AI-Associate: A Lightweight Architecture for Conversational Agents

In this paper, we introduce a novel architecture of chatbot oriented toward universal deployment of LLMs in scenarios that need small and quick LLM responses without full utilization of an LLM's computational resources. The described architectural solution would enable using a number of such models each with a lesser computational cost, thus improving the human-computer interaction process. Through experiments, we show that this architecture is resource-efficient for creating strong human-machine interaction systems. This architecture simplifies the deployment of advanced conversational AI, especially in constrained scenarios by budget and computational resources.