DiaLog – Smart Diabetes Meal Analyzer

DiaLog helps people with diabetes make safer food choices. It analyzes meals using machine learning and provides clear, personalized guidance before and after you eat.

Table of Contents

• Features
• Architecture
• Tech Stack
• Getting Started
• Run (Windows PowerShell)
• Multilingual Support 🇮🇳
• Safety & Nutrition Page
• API Overview
• Troubleshooting

Features

• Smart Meal Analysis (ML) with confidence and explanations
• Truly Personalized Recommendations based on your profile and logs
• Nutritional Facts: calories, carbs, protein, fats, fiber, GI/GL
• Real‑time Search: 1,000+ Indian foods with quick filter
• BMI & Health Insights surfaced in the dashboard
• Risk Assessment Badges: safe / caution / unsafe
• Safety & Nutrition Page to check food safety without logging
• Multilingual UI: popular Indian languages via live translation

Architecture

• Frontend: React 18, Tailwind CSS, Heroicons
• Backend: FastAPI, Pydantic, Uvicorn, joblib models
• ML: Random Forest/ensemble models + per‑user personalized models
• Data: Food Master Dataset (nutritional facts), User Logs (for personalization)
• Storage/Cloud (optional): Firebase Admin SDK for Firestore logging

See docs for visuals and flows:

• Architecture Diagram
• ML Workflow
• UI Wireframes

Tech Stack

• Frontend: React, Tailwind CSS, Heroicons
• Backend: FastAPI, Uvicorn, Pydantic
• ML: scikit‑learn, joblib
• Optional: Firebase Admin for Firestore logging

Getting Started

Prerequisites:

• Python 3.10+ (3.8+ supported)
• Node.js 16+ (18+ recommended)
• Git

Initial Setup (first time only)

Backend setup

# Navigate to backend directory
cd backend

# Create virtual environment (recommended)
python -m venv venv

# Activate virtual environment (Windows PowerShell)
./venv/Scripts/Activate.ps1

# Install Python dependencies
pip install -r requirements.txt

# Optional: Firebase Admin support (for Firestore logging)
npm install

Frontend setup

# Navigate to frontend directory
cd ../frontend

# Install Node.js dependencies
npm install

Run (Windows PowerShell)

Option 1: Start both servers together

# From project root directory
./start_dev.bat

Option 2: Start servers separately

# Backend terminal
cd backend; uvicorn main:app --reload --host 0.0.0.0 --port 8000

# Frontend terminal (new window)
cd frontend; npm start

Access URLs

• Backend (Production): https://dialog-backend.onrender.com
• Frontend (Production): https://dialog-frontend.onrender.com
• Local Backend API: http://localhost:8000
• Local Frontend Dev: http://localhost:3000 (or 3001 if 3000 is occupied)
• API Docs (Swagger): http://localhost:8000/docs

Deployment (Render)

Backend (Web Service)

• workingDirectory: backend
• buildCommand: pip install --upgrade pip ; pip install --only-binary=:all: -r requirements.prod.txt
• startCommand: uvicorn main:app --host 0.0.0.0 --port $PORT
• Health check path: /health
• Env vars: PYTHON_VERSION=3.11.9, plus any Firebase / Gemini keys.

Frontend (Static Site)

• Root Directory: frontend
• Build Command: npm ci && npm run build (or npm install && npm run build if no lockfile)
• Publish Directory: build
• REACT_APP_API_BASE_URL: https://dialog-backend.onrender.com
• Firebase env keys (if using Firestore features).

SPA Refresh Fix

Static hosting returns 404 on deep links by default. Added public/404.html which redirects to index.html and a script in src/index.jsx that restores the original path for React Router. If you use Render's rewrite settings, also add: Source /* → Destination /index.html (Rewrite).

Locking Dependencies (Optional Post‑Deploy)

After a successful deploy you can freeze versions for reproducibility:

pip freeze > backend/constraints.txt

Copy top‑level lines you care about into a pinned requirements.prod.lock.txt later. Keep the relaxed file for quick iteration until stability is confirmed.

Multilingual Support 🇮🇳

• Live translation across the app; choose your language in Profile or via the Navbar’s globe button.
• Popular Indian languages included: Hindi, Bengali, Marathi, Telugu, Tamil, Gujarati, Kannada, Malayalam, Punjabi, Odia, Urdu (plus English).
• Powered by a free translation API for live content; can be upgraded to i18next + resource files or alternate providers (Bhashini, Sarvam AI, LibreTranslate) as needed.

Safety & Nutrition Page

Check if a food is Safe / Caution / Unsafe before you eat it – without logging anything.

• UI: Navbar → “Safety & Nutrition” page
• Flow: type to search → pick a food → click “Check Safety”
• Backend: uses /food/{name} for nutrition facts and /predict for ML safety
• No side effects: nothing is written to Firestore logs from this page

API Overview

• GET /health – server/model status
• GET /foods – list of foods
• GET /food/{food_name} – nutrition for a food
• POST /predict – safety prediction for a meal
• POST /recommendations – ML meal recommendations
• POST /truly-personalized-recommendations – per‑user model recommendations

See also: docs/api-endpoints.md

Troubleshooting

• Frontend on “port 3000 in use”: it will auto‑select 3001.
• Backend “Could not import module 'main'”: run from backend folder or use uvicorn main:app.
• Ensure the backend runs on port 8000 to match frontend/src/services/api.js.