Local Food AI USER GUIDE

Local Food AI Platform

Clinician User Manual & Functional Application Walkthrough

Audience Clinical Staff & Users
Author Francois Lange
Status Ready (production)
Domain-specific Project 1 (DSP1) - Clinical Guide 1 / 9
Project Vision PRIVACY FIRST

The Vision: A Privacy-First Clinical Dietitian

  • Complete HIPAA Compliance: Patient dietary parameters, medical diagnoses, and recipe histories remain 100% inside your building.
  • Adaptive AI Specialist: The local AI engine adapts instantly to user selections (e.g., pregnancy, kidney disease, diabetes).
  • Fully Local RAG Intelligence: Bypasses slow and risky cloud queries by retrieving context entirely from in-house databases.
🔒 Air-Gapped Advantage
"Unlike ChatGPT or consumer dietary trackers, Local Food AI never leaks clinic logs or clinician search strings to third-party databases."
Local Food AI - User Presentation 2 / 9
Access Control LOGIN & SECURITY

Secure Authentication & Password Recovery

  • Bcrypt Hashed Credentials: Heavy database level encryption blocks unauthorized entry.
  • Sidebar Session Status: Displays the active user, network mode, and active LLM version details.
  • Outbound Email Recovery: Integrated password reset routing via secure internal SMTP relays.
👤 Cookie Timeout
Secures sessions on shared ward computers, preventing patient detail access.
🔑 Self-Serve Resets
Request reset loops dispatch direct recovery keys to the clinician's inbox.
Local Food AI - User Presentation 3 / 9
Clinical Search DATABASE FILTERS

Tab 1: High-Speed Database Exploration

  • Real-Time Searching: Access millions of OpenFoodFacts entries in under 0.04 seconds.
  • Dynamic Warnings: Automatically overlays warning indicators based on active patient profile constraints.
  • Flexible Column Config: Custom selections allow dietitians to inspect precise vitamins and minerals.
🥗 Product Search Cards
Unpasteurized Brie High Risk

Listeria Risk: Avoid raw/unpasteurized soft cheeses during pregnancy.

Raw Spinach Recommended

High Iron & Folate: Encourages fetal cell growth.

Local Food AI - User Presentation 4 / 9
Plate Formulation MY PLATE BUILDER

Tab 2: My Plate Builder & Portion Parser

  • Natural Unit Parser: Clinicians can enter quantities in plain English (e.g. `1.5 cups`, `2 tablespoons`).
  • Density Conversions: Converts volume measurements into metric grams based on food density maps.
  • Cumulative Totals: Displays real-time compound macro updates (Proteins, Carbs, Fats).
📊 Macro intake vs Thresholds
Protein (70g Target) 52g (74%)
Sodium (2000mg Limit) 420mg (21%)
Local Food AI - User Presentation 5 / 9
Interactive AI CONSULTATION CHAT

Tab 3: Clinician Consultation Portal

  • Client Profile Dropdown: Select medical conditions (Diabetes, Anemia, Hypertension, Pregnancy).
  • Verified database queries (RAG): The local model queries local food data first to ensure factual responses.
  • Chain-of-Thought Explanation: Explains details step-by-step to help clinicians verify reasoning path safety.
> Can my diabetic client eat dates?
<thought> Checking glycaemic index... Dates contain ~66g sugar per 100g. High GI load. Injected profile is Diabetes. </thought>
Dates are high in simple sugars (GI ~103). Recommend limiting portion to <20g and pairing with almonds to slow absorption.
Local Food AI - User Presentation 6 / 9
Diet Plans AI MEAL PLANNER

Tab 4: Automated Clinical Meal Planner

  • Daily Caloric Targets: Input targets (e.g. `2000 kcal`) to scale portion recommendations.
  • Diet constraints integration: Enter restrictions (e.g. Halal, Vegan, Low-Fat, Kosher).
  • Printable Grid output: Formats the entire breakdown (Breakfast, Lunch, Dinner) into clean tables.
Meal Proposed Food Kcal
Breakfast Oatmeal with Walnuts & Bananas 450
Lunch Quinoa & Roasted Veggie Salad 650
Dinner Lentil Dahl with Brown Rice 750
Local Food AI - User Presentation 7 / 9
Security PRIVACY GUARANTEED

How Your Patient Data is Protected

  • No External API Transmissions: Patient data stays locally in memory. Zero external sharing.
  • Anonymous SearXNG Meta-Search: External searches are stripped of cookies and clinic source tags.
  • Isolated Networks: System nodes run in private networks, blocking external access.
🔒 Hospital-Grade Data Security
All processing occurs inside your facility's firewall. Local databases and local AI engines guarantee complete privacy compliance.
Local Food AI - User Presentation 8 / 9
Application Use WORKFLOW STEPS

Step-by-Step Clinical Workflow

1. Patient Profile
Select active health restrictions and medical conditions in the sidebar selector.
2. Build Diet Plate
Search foods, add to plate, and adjust portions using natural volumes.
3. Consult AI
Generate meal plans, check safety flags, and view the AI's clinical reasoning.
Local Food AI - User Presentation 9 / 9
Slide 1 of 9