Physical activity is essential for overall health, strengthening the cardiovascular system and musculoskeletal structure while also improving mental well-being. However, adherence to personalized training programs remains challenging, particularly for individuals requiring regular monitoring. Our research focuses on developing a chatbot powered by large language models (LLMs) and a Retrieval-Augmented Generation (RAG) architecture, integrating a specialized knowledge base in sports science. This system delivers individualized training plans and real-time feedback, enhancing user engagement. The methodology involves data preprocessing, hybrid lexical-semantic search, and response generation through a locally deployed LLM. This approach ensures safe and personalized sports guidance, making exercise more accessible to both amateur athletes and individuals with health conditions. Our study demonstrates how AI can revolutionize sports training by bridging the gap between medical care and fitness through intelligent, tailored support.