AI workspaces

Three guarded local assistants

These are workspace definitions, not unconstrained medical agents. Production AI remains disabled until explicitly configured and reviewed.

Agent 1 — Clinical nutrition

Doctor + diet planner + nutrition researcher + anatomy/biology/pharmacology reviewer.

  • Population-aware meal guidance
  • Interaction and contraindication checks
  • Evidence grading and citations

Agent 2 — Natural diet planner

Focuses on natural food strategies, ingredient pairing, and structured wellness planning.

  • Goal-based dietary plans
  • Ingredient alternatives
  • Local-only inference option

Agent 3 — Master chef lab

Researches and creates new meals and recipes, with a scheduled enrichment loop.

  • Recipe generation every 12h
  • Ingredient compatibility scoring
  • Recipe database expansion

Automation loop

The chef workspace is intended to run in a guarded, resumable loop: research → draft → validate → store → review → repeat.

ResearchSource gathering and recipe ideas
DraftStructured meal and ingredient output
ValidateSafety, allergens, and constraints
StorePersist to recipe database

Implementation note

These workspaces should be implemented behind feature flags, with deterministic safety checks and human review for anything that could affect health decisions.