RLMRec is a model-agnostic framework that enhances existing recommender systems with LLM-empowered representation learning. It integrates LLMs with collaborative filtering through graph neural networks to capture semantic aspects of user behaviors and preferences. The framework generates text descriptions for users and items, develops LLM-powered user/item profiles, and aligns the semantic space of LLMs with the representation space of collaborative signals via cross-view alignment. Implemented in PyTorch and evaluated on Amazon-book, Yelp, and Steam datasets.