Abstract
This paper presents a personalized book recommendation system that generates suggestions based on a user's reading history and semantic analysis of textual preferences. Unlike conventional genre-based approaches, the proposed system captures implicit preferences by identifying thematic patterns in previously read works and modelling user behavior over time. The system architecture is based on a fine-tuned LLaMA 3.2 language model combined with retrieval-augmented generation (RAG) to dynamically construct query context. Evaluation was conducted on a proprietary dataset of 100 literary works spanning 10 genre categories. Testing results indicate that recommendation accuracy improves as user history accumulates. The system is applicable in educational settings for navigating library collections, including universities with specialized literary funds.