Managing multiple chronic conditions through diet requires satisfying disease‑ and medication‑specific constraints across meals and days while preserving existing eating habits. These constraints are publicly listed in clinical guidelines and drug labels, yet converting them into a feasible personal plan remains hard for language‑model agents. We built a dietitian‑approved weekly‑plan evaluator from NHANES and tested thirteen models, ranging from medically fine‑tuned 8B models to GPT‑6.1 Sol. Direct generation yielded a complete‑plan pass rate below 3%. Adding explicit rules, more reasoning, or model‑directed tools raised GPT‑6.1 Sol’s pass rate to at most 8%.
We introduce Noura, which pairs LLM planning with executable personal specifications derived from public clinical guidance and individual health and dietary profiles. During planning, the specification grounds foods, optimizes portions, checks the multi‑day plan, and guides revisions. The key insight is that a specification that rejects a plan can also dictate what to learn from it. Noura turns executed plans and context‑aware revisions into supervision for compact models, recovering targets even from trajectories that never achieve full success.
The same specification supplies supervised targets, preferences, and rewards for post‑training. On 4,993 held‑out participants, Noura‑SFT raises Qwen3‑14B’s complete‑plan pass rate from 24.0% to 72.5%, cuts rule violations by 97.5% with no critical failures, and improves dietary quality. The locally hosted 14B planner incurs fewer violations than its GPT‑6 Luna teacher. A matched comparison shows executed targets reduce violations by 41% over model‑written proposals. Across Qwen, Phi, and Llama backbones, self‑training alone reduces violations by 41%‑61% on a 1,249‑person subset.
Review: Noura demonstrates that turning public clinical constraints into executable supervision can efficiently fine‑tune large language models, dramatically improving the feasibility and safety of multi‑condition dietary planning and offering a reproducible pathway for personalized health management.