NutriSnap — Take back control of your plate.

We only get one body. And 80% of the work of losing weight, building muscle, or simply feeling better happens at the table — not at the gym.

Most nutrition apps ask you to create an account, hand over your data, and trust that a remote server keeps your most intimate health information safe. NutriSnap takes the opposite approach: your data never leaves your device.

---

What it does

Snap a photo of your meal. In seconds, Gemini AI identifies the food, estimates calories, and breaks down macronutrients — proteins, carbs, fats. Your daily targets are calculated from your actual metabolic profile (Mifflin-St Jeor or Katch-McArdle, depending on your data). No generic 2000 kcal default.

   

---

Features

Feature	Detail
AI meal scanning	Photo → Gemini 2.5 Flash → calories + macros in seconds
Metabolic engine	BMR, TDEE, macro targets calculated from your actual profile
Daily AI recap	Coach summary of your day, cached locally — no repeat calls
Multi-day analysis	7-day or 30-day trend analysis with AI coaching
Meal history	Monthly calendar with color-coded daily intake
Meal detail	Photo, macros, AI summary and coach tip per meal
Privacy-first	Zero server, zero telemetry, zero account — IndexedDB only
PWA	Installable on iOS & Android, works offline
Dark / Light / System	Theme persisted per device
FR / EN	Language switch in profile settings
BYOK	Bring your own Gemini API key — free tier available

---

Privacy model

NutriSnap has no backend. There is no server to breach, no database to leak, no account to compromise.

• All data (profile, meals, photos, settings) lives in IndexedDB on your browser.
• The only external call is to the Google Gemini API — initiated by you, using your own key, at the moment of each meal scan.
• No cookies, no analytics, no third-party SDKs beyond the Gemini client library.

The full privacy policy is available in-app at /privacy.

---

Getting started

Prerequisites

• Node.js 18+
• A Google Gemini API key (free tier available)

Local development

git clone https://github.com/your-username/nutrisnap.git
cd nutrisnap
npm install
npm run start        # ng serve — dev mode, no service worker

Deploy to a static host (OVH, Netlify, etc.)

npm run build        # production build → dist/nutrisnap/browser/

The build output is a static SPA. Drop it on any HTTPS host. The included .htaccess handles SPA routing and PWA cache-busting headers for Apache.

Auto-deploy to OVH via FTP (requires a .env file with FTP credentials):

cp .env.example .env   # fill in your FTP credentials
npm run start:sync-prod

This watches for file changes, rebuilds in production mode, and syncs to FTP automatically.

---

Stack

Layer	Tech
Framework	Angular 21 (standalone components, signals)
UI	Angular Material 3
Storage	Dexie.js (IndexedDB)
AI	Google Gemini 2.5 Flash (@google/generative-ai)
PWA	@angular/service-worker (auto-update snackbar)
Deploy	Static SPA — any HTTPS host

---

Project structure

src/app/
  core/
    services/     translate, theme, gemini, log, storage, update
    models/       profile.model, meal.model
  features/
    onboarding/   5-step setup: profile → goal → API key → AI coach
    dashboard/    daily stats ring, macro cards, meal timeline
    scanner/      camera + gallery + Gemini analysis
    history/      monthly calendar + day panel + AI trend analysis
    meal-detail/  photo, macros, coach tip, delete
    profile/      edit profile, theme, language, data export/import
    privacy/      full GDPR policy page
  shared/
    components/   NsCardComponent, RecapSheetComponent, LegalFooterComponent
    pipes/        TranslatePipe
public/
  assets/i18n/    fr.json, en.json

---

AI architecture

BYOK (Bring Your Own Key): the Gemini API key is stored in IndexedDB on your device. It is passed directly from your browser to Google's API — NutriSnap never sees it, and it never touches any intermediate server.

Prompts used:

• analyzeMeal — image(s) + daily context → JSON (food name, calories, macros, coach tip, confidence score)
• getDailyRecap — today's logs + profile → plain-text coach summary (cached in DB to avoid repeat calls)
• getWeeklyAnalysis — aggregated n-day logs → plain-text trend analysis
• getCoachFeedback — metabolic profile → onboarding coach message

---

Modules

Module	Status
M1 — Onboarding (5 steps)	✅
M2 — Data engine (Dexie v2)	✅
M3 — Gemini meal scanner	✅
M4 — Dashboard & dark mode	✅
M5 — PWA / Offline	✅
M6 — Meal detail + History calendar	✅
M7 — AI Coach (daily recap + trend analysis)	✅
M8 — i18n FR/EN	✅

---

License

MIT — do whatever you want with it.

---

Built with Claude Code by Dorian Naaji.

Tags: nutrition calorie-tracker macro-tracker pwa privacy-first gemini-ai angular indexeddb byok offline no-backend health fitness meal-scanner ai · suivi nutritionnel compteur calories application nutrition intelligence artificielle vie privée sans serveur installable