Dietary nutrition planning is crucial for chronic disease management and maintaining health. In practice it must satisfy personalized constraints and multidimensional nutritional goals simultaneously, which often conflict, and user constraints evolve with feedback, creating a gap between generic guidelines and executable plans. To bridge this gap we first define the personalized fully quantified multi‑objective dietary planning problem (MDP). To solve MDP we build the nutrition agent ShanLiangRen. The system first translates dietary specifications, nutrient data, user attributes and natural language requirements into an individualized constrained planning instance. It then applies an exact retrieval‑augmented generation method to shrink the feasible candidate set from a large ingredient and recipe space. Finally, it performs Pareto‑guided refinement: an LLM iteratively revises candidate plans under deterministic nutrition computation and constraint verification feedback. The output consists of fully quantified meal plans with explicit ingredients and portion sizes, together with nutrition compliance reports that show constraint satisfaction and nutrient interval attainment. The system is deployed as a WeChat mini‑program, with a demo video linked in the paper.
Review