Towards a Safe and Trustworthy LLM-RAG Chatbot for Hemodialysis Patient Education: Design, Implementation, and Preliminary Evaluation

Abstract

The increasing demand for accurate, timely, and personalized medical information in healthcare, especially for chronic disease management such as hemodialysis, has prompted the development of advanced technological solutions. This paper presents Chatbot Lisa, a Retrieval-Augmented Generation (LLM-RAG) system designed to assist hemodialysis patients by providing relevant and accurate medical information. The focus of this paper is on the system architecture, highlighting the use of the Django framework for the frontend, Sanic for asynchronous request handling, and ChromaDB for semantic retrieval. The chatbot engine leverages the best model, Sahabat AI LLM Model with BGE-M3 embedding model. Additional features, such as user authentication and continuous feedback mechanisms, are included to improve the user experience and support iterative system refinement.

Ähnliche Arbeiten