Development of an Intelligent Educational Platform Module using LLM and RAG Technologies
Abstract
The article discusses the design and implementation of an intelligent assistant for the "PROPS" educational platform based on the Retrieval-Augmented Generation (RAG) architecture. The relevance of the transition from static learning management systems (LMS) to adaptive environments using large language models (LLM) to form individual educational trajectories is substantiated. The technological cycle of the "AssistenPROPS" assistant is described, including the use of the Llama 3.1 model, vector databases, and semantic matching mechanisms. Special attention is paid to the method of dynamic generation of prompt templates that take into account the student's cognitive load, the context of the dialogue, and pedagogical goals. The scientific novelty of the study lies in the implementation of the principle of dynamic thematic granularity of content, which provides automatic branching of thematic nodes within competencies without pre-setting rigid hierarchies. The authors show that the integration of the RAG framework allows minimizing the risk of generative "hallucinations" and ensuring the factual accuracy of educational materials. The results of the study confirm the effectiveness of using RAG architectures to create scalable intelligent educational systems of the next generation.