🏠 EverShelf

Self-hosted pantry management system — Track your food inventory, scan barcodes, get AI-powered recipe suggestions, and reduce waste.

---

🚀 Try the live demo — no installation required!

▶ Open Live Demo  ·  🌐 Project Website  ·  📖 Wiki

The demo runs with mock pantry data. AI features are fully enabled. All write operations are safely sandboxed.

---

---

✨ Features

📦 Inventory Management

• Barcode scanning — Scan products with your phone camera using QuaggaJS
• AI identification — Take a photo and let Google Gemini identify the product, with suggestions from your existing inventory; gracefully shows a friendly message when AI quota is exhausted instead of a raw API error
• Smart locations — Track items across Pantry, Fridge, Freezer, and custom locations
• Expiry tracking — Automatic shelf-life estimation based on product type and storage
• Opened product tracking — Reduced shelf-life calculation when packages are opened; opened-product expiry is now also checked when building banner alerts (not just the dashboard section)
• Vacuum-sealed support — Extended expiry dates for vacuum-sealed items
• Anomaly detection — Banner alerts for suspicious quantities and consumption predictions with inline correction; dismiss button now shows the current inventory quantity so the action is unambiguous ("La quantità è giusta (2 pz)")

🤖 AI-Powered (Google Gemini)

• Expiry date reading — Photograph a label and extract the expiry date automatically
• Product identification — Point your camera at any product for instant recognition
• Existing product matching — AI scan shows matching products already in your pantry before suggesting new ones
• Storage & shelf-life hint — When adding a new product, Gemini suggests the optimal storage location and shelf-life in the background; shown as an inline AI badge next to the expiry estimate
• Recipe generation — Get personalized recipes based on what's in your pantry; streams live via Server-Sent Events so results appear as they are generated
• Smart chat assistant — Ask questions about your inventory, get cooking tips
• Shopping suggestions with tips — AI-powered purchase recommendations, each enriched with a short practical buying/storing tip
• Anomaly explanation — "🤖 Spiega" button on anomaly banners explains in plain language why a discrepancy likely occurred and what to do
• Model fallback — All AI endpoints try gemini-2.5-flash first and fall back to gemini-2.0-flash automatically
• Graceful no-key state — When no Gemini key is configured, AI entry points show a friendly message; the header button is visually greyed with an amber dot

🛒 Shopping List

• Bring! integration — Sync with the Bring! shopping list app
• Generic shopping names — Products are grouped by type ("Latte", "Affettato", "Panna da cucina") rather than brand, keeping the Bring! list clean and consolidated
• Smart predictions — Know what you'll need before you run out
• Auto-add on depletion — When a product reaches zero the app adds it to Bring! automatically, no confirmation needed
• Auto-remove on scan — Products are removed from the shopping list when scanned in - Auto-migration — Items already on the Bring! list are silently renamed to their generic name in the background (throttled, runs on list load)
  • Catalog coverage — All product types resolve to a German Bring! catalog key for icon and category display in the Bring! app

🍳 Cooking Mode

• Step-by-step guidance — Follow recipes with a hands-free cooking interface
• Text-to-Speech — Voice readout of recipe steps; supports browser Web Speech API, native Android TTS (kiosk), or a custom REST endpoint (Home Assistant, etc.); retries voice loading for up to 10 seconds with a fallback refresh button; TTS activates automatically without requiring the global TTS setting to be enabled
• Auto-read on navigate — Each step is read aloud automatically when you tap Next or Previous; the first step is read when entering cooking mode
• Timer voice alerts — 10-second countdown warning spoken aloud before each timer expires; expiry announced vocally when time is up
• Recipe completion — "Buon appetito!" spoken when the last step is confirmed
• Built-in timer — Automatic timer suggestions based on recipe instructions
• Ingredient tracking — Mark ingredients as used during cooking; leftover quantities prompt a "move to another location" flow

📊 Dashboard

• Waste tracking — Monitor consumed vs. wasted products over 30 days
• Anti-waste report — Personalised waste rate vs. national average with annual kg estimate; shown above the expiring-items list
• Expiry alerts — Visual warnings for expired and soon-to-expire items
• Opened products panel — Tracks partially-used items; expiry is recalculated from the opening date using AI (Gemini) + per-category rule fallback; whole sealed packages always keep their original manufacturer expiry; conf items with mixed whole + fractional units are shown as two separate entries
• Freezer shelf-life — Granular per-product estimates (USDA/EFSA): fish 120 d, poultry 270 d, whole red-meat cuts 365 d, mince 120 d, vegetables/fruit 270 d, generic 180 d; AI + cache still take priority over rules
• Safety ratings — Smart assessment of expired product safety (by category and location); expired unsafe items shown with a red danger banner and "L'ho buttato" as the primary action
• Expired product banner — Products that have passed their effective shelf-life (including opened-product reduced expiry) appear in the top notification banner; icon, colour and title adapt to the actual safety level (✅ green for safe, 👀 amber to check, 🚫 red for danger); high-risk items get a prominent discard action
• Quick recipe bar — One-tap recipe suggestion using expiring products
• Anomaly banner — Scrollable banner with suspicious quantities and consumption prediction mismatches, with one-tap correction or inline edit
• Expired/expiring alerts — Priority-sorted banner notifications for expired and soon-to-expire products with use, throw, edit, and dismiss actions
• Swipe navigation — Touch swipe or tap arrows/dots to browse banner notifications
• Quick-access buttons — Recently used and most popular products shown on the inventory page for fast access

📱 Progressive Web App

• Mobile-first design — Optimized for phones, works on tablets and desktop
• Installable — Add to home screen for a native app experience
• Multi-device — Settings and data sync across devices on the same server

⚖️ Smart Scale Integration (Add-on)

• Bluetooth gateway — Connects a BLE smart scale to EverShelf via local WebSocket
• SSE relay — Server-side relay avoids mixed-content (HTTPS→WS) issues
• Auto-discovery — Server scans LAN to find the gateway automatically
• Auto weight reading — When adding/using a product with unit g/ml, weight fills automatically
• 10g threshold — Ignores readings that haven't changed enough between products - Duplicate-reading prevention — Server-side 12-second dedup window rejects a second scale-triggered deduction of the same product, guarding against BLE multi-fire- ml conversion hint — Shows "weight in grams → will be converted to ml" when product unit is ml
• Stability + auto-confirm — 10s stable wait + 5s countdown before confirming
• Real-time status — Scale connection indicator always visible in the header
• Multi-protocol — Supports Bluetooth SIG Weight Scale, Body Composition, Xiaomi Mi Scale 2 and 100+ models
• Built into kiosk (v1.6.0+) — BLE gateway runs as an integrated foreground service inside the EverShelf Kiosk app; no separate APK needed. The standalone gateway app in evershelf-scale-gateway/ is deprecated but kept for non-kiosk use cases.

📺 Android Kiosk Mode (Add-on)

• Dedicated tablet app — Full-screen WebView wrapper for wall-mounted kitchen tablets
• True kiosk lock — Screen pinning blocks home/recent buttons
• Setup wizard — 6-step guided configuration (language, welcome, permissions, server URL, BLE scale scan, screensaver, summary)
• Smart auto-discovery — Scans the LAN in parallel (60 threads, TCP pre-check, ports 80/443/8080/8443) with real-time UI feedback; correctly identifies the device's Wi-Fi/Ethernet subnet (VPN and cellular interfaces are filtered out)
• Built-in BLE scale gateway — GatewayService foreground service; BLE scanning + WebSocket server :8765 run directly inside the kiosk app. Select your scale in step 5 of the wizard — no external app required
• Scale auto-configuration — After selecting the BLE device, the wizard writes scale_enabled and scale_gateway_url=ws://127.0.0.1:8765 to the server automatically
• Camera & mic permissions — Full hardware access for barcode scanning and voice; grant button transforms to a green confirmation after granting
• Native TTS bridge — Cooking mode voice readout uses the Android TextToSpeech engine directly, bypassing Web Speech API voice limitations; no offline voice packs required
• Hard refresh — ↻ button clears WebView cache to pick up web app updates
• Update notifications — Checks GitHub releases every 6h, shows banner when updates available
• SSL support — Accepts self-signed certificates
• Android kiosk app — evershelf-kiosk/ — downloadable APK

---

🚀 Quick Start

Prerequisites

• Web server with PHP 8.0+ (Apache or Nginx)
• PHP extensions: pdo_sqlite, curl, mbstring, json
• HTTPS recommended (required for camera access on mobile)

Installation

Option A: Docker (recommended)

# 1. Clone the repository
git clone https://github.com/dadaloop82/EverShelf.git
cd EverShelf

# 2. Create configuration file
cp .env.example .env
nano .env

# 3. Start with Docker Compose
docker compose up -d

# → Open http://localhost:8080

Option B: Manual

# 1. Clone the repository
git clone https://github.com/dadaloop82/EverShelf.git
cd EverShelf

# 2. Create configuration file
cp .env.example .env

# 3. Set permissions
chmod 755 data/
chmod 664 data/.gitkeep
chown -R www-data:www-data data/

# 4. Edit your configuration
nano .env

Configuration (.env)

# Required for AI features (get a key at https://aistudio.google.com/app/apikey)
GEMINI_API_KEY=your_api_key_here

# Optional: Bring! shopping list integration
BRING_EMAIL=your_email@example.com
BRING_PASSWORD=your_password

# Optional: Text-to-Speech for cooking mode
TTS_URL=http://your-home-assistant:8123/api/events/tts_speak
TTS_TOKEN=your_long_lived_token
TTS_ENABLED=true

# Optional: Security — protect the save_settings endpoint
# Set a strong random string; the Settings UI will ask for it before saving
SETTINGS_TOKEN=

# Optional: Demo mode — block all write operations at the router level
DEMO_MODE=false

Web Server Configuration

Apache (.htaccess)

The app works out of the box with Apache if placed in the web root or a subdirectory. Make sure mod_rewrite is enabled and AllowOverride All is set.

<Directory /var/www/html/evershelf>
    AllowOverride All
    Require all granted
</Directory>

Nginx

server {
    listen 80;
    server_name your-server.local;
    root /var/www/html/evershelf;
    index index.html;

    location /api/ {
        try_files $uri $uri/ =404;
        location ~ \.php$ {
            fastcgi_pass unix:/run/php/php8.2-fpm.sock;
            fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
            include fastcgi_params;
        }
    }

    # Deny access to sensitive files
    location ~ /\.env { deny all; }
    location ~ /data/ { deny all; }
    location ~ /backup\.sh { deny all; }
}

HTTPS Setup (Recommended)

Camera access requires HTTPS on most mobile browsers. Options:

• Let's Encrypt with Certbot (for public-facing servers)
• Self-signed certificate (for local network only)
• Reverse proxy (e.g., Caddy, Traefik) with automatic TLS

Cron Job (Optional)

Set up a cron job for smart shopping predictions:

# Run every 5 minutes
*/5 * * * * php /path/to/evershelf/api/cron_smart_shopping.php >> /path/to/evershelf/data/cron.log 2>&1

Backup (Optional)

The included backup.sh creates local daily backups of your database:

# Run daily at 3 AM
0 3 * * * /path/to/evershelf/backup.sh

---

🏗️ Architecture

evershelf/
├── index.html              # Single-page application (SPA)
├── manifest.json           # PWA manifest
├── .env.example            # Configuration template
├── backup.sh               # Local database backup script
├── LICENSE                 # MIT License
│
├── api/
│   ├── index.php           # Main API router (all endpoints)
│   ├── database.php        # SQLite schema, migrations, helpers
│   └── cron_smart_shopping.php  # Background job for predictions
│
├── assets/
│   ├── css/style.css       # All application styles
│   ├── js/app.js           # All application logic
│   └── img/                # Static images
│
└── data/                   # Runtime data (gitignored)
    ├── evershelf.db         # SQLite database (auto-created)
    ├── backups/            # Local DB backups
    └── *.json              # Token/cache files

evershelf-scale-gateway/    # ⚖️ Android BLE gateway [DEPRECATED — integrated into kiosk v1.6.0+]
    ├── README.md           # Deprecation notice + legacy docs
    └── app/src/            # Kotlin Android source (WebSocket + BLE)

evershelf-kiosk/            # 📺 Android kiosk app (add-on)
    ├── README.md           # Setup & feature docs
    └── app/src/            # Kotlin Android source (WebView wrapper)

API Endpoints

Category	Action	Method	Description
Products	search_barcode	GET	Find product by barcode
	lookup_barcode	GET	Look up barcode on Open Food Facts
	product_save	POST	Create or update a product
	products_list	GET	List all products
Inventory	inventory_list	GET	List inventory items
	inventory_add	POST	Add product to inventory
	inventory_use	POST	Use/consume from inventory
	inventory_summary	GET	Count by location
AI	gemini_identify	POST	Identify product from photo
	gemini_expiry	POST	Read expiry date from photo
	gemini_chat	POST	Chat with AI assistant
	generate_recipe	POST	Generate recipe from inventory
	gemini_product_hint	POST	Storage location + shelf-life hint
	gemini_shopping_enrich	POST	Enrich shopping suggestions with tips
	gemini_anomaly_explain	POST	Plain-language anomaly explanation
Shopping	bring_list	GET	Get Bring! shopping list
	bring_add	POST	Add items to Bring!
	smart_shopping	GET	Smart shopping predictions
Settings	get_settings	GET	Get server configuration
	save_settings	POST	Update server configuration

---

🔒 Security Notes

• Credentials are stored in .env (server-side, never committed to Git)
• Database stays local — never pushed to remote repositories
• API keys are never exposed to the browser — get_settings returns only boolean flags (gemini_key_set, settings_token_set), never raw key values
• Settings write protection — set SETTINGS_TOKEN in .env to require a secret token (X-Settings-Token header) for all save_settings calls; validated with hash_equals to prevent timing attacks
• Demo / public mode — set DEMO_MODE=true to block all write operations at the PHP router level before any business logic runs
• The API uses parameterized SQL queries (PDO prepared statements) against injection
• Input validation on all inventory operations (quantity bounds, location whitelist)
• Consider adding reverse-proxy authentication (e.g. Authelia, Nginx auth_basic) if the server is accessible from the internet

---

🛠️ Development

# Run PHP's built-in server for local development
php -S localhost:8080 -t /path/to/evershelf

# Check PHP syntax
php -l api/index.php
php -l api/database.php

The application uses no build tools — edit files directly and refresh.

---

📋 Roadmap

High Priority

• Cooking mode — 3D wheel JS — swipe navigation, gyroscope tilt, haptic feedback
• Cooking mode — step timers — auto-detect "X minutes" in recipe steps, countdown + alert
• Push notifications — daily expiry alerts via PWA Service Worker + VAPID
• Quick search / quick-add bar — always-visible search above the nav, PWA shortcuts

Medium Priority

• Receipt OCR → bulk add — photo of receipt → Gemini Vision → auto-fill inventory
• CSV/JSON export & import — download/upload inventory from Settings
• Custom storage locations — user-defined locations beyond Fridge/Freezer/Pantry
• Multi-user support — PIN-based user distinction, action log with user label
• AI optimal purchase prediction — suggest "buy X units of Y within Z days"
• Price history sparklines — per-product price chart from the AI cache data

Low Priority / Nice to Have

• Dark mode — CSS custom properties are already structured to support it
• Full offline mode — Service Worker cache to show inventory read-only when server is down
• French & Spanish translations (fr.json, es.json)
• Swipe actions on inventory rows — swipe left to use/discard, right to edit
• PHP unit tests — PHPUnit coverage for shelf-life, price calc, and key helpers

Completed ✅

• ✅ AI price estimation in shopping list
• ✅ Server heartbeat + offline banner
• ✅ In-app bug reporter → automatic GitHub issue creation
• ✅ Cooking mode (start, steps, 3D wheel CSS)
• ✅ Kiosk ⚙️ Settings overlay button (replaces Android native button)
• ✅ Adaptive consumption anomaly detection
• ✅ CI/CD pipeline (PHP lint, JS lint, Docker build, Trivy security scan)

---

🌐 Translations

The app supports multiple languages via JSON translation files in the translations/ folder.

Language	Status
🇮🇹 Italian (it)	✅ Complete (base)
🇬🇧 English (en)	✅ Complete
🇩🇪 German (de)	✅ Complete

Want to add your language? See the Translation Guide — just copy translations/it.json, translate the values, and submit a PR!

---

🤝 Contributing

Contributions are welcome! See CONTRIBUTING.md for detailed guidelines.

1. Fork the repository
2. Create a feature branch (git checkout -b feature/my-feature)
3. Commit your changes (git commit -m 'Add my feature')
4. Push to the branch (git push origin feature/my-feature)
5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

👨‍💻 Author

Stimpfl Daniel — evershelfproject@gmail.com

• Website: evershelfproject.dadaloop.it
• GitHub: @dadaloop82

---

📸 Screenshots

For a live walkthrough with real data and full AI enabled, visit the live demo — no installation required.

Want to contribute a GIF or screenshots? See CONTRIBUTING.md — PRs welcome!