ab"
    echo $chunk->object; // "chat.completion.chunk"
    echo $chunk->created; // 1755530359
    echo $chunk->model; // "meta-llama/Llama-3.3-70B-Instruct"

    echo $response->choices[0]->message->role; // "assistant"
    echo $response->choices[0]->message->content; // "Hello!", "How", "can", "I", "help", "you", "?"
}

Models

Lists the various models available in the API.
A single method is available, list.

$response = $client->models()->list();

foreach ($response->data as $model) {
    echo $model->id; // "llama-3.3-70b-instruct"
    echo $model->name; // "LLama 3.3 70b instruct"
    echo $model->description; // "Very powerful open-weights model on par with the capabilities of GPT-4o for many types of tasks."
    print_r($model->sensitivityLevel); // ["PA", "SNC", "CU"]
    print_r($model->defaultFor); // ["SNC"]
    echo $model->created; // 1744201138356
    echo $model->ownedBy; // "meta"
    echo $model->object; // "model"
}

Quota consumption

Provides the quota consumption details for a specific AI model. A single method is available, retrieve.

$response = $client->quotaConsumption()->retrieve('gpt-4o');

echo $response->consumedPromptTokens; // 1234
echo $response->consumedCompletionTokens; // 5678
echo $response->consumedTotalTokens; // 6912
echo $response->quota; // 100000

Tests

The library is full covered by unit tests using the PHPUnit framework. To run the tests, execute:

./vendor/bin/phpunit --coverage-html folder-for-coverage-results

Versioning

We use SemVer for versioning.

openeuropa / gpt-at-ec-php-client

Maintainers

Package info

Statistics

Security