Penerapan RAG pada Chatbot Edukasi dan Deteksi Dini Penyakit THT

Abstract

Limited access to fast and accurate medical information is often a major constraint in the early detection of Ear, Nose, and Throat (ENT) diseases. This study proposes the development of an intelligent chatbot using the Retrieval-Augmented Generation (RAG) architecture to minimize hallucinations in Large Language Models (LLMs). The system is built using the n8n low-code automation platform, integrated with the WhatsApp API as the user interface and the gpt-4o-mini model as the inference engine. The system's knowledge base is sourced from external databases, including clinical references and visit data, processed through a vector store to ensure that responses remain within the context of valid data. Testing results indicate that the implementation of RAG increases information accuracy compared to the standard model. Furthermore, the use of n8n has proven to provide operational cost efficiency and accelerate the deployment cycle. The system successfully achieved an average latency of under 5 seconds with a Success Rate of 95%. This study concludes that the integration of RAG on a no-code platform is an effective solution for providing precise and economical health informatics services.

Ähnliche Arbeiten