The Tanzu team has been hard at work on some new essential capabilities for organizations that are delivering AI Applications. With the release of VMware Tanzu Platform 10.2, customers building agentic, retrieval-augmented generation (RAG) or Generative AI (GenAI) applications, can now use new features in Tanzu AI Solutions that enable faster AI app delivery, safer operation of AI Apps, and new model optimization tools.
Here’s more about what’s new for AI application delivery in Tanzu Platform 10.2.
Tanzu AI Solutions: The Polyglot AI Application Delivery Platform
With the general availability of Spring AI, the Tanzu AI Solutions team has enabled Java developers to shift into AI development without having to reskill. But what about Python enthusiasts? After all, a lot of the data science discipline, which feeds AI models, utilizes Python. At Tanzu, we are committed to meeting the cross-team collaboration needs of our customers, so our innovation approach continues to enable polyglot capabilities in Tanzu Platform. With the AI space having a strong Python development community, we are pleased to announce that we are adding an application accelerator based on Python and Langchain, that provides similar functionality to spring-ai-chat.
Application Accelerators for VMware Tanzu Platform help developers deploy their applications in a repeatable way. They contain ready-made, enterprise-conformant code and configurations that are published in Git repositories. The accelerator file in the repository declares input options for the accelerator. This file also contains instructions for processing the files when generating a new project. Because it utilizes consistent bindings to services and deployment paths, applications built with accelerators can reduce operational burdens.
The Python accelerator for Tanzu Platform enables developers to rapidly and consistently deliver AI chat applications while utilizing their preferred programming language. Data teams can still build models in Python and enterprise developers can use the right language for the applications that consume those models- including Java, .net and now Python. So no one has to compromise or reskill.
With a polyglot platform, organizations can boost efficiency and reduce the risk of errors because teams are working with the programming languages they are most familiar with and are best suited for their work. Furthermore, polyglot teams using the Python accelerator and Spring AI can smoothly manage and maintain AI apps side-by-side on Tanzu Platform because it streamlines collaboration between development and platform engineering.
More Model Choices with Claude Model Family Support
Because the pace of change for foundation models is increasing, we’ve long counseled customers that a single AI model won’t cut it. Organizations will need to consider utilizing multiple models to address different phases of their applications’ evolution (e.g. test versus production), as well as to manage costs and enable new innovations. This is why Tanzu AI Solutions has been designed for operational agility.
Frequent model swapping is the name of the AI application game, so having a library of pre-vetted AI models is essential to maintaining AI application agility. Tanzu AI Solutions in Tanzu Platform makes it easier to swap models by utilizing OpenAI API for all model integrations so that you can quickly swap models without having to refactor your application. This consistency enables organizations to adapt to model innovations quickly.
Model flexibility is critical for AI competitiveness. This is why Tanzu AI Solutions supports both on-platform and endpoint model brokering so organizations have more choices of how to run their models. With the Tanzu Platform 10.2 release, we’ve expanded support for Anthropic’s Claude model family including a complete API integration. Full integrations of the Claude Models with Tanzu Platform offers more ways for organizations to efficiently operate their AI applications. In addition, the AI middleware in Tanzu AI Solutions enables organizations to apply governance features like rate-limiting, role-based-access-control and audit to their Claude models.
Model Optimization and Cost Control with Tanzu AI Solutions
AI models also frequently need to go through performance and cost evaluations to measure their return on investment (ROI). The constant iteration process that is needed to test, launch to production and then relaunch with different models can take a toll on organizations that are utilizing DIY solutions. Also, the models themselves need to be optimized to ensure efficient cost management.
Model distillation is like teaching a smart student (a smaller model) by having them learn from a really brilliant professor (a larger model). Instead of training the small model from scratch with just raw data, you let it learn from the outputs of the big model, which already understands the patterns really well. The goal is to make the small model almost as good as the big one, but faster and lighter. This is useful when you need powerful AI model performance on devices that can’t handle huge models, like phones or edge devices.
With Tanzu Platform 10.2 we have added a journaling feature to help organizations format their inference data, and then export it in the preferred format to third party distillation tools. By automatically formatting journals, the platform engineering team can streamline their exports and expedite the distillation process.
What’s next for Tanzu AI Solutions?
Tanzu is continuing to focus on innovation for Cloud Foundry based workloads and help our customers become AI native enterprises. This means delivering a Platform as a Service experience that incorporates security, reliability, governance and while supporting AI native innovation. From fine-tuned GenAI experiences to scalable agentic systems, Tanzu Platform is built to support innovation without compromise.
This is just the beginning—next, we’re focused on expanding our ecosystem with more integrations, deeper optimization tools, and continued improvements to help organizations stay ahead in the rapidly evolving AI landscape. Stay tuned— and join us at VMware Explore in Las Vegas August 25- 28 to learn more!
