Akeneo Rekognition Bundle Laravel Package

Getting Started

Minimal Steps to First Use

1. Install the Bundle

   composer require clickandmortar/akeneo-rekognition-bundle
   

   Ensure compatibility with your Akeneo version (e.g., v0.3.* for Akeneo v4.0.*).

2. Enable the Bundle Add to config/bundles.php:

   ClickAndMortar\AkeneoRekognitionBundle\ClickAndMortarAkeneoRekognitionBundle::class => ['all' => true]
   

3. Configure AWS Credentials Set AWS credentials in config/services/services.yml (or via environment variables):

   parameters:
       aws.access_key_id: '%env(AWS_ACCESS_KEY_ID)%'
       aws.secret_access_key: '%env(AWS_SECRET_ACCESS_KEY)%'
       aws.region: '%env(AWS_REGION)%'
   

4. Trigger First Analysis Use the CLI command to process a product:

   php bin/console akeneo-rekognition:analyze --product-sku="PRODUCT_SKU"
   

   Verify results in the product’s rekognition_labels or rekognition_text attributes (if configured).

Implementation Patterns

Workflows

1. Batch Processing Use the akeneo-rekognition:analyze:batch command to process multiple products via a CSV file:

   php bin/console akeneo-rekognition:analyze:batch --file="path/to/products.csv"
   

   • CSV Format: sku,image_path (e.g., PROD001,/path/to/image.jpg).

2. Event-Driven Integration Subscribe to Akeneo’s product_update event to auto-trigger analysis:

   // src/EventListener/RekognitionListener.php
   use ClickAndMortar\AkeneoRekognitionBundle\Event\RekognitionEvent;
   
   class RekognitionListener {
       public function onProductUpdate(RekognitionEvent $event) {
           $product = $event->getProduct();
           if ($product->hasImage()) {
               $this->rekognitionService->analyze($product);
           }
       }
   }
   

   Register in services.yml:

   services:
       App\EventListener\RekognitionListener:
           tags:
               - { name: kernel.event_listener, event: akeneo_rekognition.analyze, method: onProductUpdate }
   

3. Custom Attribute Mapping Extend the bundle to map Rekognition results to custom product attributes:

   # config/akeneo_rekognition.yml
   attributes:
       labels:
           - { attribute: "product_color", field: "color_name" }
       text:
           - { attribute: "product_description", field: "extracted_text" }
   

Integration Tips

• Image Validation: Pre-check image URLs/paths for accessibility before analysis.
• Rate Limiting: AWS Rekognition has usage limits. Implement retries with exponential backoff:

  $client = new \Aws\Rekognition\RekognitionClient([
      'region'  => 'us-east-1',
      'version' => 'latest',
      'retry'   => [
          'max_attempts' => 3,
          'mode'         => 'standard',
      ],
  ]);
  

• Async Processing: Use Symfony Messenger or Akeneo’s queue system for large catalogs:

  $message = new AnalyzeProductMessage($product);
  $bus->dispatch($message);
  

Gotchas and Tips

Pitfalls

1. AWS Credentials

   • Issue: Hardcoded credentials in services.yml expose security risks.
   • Fix: Use environment variables or AWS IAM roles (for EC2/ECS):

     parameters:
         aws.credentials.role_arn: '%env(AWS_IAM_ROLE_ARN)%'
     

2. Image Format/Size

   • Issue: Rekognition requires images ≤ 5MB and specific formats (JPEG/PNG).
   • Fix: Validate and resize images before processing:

     use Intervention\Image\ImageManager;
     
     $image = (new ImageManager())->make($imagePath);
     $image->resize(1024, 1024)->save();
     

3. Attribute Conflicts

   • Issue: Overwriting existing product attributes (e.g., description).
   • Fix: Use unique attribute names (e.g., rekognition_description) or merge logic:

     $product->setDescription(
         $product->getDescription() . "\n\nDetected: " . $rekognitionText
     );
     

4. Deprecated Methods

   • Issue: Bundle uses rekognition-php v1.x (abandoned in 2020).
   • Fix: Fork the package or migrate to AWS SDK v3 directly:

     $result = $client->detectLabels(['Image' => ['S3Object' => ['Bucket' => 'bucket', 'Name' => 'image.jpg']]]);
     

Debugging

• Enable Verbose Logging Add to config/packages/monolog.yaml:

  handlers:
      akeneo_rekognition:
          type: stream
          path: "%kernel.logs_dir%/akeneo_rekognition.log"
          level: debug
  

• Test with Mocked AWS Use aws-sdk-mock for local testing:

  composer require aws/aws-sdk-php-mock
  

  use Aws\Rekognition\RekognitionClient;
  use Aws\Sdk\MockHandler;
  
  $mock = new MockHandler();
  $mock->append(new MockDetectLabels());
  $client = new RekognitionClient([], ['handler' => $mock]);
  

Extension Points

1. Custom Analyzers Implement ClickAndMortar\AkeneoRekognitionBundle\Analyzer\AnalyzerInterface for additional AWS services (e.g., Celebrity Recognition):

   class CelebrityAnalyzer implements AnalyzerInterface {
       public function analyze($image): array {
           $result = $client->recognizeCelebrities(['Image' => ['Bytes' => file_get_contents($image)]]);
           return $result->getCelebrityFaces();
       }
   }
   

   Register in services.yml:

   services:
       App\Analyzer\CelebrityAnalyzer:
           tags:
               - { name: akeneo_rekognition.analyzer, type: "celebrity" }
   

2. Post-Processing Filters Filter Rekognition results before storage (e.g., exclude "background" labels):

   // src/Service/RekognitionFilter.php
   class RekognitionFilter {
       public function filterLabels(array $labels): array {
           return array_filter($labels, fn($label) => $label['Name'] !== 'Background');
       }
   }
   

   Inject into the bundle’s service container.

3. Database Storage Override default attribute storage to use a dedicated table:

   // src/Entity/RekognitionResult.php
   #[ORM\Entity]
   class RekognitionResult {
       #[ORM\Id, ORM\GeneratedValue]
       private ?int $id = null;
   
       #[ORM\ManyToOne(targetEntity: Product::class)]
       private ?Product $product = null;
   
       #[ORM\Column(type: 'json')]
       private array $labels = [];
   
       // ...
   }