Presentation Information
[3Yin-A-54]Automated Spring Boot Unit Test Generation Using Workflow-Based AI AgentsImplementation of an Iterative Test Completion System Using JaCoCo and GPT-5
〇Kotaro Okamoto1, Kohei Itaya1 (1. IBM Japan Systems Engineering Co.)
Keywords:
Workflow-Based AI Agent,Automated Unit Test Generation
In software development, manual test code creation for ensuring test coverage requires enormous time costs, making the improvement of test development efficiency a critical challenge. In frameworks employing dependency injection such as Spring Boot, the complexity of the mechanism poses a significant burden for beginners in test creation. This study developed a fully automated unit test generation system using workflow-based AI agents compatible with Spring Boot. We constructed an agent using Python, OpenAI Agents SDK, and GPT-5, integrating coverage measurement with JaCoCo. The system automatically executes an iterative process of source code reading, test generation, test execution and report analysis, and supplementary test addition. We conducted an empirical experiment to automatically generate test code for an official Spring Boot sample application, achieving high-level code coverage without human intervention. This study demonstrated that integrating AI agents and coverage tools enables both quality and speed. Furthermore, we enhanced practicality by designing the system to be integrated into development workflows with only Python code invocation.