Alternative - AI generated metatags for images in TYPO3 with Google Gemini

Introduction

This TYPO3 extension allows setting alternative texts, title labels and a description for images in a filestorage. This can be done via file list backend module or via command on the CLI.

Example metadata labels from AI:

Example backend integration:

Example CLI command:

Google Gemini API

• To use the extension, you need a Google Gemini API key. You can register for one at https://aistudio.google.com/app/api-keys.
• Alternatively, you can implement your own LLM provider (see Custom LLM Integration below).

Installation

composer req in2code/alternative

After that, you have to set some initial configuration in Extension Manager configuration:

Title	Default value	Description
setAlternative	1	Toggle function: Set a value for alternative text
setTitle	1	Toggle function: Set a value for image title
setDescription	1	Toggle function: Set a value for a description
showButtonInFileList	1	Show or hide button in backend module file list
apiKey	-	Google Gemini API key. You can let this value empty and simply use ENV_VAR "GOOGLE_API_KEY" instead if you want to use CI pipelines for this setting
limitToLanguages	-	If set, limit to this language identifiers only. Use a commaseparated list of numbers

Note: It's recommended to use ENV vars for in2code/alternative instead of saving the API-Key in Extension Manager configuration

GOOGLE_API_KEY=your_api_key_from_google

CLI commands

# Set metadata for all image files in storage 1
./vendor/bin/typo3 alternative:set "1:/"

# Set metadata for all image files in a subfoler in storage 1 (maybe "fileadmin/in2code/folder/")
./vendor/bin/typo3 alternative:set "1:/in2code/folder/"

# Enforce to set metadata for all image files in storage 1
./vendor/bin/typo3 alternative:set "1:/" 1

Custom LLM Integration (like ChatGPT, Claude, Mistral, etc.)

Alternative uses a factory pattern to allow custom LLM providers. By default, it uses Google Gemini, but you can easily integrate other AI services (OpenAI, Claude, local models, etc.).

Implementing a Custom LLM Repository

1. Create a custom repository class implementing RepositoryInterface - see example for OpenAI ChatGPT:

<?php

declare(strict_types=1);

namespace Vendor\MyExtension\Domain\Repository\Llm;

use In2code\Alternative\Domain\Repository\Llm\AbstractRepository;
use In2code\Alternative\Domain\Repository\Llm\RepositoryInterface;
use In2code\Alternative\Exception\ApiException;
use In2code\Alternative\Exception\ConfigurationException;
use TYPO3\CMS\Core\Http\RequestFactory;
use TYPO3\CMS\Core\Resource\File;

class ChatGptRepository extends AbstractRepository implements RepositoryInterface
{
    private string $apiKey = '';
    private string $apiUrl = 'https://api.openai.com/v1/chat/completions';

    public function __construct(
        protected RequestFactory $requestFactory,
    ) {
        parent::__construct($requestFactory);
        $this->apiKey = getenv('OPENAI_API_KEY') ?: '';
    }

    public function checkApiKey(): void
    {
        if ($this->apiKey === '') {
            throw new ConfigurationException('OpenAI API key not configured', 1735646000);
        }
    }

    public function getApiUrl(): string
    {
        return $this->apiUrl;
    }

    public function analyzeImage(File $file, string $languageCode): array
    {
        $this->checkApiKey();
        $this->setLanguageCode($languageCode);
        $imageData = base64_encode($file->getContents());
        return $this->generateMetadataWithChatGpt($imageData, $file->getMimeType());
    }

    protected function generateMetadataWithChatGpt(string $imageData, string $mimeType): array
    {
        $payload = [
            'model' => 'gpt-4o',
            'messages' => [
                [
                    'role' => 'user',
                    'content' => [
                        ['type' => 'text', 'text' => $this->getPrompt()],
                        [
                            'type' => 'image_url',
                            'image_url' => [
                                'url' => 'data:' . $mimeType . ';base64,' . $imageData,
                            ],
                        ],
                    ],
                ],
            ],
            'temperature' => 0.1,
            'max_tokens' => 500,
        ];

        $options = [
            'headers' => [
                'Authorization' => 'Bearer ' . $this->apiKey,
                'Content-Type' => 'application/json',
            ],
            'body' => json_encode($payload),
        ];

        $response = $this->requestFactory->request($this->getApiUrl(), $this->requestMethod, $options);

        if ($response->getStatusCode() !== 200) {
            throw new ApiException(
                'Failed to analyze image with ChatGPT: ' . $response->getBody()->getContents(),
                1735646001
            );
        }

        $responseData = json_decode($response->getBody()->getContents(), true);
        return $this->parseResponse($responseData);
    }

    protected function parseResponse(array $responseData): array
    {
        if (isset($responseData['choices'][0]['message']['content']) === false) {
            throw new ApiException('Invalid ChatGPT API response structure', 1735646002);
        }

        $text = $responseData['choices'][0]['message']['content'];

        // Extract JSON from markdown code blocks if present
        if (preg_match('/```json\s*(\{.*?\})\s*```/s', $text, $matches)) {
            $text = $matches[1];
        } elseif (preg_match('/```\s*(\{.*?\})\s*```/s', $text, $matches)) {
            $text = $matches[1];
        }

        $data = json_decode($text, true);
        if ($data === null) {
            throw new ApiException('Could not parse ChatGPT response as JSON', 1735646003);
        }

        return [
            'title' => $data['title'] ?? '',
            'description' => $data['description'] ?? '',
            'alternativeText' => $data['alternativeText'] ?? '',
        ];
    }
}

2. Register your custom repository in ext_localconf.php:

<?php
defined('TYPO3') || die();

// Register custom LLM repository
$GLOBALS['TYPO3_CONF_VARS']['EXTENSIONS']['alternative']['llmRepositoryClass']
    = \Vendor\MyExtension\Domain\Repository\Llm\ChatGptRepository::class;

Hint: Don't forget to register your Repository in your Services.yaml and flush caches

Changelog and breaking changes

Version	Date	State	Description
2.0.0	2025-12-31	Feature	Prepare overruling of the GeminiRepository
1.2.0	2025-12-07	Feature	Support TYPO3 14, use better titles and icons, decrease length for metadata, skip missing images
1.1.0	2025-12-04	Feature	Add ddev as local environment, prevent syntax error in PHP 8.2
1.0.0	2025-12-03	Task	Initial release of in2code/alternative

Contribution with ddev

This repository provides a DDEV-backed development environment. If DDEV is installed, simply run the following commands to quickly set up a local environment with example usages:

• ddev start
• ddev initialize

Backend Login:

Username: admin
Password: admin

Installation hint:

1. Install ddev before, see: https://ddev.readthedocs.io/en/stable/#installation
2. Install git-lfs before, see: https://git-lfs.github.com/
3. Add the API key in .ddev/.env like

GOOGLE_API_KEY=your_api_key_from_google