METHOD OF AUTOMATED CONTINUOUS GENERATION OF UML DIAGRAMS IN LARGE-SCALE SOFTWARE REPOSITORIES

Abstract

Large-scale software projects often struggle to keep design documentation, such as UML diagrams, up-to-date with rapid code changes. This article proposes a method to integrate automated UML diagram generation into the continuous integration/continuous deployment (CI/CD) pipeline of a large .NET code repository. Upon each commit to the main branch, the CI pipeline triggers a secondary process that uses an AI component to detect code structure changes and update or create corresponding UML diagrams. Existing diagram definitions are retrieved and provided to the AI alongside code diffs, enabling it to generate revised UML. The updated diagrams are then published back to the documentation page, replacing outdated content. This ensures that architectural diagrams continuously evolve with the codebase, eliminating documentation drift and reducing manual effort. The main pipeline comprises: building the primary solution, executing the UML generation module, updating documentation, and then proceeding with further build steps. The approach leverages text-based “diagrams-as-code” and AI to maintain living documentation in real time. The paper analyzes recent research in continuous documentation and AI-driven code visualization, outlines the implementation in detail, and discusses the benefits and challenges.