End-to-End Deployment Framework for LLM-Powered Test Case Generation from Software Requirement Specifications

Abstract

It is a world of artificial intelligence. Terms which excite gen-z’s the most are Chatgpt, Llama, Gemini. The backbones of any such Chabot’s are large language models. There are numerous applications of LLM’s in every phase of software development life cycle. From requirement generation to maintenance of software almost every phase can be automated using LLM’s. The testing phase is not left behind. Using large language models in test data generation, test suit creation, creating test cases from bug reports, creation of unit test cases etc. Writing the test cases initially even before the development of software starts gives a cutting edge to STLC as a lot of human efforts are needed to write them and is a mundane task for testers. The aim of this study is to generate an end to end framework which helps in automatic generation of test cases from user requirements using various LLM’s available. We focus only in the area of TDD (Test driven Design) where we aim to generate test cases from the initial requirement document. The study also compares the quality of the generated test cases with ground truths to reveal the percentage of manual efforts reduced by automating the test case generation process.