Moderation Call

Moderation

Moderation calls are calls that can give feedback if a prompt is offensive or in certain provider cases against the terms and conditions to send in to the provider.

Example normalized Moderation call

The following is an example of sending the text "I want to kill all of them" into OpenAI using the text-moderation-latest model. The output is a ModerationResponse object.

use Drupal\ai\OperationType\Moderation\ModerationInput;

$input = new ModerationInput('I want to kill all of them');
/** @var \Drupal\ai\OperationType\Moderation\ModerationOutput $moderation_object */
$moderation_object =  \Drupal::service('ai.provider')->createInstance('openai')->Moderation($input, 'text-moderation-latest', ['my-custom-call']);

/** @var \Drupal\ai\OperationType\Moderation\ModerationResponse $moderation */
$moderation = $moderation_object->getNormalized();

// If moderation is flagged, we wrote something bad.
if ($moderation->isFlagged()) {
  print_r($moderation->getInformation());
  // This will output reasoning for being flagged.
  throw new \Exception('Exit here, so we will not get banned');
}

Moderation Interfaces & Models

The following files defines the methods available when doing a moderation call as well as the input and output.

• ModerationInterface.php
• ModerationInput.php
• ModerationOutput.php

Moderation Explorer

If you install the AI API Explorer, you can go configuration > AI > AI API Explorer > Moderation Generation Explorer under /admin/config/ai/explorers/ai-moderation to test out different calls and see the code that you need for it.