Moderne today added an artificial intelligence (AI) agent to its platform for refactoring codebases residing in multiple repositories.
Known as Moddy, the AI agent developed by Moderne is specifically trained to automate application modernization tasks such as converting SQL syntax into another format.
Moderne CEO Jonathan Schneider said Moddy has been trained using the OpenRewrite Lossless Semantic Trees (LSTs) code representation platform, an open source project that uniquely serializes LSTs in a way that captures the relationships between IT infrastructure and code bases, including dependencies, that span multiple repositories.
OpenRewrite is at the core of a Moderne application modernization platform that makes it possible to access code residing in multiple repositories. The ability to span multiple code bases is crucial because most legacy applications that organizations seek to modernize typically reside in one or more repositories. Rather than having to build AI agents for each repository that organizations would then need to integrate, Moddy makes it possible to use a single agent to automate a range of tasks versus having to invoke and integrate multiple AI agents for each codebase, said Schneider.
Moderne already provides multiple recipes that DevOps teams employ to modernize a range of legacy applications. Those recipes were used to train the Moddy AI agent, which will continue to be trained as additional recipes are created.
Moddy, designed to work with multiple large language models (LLMs), is also able to access knowledge graphs created via the interaction between LLMs and the recipe created by Moderne to provide deep understanding of the business logic embedded within an application. That capability makes it possible, for example, to determine where payments processing occurs within a specific business unit.
The OpenRewrite LST platform is already being employed by a range of providers of application modernization services, including vendors that rely on it to train their own AI agents to update code bases. The Moderne platform, in addition to providing organizations with commercial support for OpenRewrite LST, makes it possible to apply those capabilities across multiple code bases in a way that provides organizations with more visibility into and granular control over those workflows, said Schneider.
Ultimately, the goal is to make it simpler for organizations to swap out one IT platform for another as they best see fit, he added. Many organizations today are locked into legacy platforms simply because replacing the extensions that have been programmatically made to that platform is too complex to reverse engineer. The Moderne platform streamlines those efforts in a way that reduces the need to rely on a small army of consultants that would otherwise be required to accomplish that goal, said Schneider.
At a time when many of the providers of those legacy platforms are increasing their pricing, interest in application modernization has never been higher, he added.
The level of complexity involved with modernizing applications will, of course, vary widely. Updating from one version of Java to another is not nearly as complex an endeavor as trying to convert COBOL code into another programming language that might let an application run on another platform. Regardless of the scope of the project, however, there are a host of tasks that clearly lend themselves to being completed by an AI agent versus a software developer that should be squarely focused on ensuring that the business logic that drives the application is running in the most optimal way possible.
