Boosting Business Capabilities with Pimcore Platform Version Release 2024.3

Welcome to the future of digital business solutions with the latest release of Pimcore Platform 2024.3. As we continue to embrace the transformative power of digital technology, this new version promises unprecedented advancements tailored to enhance your operational efficiency and data management capabilities. From revolutionary AI-driven enhancements to unique workflow optimizations, Pimcore 2024.3 is designed to empower businesses like yours to transform operations further.

Pimcore Copilot Enhancements: Elevating AI and Workflow Automation

With Pimcore Platform Version 2024.3, the Pimcore Copilot has evolved to handle an even more extensive array of complex business needs and use fine-tuned AI models based on your data. These enhancements solidify its position as an indispensable tool for AI-powered automation, enabling more precise and dynamic management of data and workflows across your enterprise.

New Automation Action: Automated data object classification with fine-tuned models

This new automation action leverages fine-tuned models to enhance data categorization by assigning attributes such as select, multi-select, and tags to various data objects. It offers a more flexible and precise approach compared to the traditional 'Hugging Face Zero-Shot Text Classification' by allowing adjustments to categories to fit the specific needs of your data better.

Key Features

1. Use your fine-tuned model
   You now have the option to integrate your own fine-tuned model into the system, allowing you to leverage its specialized capabilities and knowledge to identify the best matches for your classification tasks. By incorporating a model that has been tailored or optimized for your specific domain or dataset, you can enhance the accuracy and relevance of the classification results.
    
2. No limitation on possible results
   In contrast to the Hugging Face Zero Shot Text Classification approach, which operates within limited predefined categories and has constraints on the range of possible classification results, this alternative method does not limit the outcomes. This flexibility allows for a broader spectrum of potential results, enabling the model to generate more diverse and nuanced classifications.
    
3. Add attribute values to select, multi-select attributes and tags
   Determine the specific information that needs to be sent to the AI model for classification purposes and clearly define the attributes or features used in this classification process. Once the model processed the data, store the resulting classifications as values associated with different attributes.

Use Cases

Data object attribute extraction

• The challenge: You received product descriptions from one of your suppliers and stored them in Pimcore. For specific output channels, you need attributes (e.g., color, size…, etc.) as a secondary value in a multi-select field. You now need to extract this information from the description and store it in a multi-select field.
• The solution: Use a fine-tuned model for best results. Execute the action for all data objects that require an attribute extraction. The values will be stored separately in the defined multi-select field and can be used for various output channels.
• The benefit: With this action, you can speed up your data object maintenance process by automating the extraction of specific attributes. This reduces manual effort and decreases the risk of errors.

Automated classification of data objects

• The challenge: You use a taxonomy system built with Pimcore tags to classify data objects. This taxonomy system will also be used to specify which data objects should be used in which output channel. Lately, you have received a more significant amount of new data objects in Pimcore, and now you need to assign the correct category in the taxonomy system.
• The solution: Use a fine-tuned model for best results. Execute the action for all data objects that require a classification.
• The benefit: With this action, you can speed up your data object maintenance process by automating the assignment of tags. This reduces manual effort and decreases the risk of errors.

New Automation Action: Automated image classification with fine-tuned models

Like text classification, this automation action applies fine-tuned models to assign tags to images, facilitating a more detailed and nuanced categorization of visual content. This method extends beyond the limitations of 'Hugging Face Zero-Shot Image Classification,' providing a broader range of tagging options that enhance the specificity and relevance of image categorization.

Key Features

1. Use your fine-tuned model
   You have the option to integrate your own fine-tuned model into the classification system, allowing you to leverage its specialized training and expertise to identify the best matches for your specific classification needs. By incorporating a model that has been fine-tuned or optimized for your particular domain or dataset, you enhance the system's ability to deliver more accurate and relevant classification results. This customization enables you to build upon the existing knowledge embedded in the fine-tuned model, improving performance in detecting and categorizing data according to your unique requirements.
    
2. No limitation on possible results
   Compared to Hugging Face Zero Shot Image Classification, which operates within predefined categories and has constraints on the range of possible classification outcomes, this alternative method offers greater flexibility with no limitations on the results. This unrestricted approach allows for a more comprehensive exploration of possible classifications, enabling the model to generate a wider variety of results based on the input data.
    
3. Add tags to images
   Define the specific tags that should be used for classifying your images by establishing a tagging schema that outlines the categories or attributes you want to apply. Once you have determined these tags, you can send the images to the AI model for analysis. The model will process the images, identify relevant features and patterns, and assign the predefined tags accordingly. The results of this analysis will then be stored as tags associated with each image, providing a structured and organized way to categorize and manage your image data.

Use Cases

Classify images based on custom information

• The challenge: You want to tag your images based on some custom information (like emotion, fashion style, seasonal tags (winter, summer, …), image style, etc.) to improve finding the images via the search functionality.
• The solution: Use a fine-tuned model for best results. Execute the action for all images that require this custom information.
• The benefit: With this action, you can reduce the manual maintenance effort required for assigning custom information to images, ensuring you find the correct image even faster.

Enable AI fine-tuning: Empower Your Models

Pimcore Platform Version 2024.3 introduces a suite of AI training actions to support your initiatives in fine-tuning custom models using Hugging Face. These enhancements provide a detailed and actionable pathway for fine-tuning models tailored for text and image classifications. For this process, your data in Pimcore is used to deliver the best results for your specific requirements. These action steps can be found in the Copilot Showcase bundle.

Using our action steps to optimize AI models is just one option; alternatively, you can use your own training methods to fine-tune models to meet your custom needs.

These models can be used for new actions, such as AI Text Classification and AI Image Classification, which will also be included in this platform version and are described above.

New Automation Action: Change Workflow State

This action step integrates Pimcore Copilot with the Pimcore workflow engine. It enables efficient modification of the workflow state of various elements within a single operation. It helps maintain a cohesive process flow, allowing for rapid updates across multiple workflow components, enhancing efficiency, and reducing potential errors.

Key Features

1. Change the workflow state of manually selected elements
   Select one or more elements within the system and modify their workflow state by executing the appropriate action from the Pimcore Copilot. This transition can either be predefined in the configuration settings of the action or specified dynamically during the action's execution by the user.
2. Change the workflow state of automatically selected elements
   Select elements using previous action steps, such as applying a PQL (Pimcore Query Language) filter to retrieve all elements with a specific attribute. Once you have identified the relevant elements through these filters, you can change their workflow state by executing the appropriate action. This process allows you to target and modify the state of a precise subset of elements based on predefined criteria, streamlining the management of elements and ensuring that the workflow adjustments are accurately applied to the intended items.

Use cases

Change the workflow state of images in a specific folder

• The challenge: Assume you have a folder called ‘archive.’ All images moved to this folder should get the last workflow state without being changed manually.
• The solution: Use the ‘Change Workflow State’ and configure the action to be executed if an image gets moved to a specific folder. Therefore, you can use our new action step, ‘Events as Triggers for Actions.’ Every time an image gets transferred to a particular folder, the action gets executed.
• The benefit: This action reduces the manual maintenance effort and ensures that the workflow state of images is always up to date.
   

Automatically assign a 'review-needed' state after a text was generated in a Copilot Action

• The challenge: You generated texts with a Pimcore Copilot action and saved them to the corresponding data objects. A second person should now review these texts, and you need to change the workflow state to make that happen.
• The solution: Add a new action step to the existing action where the texts are created. With this action step (Change Workflow State), you can set the required workflow state as soon as a new text is created for a data object.
• The benefit: With this action step, you can automate a review process to minimize manual maintenance work and keep track of all changes made by this action.

Automatically assign a specific workflow state based on specific criteria

• The challenge: You want to change the workflow state of data objects to 'ready' as soon as specific data quality criteria are fulfilled. You want to maintain a specific text in all languages before the data object is used in another channel.
• The solution: Set up an action that filters for all data objects via a PQL action step where all languages for a specific text are maintained. These objects are now the scope for the following action step where you change the workflow state to ‘ready’. This action will run every 30 minutes triggered by a scheduler.
• The benefit: Only data objects that fulfill specific data quality criteria will be used for other output channels to guarantee optimal data quality throughout the different channels.           

Maximize the flexibility for the execution of Pimcore Copilot Actions

With the release of Platform Version 2024.3, we are excited to introduce new triggers that significantly enhance the capabilities for building more advanced and sophisticated automation scenarios within the Pimcore Copilot. These new triggers offer greater flexibility and control, empowering users to automate a broader range of tasks. In addition to the existing functionality, you can now utilize events and workflows as triggers to initiate the execution of Pimcore Copilot actions, enabling a more seamless and dynamic integration of automated processes across your Pimcore platform.

Key Features

1. Define an event or workflow to trigger action
   You can define which event or workflow should trigger the action. A dropdown menu allows you to select events and workflows for different data elements.
2. Define workspace
   You can define specific workspaces that should be considered for triggering the Pimcore Copliot action.
3. Define precondition
   If you want to apply more filtering, you can use the precondition dropdown or the PQL precondition field to filter based on more specific criteria.

Use Cases

Notification for deletion of objects in a specific folder

• The challenge: Only a few individuals within an organization usually have permission to delete objects. If an object is deleted from this archive folder, all product managers should be informed.
• The solution: Set up an action with a trigger to execute a notification action after a delete event. The notification will be sent to all product managers. In the trigger section of the action, you can specify the required folder (archive).
• The benefit: Proactive information about significant changes without manually scanning the affected folder.

Automatically assign metadata and change the parent of an asset

• The challenge: You have an import folder where many people can import new assets. These assets need to be checked manually and moved to another folder after approval. Additionally, you must add some text that describes the image to find it faster and apply the first workflow state.
• The solution: Set up an action that will be executed after a workflow change. Limit this action to the import folder. Add an action step to extract a description text from the image and another action step that moves the image to another folder after you have changed the workflow state so that the image is also available for other users.
• The benefit: With this action, you will speed up data maintenance by minimizing manual steps and ensuring the quality of your assets' data.

Further Advancing Automation Scope in Pimcore Copilot

Pimcore Platform Version 2024.3 significantly improves the workflow and automation control capabilities of Pimcore Copilot. These enhancements fine-tune data management, boosting operational precision within your enterprise workflows:

Improvement: Change the Parent of Assets
This functionality now includes modifying the parent of associated assets and enhancing the flexibility and control within your data management processes.

PQL Result as Scope for Other Actions
Actions can now utilize the results from PQL queries as scopes for subsequent operations, streamlining the execution of complex action chains and enhancing efficiency.

Add PQL as a Precondition
Allows for implementing complex filtering criteria through PQL, offering more refined control over the triggers of various actions based on specific data attributes.

Allow environment variables to be used in action step configuration
This improvement will enable users to have more influence on actions. Instead of defining all configuration options upfront in the configuration window, you can empower your users to input specific options when executing the action - e.g., allowing users to enter a new path when moving elements within the structure using a Pimcore Action.

Streamlining for Excellence: Additional Technical Improvements

We are excited to share a series of housekeeping updates that enhance the core framework and various functionalities. These improvements are part of our ongoing commitment to adapt to evolving technological landscapes based on your feedback!

We also dedicated some focus to reviewing and merging community contributions into this platform version. Overall, we have successfully addressed and resolved over 300 PRs since our last Platform Version Release, demonstrating significant progress and dedication from all of you. We sincerely want to thank each of you for your contributions! Our heartfelt appreciation also extends to our valued partners, twocream and Blackbit, for their ongoing dedication and collaboration in contributing to innovative improvements.

For this platform version release, we focused on the following bundles:

• Core Framework Enhancements: New execution mode for the Generic Execution Engine.
• Admin UI Classic Bundle: New grid preview for field collections, column filtering for relational fields, and optimized admin loading for reverse relations

• Data Hub Improvements: The improvements include the 'Add All Definitions' button, multi-drag and drop, query type for version on assets, and language multi-select mutation datatype.

• Generic Data Index: New operators (!=, not like)  for the Pimcore Query Language (PQL).

These updates, driven by our commitment to quality and user satisfaction, ensure that Pimcore meets and exceeds our users' needs, paving the way for more streamlined and effective digital business solutions.

Development News: Pimcore Studio Sneak Preview

In addition to rolling out Pimcore Platform Version 2024.3, we are excited to offer a sneak peek at the ongoing development of Pimcore Studio. Below, you'll find some exclusive enhancements that aim to transform the handling of broad and intricate tasks within Pimcore. Please note that these developments are tentative and may evolve.

Furthermore, this preview is accessible as a repository, allowing you to experience it directly. A concise step-by-step guide to activating the Pimcore Studio Preview is available on our GitHub page.

Keep an eye out for more updates and news in the upcoming chapters—plenty more are coming!