This repository contains the codes of the front-end developed for the Final Course Work entitled "Violet: Uma Arquitetura para Facilitar a Interação e Acesso a Informações por meio de Perguntas sobre o Mundo Real (VIOLET: A Real-Time Architecture for Understanding and Answering Natural Language Questions)", authored by Lucas Henrique Lírio Silva Loureiro, supervised by Felipe Augusto Lima Reis, at the Instituto de Ciências Exatas e Informática (ICEI) of the Pontifical Catholic University of Minas Gerais (PUC Minas).
This work presents the development and evaluation of an architecture called Violet, designed to assist people in interacting with their surroundings. The architecture integrates speech recognition, speech synthesis, advanced language models, and computer vision technologies to answer audio questions about images. A mobile application was implemented using the Violet architecture to evaluate its feasibility and performance. The backend was developed in Java, using the Spring Framework WebFlux module, running on Google Cloud Compute Engine, while the frontend for Android mobile devices was built with Flutter. The application was subsequently evaluated through load and stress tests. The results demonstrate that the application provided satisfactory and timely responses, despite the challenges related to the complexity of the questions. Thanks to its asynchronous processing approach, the application efficiently handled multiple simultaneous requests. This work contributes to an architecture that can be integrated into several applications, combining different AI tools to improve the accessibility and autonomy of people.
It is developed in Java, using a non-blocking thread approach to allow asynchronous processing of requests through the Spring Framework's WebFlux module.
The backend repository is available here.