Agente Autónomo Baseado em LLM para Testes Visuais de Interface de Utilizador
Student
André Ricardo Gonçalves de Freitas (B.S.)
Abstract
Software testing plays a fundamental role in the application development lifecycle, ensuring the quality, reliability, and usability of the developed systems. Although existing automation tools can execute functional tests, they are often dependent on rigid selectors and predefined flows, making them fragile to application changes. Additionally, usability testing remains heavily dependent on human intervention, making the validation process time-consuming and expensive.
This project proposes the development of a dynamic test automation system based on Large Language Models (LLMs), Appium, and Selenium WebDriver. The developed solution combines traditional automation techniques with function calling and structured outputs, allowing the system to interact with web applications in an Android environment in a more flexible and contextual manner. Operating as an autonomous agent based on a multimodal vision-language model, the system interprets purely visual inputs and takes dynamic navigation and interaction decisions, significantly reducing manual effort and automating usability feedback collection.
More information
- Date
- 2026
