With the rise of digital transformation, virtual assistants are increasingly playing a vital role in higher education. This paper presents a framework for a Vietnamese virtual assistant designed to support student affairs in universities, leveraging large language models (LLMs) and retrieval-augmented generation (RAG). By integrating LLMs with RAG, the system efficiently retrieves relevant information from institutional databases while ensuring context-aware and accurate responses to student inquiries. The assistant provides real-time support for administrative tasks, academic queries, and campus services, reducing administrative workload and enhancing student engagement. A prototype was developed and evaluated based on response accuracy, efficiency, and user satisfaction. The results indicate the feasibility of LLM and RAG-based virtual assistants in improving student support services. Future research will focus on expanding domain adaptability, optimizing retrieval mechanisms, and integrating multimodal interaction features. The source code are free and open at https://github.com/hodachung1709/tdmubot .