Abstract: This paper presents a conversational smart assistant architecture that integrates a resource-constrained microcontroller frontend with a cloud-based large language model (LLM) backend. The ...