Barbara platform release 2.8.0: New AI frameworks, improved network management and much more

Enhancing your MLOps experience

Barbara allows you to deploy your trained AI models directly, as simply as just exporting them and sending them to the platform. This was possible with Tensor Flow models but we are now allowing PyTorch and ONNX frameworks too!

‍

In order to give you more flexibility, we have also included Triton as a new inference server option when deploying your models to your Edge Nodes.

We have also included GPU support so you can run your AI models on GPU-capable edge nodes too.

‍

And finally, for clarity sake, we are now showing the inference server URL as part of the information available at the model’s card view.

‍

Avoid connectivity losses due to a wrong Wifi configuration

We have improved our network management capabilities and we have just included the “rollback” functionality. Now, if you set up a wrong wifi configuration to any of your nodes, they will rollback to its previous one so you don’t lose connectivity to them. 

2.8.0 version also allows you to connect your edge nodes to hidden wifi networks

‍

More flexible configuration of the Marketplace apps you deploy

In order to speed up your configuration process during the deployment and execution of Marketplace apps, we have included a few new features such as:

• Ports and volumes configuration: during the deployment wizard you are now able to specify the ports and docker volumes of the apps
• App secrets edition: once you have a Marketplace app deployed, you can now edit its secrets directly from the app card in the details view of your edge node
  

• Improved UX of the wizard: we have redesigned the deployment wizard so it’s cleaner and easier to use.
  

‍

Get a better experience with the latest UI/UX improvements

You will see a couple of renamings in the platform. This is a consequence of reviewing the wordings and adapting them to the usual naming conventions and an effort to maintain consistency and coherence throughout the whole platform:

• We have implemented new naming conventions for secrets and configurations: Global Env Secrets, App Env Secrets, Compose Config, Global Config, and App Config.
• We have renamed the “Spaces” view as the “Workloads” view. Workload is a term typically used for applications and AI models that are already deployed.
  

We have redesign a few views to help you get the most out of them:

• The Node List view: we have enhanced this view so you can personalize it with additional information and other customization possibilities.
  

• Configuration of the apps: We have redesigned the global config card (learn more about the Global Config card in here) and the local app config segment by revising the presentation of the configuration within.

Additional changes

We have included a security update and some minor bugs have been fixed in order to address any potential vulnerabilities and maintain a secure and stable environment.

Additionally, we have migrated our platform domains to barbara.tech. The official URL for all the platform modules are now:

• Barbara Panel: https://panel.barbara.tech
• Barbara Marketplace: https://market.barbara.tech
• Barbara Academy: https://academy.barbara.tech

The old barbaraiot.com domain will still be accessible though.

‍

Explore these new features and see how they make your work with Barbara more intuitive and efficient. We welcome your thoughts and feedback on these updates!