RAGatar: Enhancing LLM-driven Avatars with RAG for Knowledge-Adaptive Conversations in Virtual Reality

We present a virtual reality system that enables users to seamlessly switch between general conversations and domain-specific knowledge retrieval through natural interactions with AI-driven avatars. By combining MetaHuman technology with self-hosted large language models and retrieval-augmented generation, our system demonstrates how immersive AI interactions can enhance learning and training applications where both general communication and expert knowledge are required.