MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202611048825 A) filed by Ayush Kumar; Neeti Gupta; Rajshree Dhoundiyal; and Shashi Kant, Greater Noida, Uttar Pradesh, on April 16, for 'intelligent prompt chaining framework for generating software component diagrams from natural language requirements.'
Inventor(s) include Ayush Kumar; Neeti Gupta; Rajshree Dhoundiyal; and Shashi Kant.
The application for the patent was published on May 29, under issue no. 22/2026.
According to the abstract released by the Intellectual Property India: "An automated software architecture diagrams generation system and method are revealed based on prompt-engineered large language models that create software architecture diagrams given, as opposed to textual requirements. Natural language requirement specifications are inputted into the invention which then processes them under an executed workflow of a chained reasoning process consisting of requirement analysis, detecting ambiguities, harnessing the system boundaries, identifying internal components, connector modelling, architecture style enforcement, providing a diagram code synthesis, syntax checking and optimizing its output. In contrast to the traditional single-prompt AI models, the presented framework operates predetermined prompts in a row, with each step narrows the context of the next step. This method can be used to extract with accuracy software modules, interfaces, dependencies, databases, event brokers, external services, and communication paths. The system embraces various architecture paradigms such as monolithic systems, layered systems, service oriented systems and event based microservices. It produces formal definitions of diagrams in PlantUML or Mermaid or similar languages, which are converted to visual diagrams that conform to UML. Syntactic correctness, semantic consistency, naming rules, completeness, connector accuracy and visual readability are verified by a validation engine. In case of defects, auto corrective iterations are elicited. In others, a set of candidate diagrams is produced and their ranking is done based on quality metrics and an optimum design is then chosen. The invention may be a web platform, enterprise assistant, IDE extension, educational tool or cloud service. It also enhances design cycles, minimizes manual architecture development effort, standardizes the quality of modelling and expands architecture development to non-experts. The reported invention offers an effective technical answer to transforming unstructured software prerequisites into structured engineering products by controlled AI coordination and responsive chaining, hence enhance the software planning, communication, and implementation preparedness across various industries."
Disclaimer: Curated by HT Syndication.
