Desenvolvimento de um ambiente web inteligente para geração automática de código para ESP32 a partir de linguagem natural

Abstract

This project focused on the development of an automated programming environment for the ESP32 microcontroller, supported by Artificial Intelligence techniques and a Django-based backend. The proposal aimed to reduce the technical barrier faced by beginners in the field of embedded systems, enabling the generation of C/C++ code from natural language instructions. The system was structured to initiate a conversation with the user, guided by a conversational assistant based on the ChatGPT model, adapted to the specific domain of the ESP32. Although direct integration with OpenAI’s custom model was not possible due to API limitations, this restriction was overcome by incorporating the assistant’s prompt into the backend, ensuring that the model reproduced its intended behavior. In addition to the integration with the OpenAI API, a local fallback mechanism was implemented to ensure system resilience, generating basic examples when no valid API key or credits are available. The Django architecture enabled the creation of a simple, responsive, and scalable web interface, including REST routes, a database for logging interactions, and preparation for OTA (Over-The-Air) deployment. From a practical standpoint, the work demonstrates initial technical feasibility, evidencing that the developed prototype signals potential to contribute to simplifying access to embedded programming and to the practical use of the ESP32 microcontroller in educational and rapid prototyping contexts. The project points to promising avenues for future discussions regarding the integration of Artificial Intelligence as a supporting tool for learning embedded systems.