Harnessing generative AI to create and understand architecture diagrams

Architecture diagrams are required tools for software development, system design, and communication. They facilitate the understanding of complex systems by providing a visual representation of components, relationships, and data flow. However, creating and interpreting these diagrams can be time-consuming and require significant expertise. Generative Artificial Intelligence (AI) offers a potential solution to automate the creation process and improve comprehension. This paper explores how generative AI can be leveraged to automatically generate diverse architectural diagrams from textual descriptions and code repositories. Additionally, the research investigates how AI techniques can assist in understanding and analyzing existing diagrams, thereby easing maintenance, documentation, and stakeholder communication. This paper discusses existing approaches, emerging techniques, challenges, and future directions in this evolving field. Our findings indicate that generative AI can significantly reduce effort in diagram creation and improve analysis while also exploring the limitations of current models.