Sometimes sustainability isn't about doing something grand—it's about starting small.
We're tackling food waste, one produce shelf at a time.
Every year, 40% of food in America goes to waste—that's $408 billion worth of food thrown away. Meanwhile:
- Grocery stores lose money on expired produce
- Customers miss out on deals on perfectly good food
- Our planet suffers from unnecessary environmental impact
The current approach? Random discounts, generic sales, and hoping someone buys it. It doesn't work.
We're bringing intelligence to the edge of the supply chain—right where food meets shelf.
Cameras point directly at produce shelves, continuously monitoring fruits, vegetables, and perishables in real-time using:
- YOLO object detection to identify what's on the shelf
- ResNet freshness model scoring items from 100 (just arrived) to 0 (spoiled)
- Gemini blemish detection for precise quality assessment
Our AI detects visual changes—browning bananas, soft spots on tomatoes, wilting lettuce—creating a real-time decay map of all inventory.
Customers download our app and connect their bank/cards through Knot API, which reveals:
- Complete purchase history across all grocery stores
- What they buy, how often, from which stores
- Price sensitivity and spending patterns
The system learns: "User A buys avocados 2x per week, always organic, usually spends $3-4 each" and "User B bought strawberries 5 times last month, only when under $5."
When cameras detect produce entering the "risky" freshness zone (70% freshness, 12 hours until spoilage):
- AI automatically calculates dynamic discounts (lower freshness = deeper discount)
- System identifies customers who actually buy that item
- Push notification: "Your favorite organic avocados now 40% off at Store X, perfect ripeness for tonight!"
No spam. No random deals. Only relevant offers for groceries you were already going to buy.
Our recommendation engine powered by xAI's Grok analyzes:
- Purchase Pattern Analysis - When are they due for their next shopping trip?
- Timing Relevance - Weekend bulk buyers vs. weekday shoppers
- Value Perception - Does the discount exceed their usual threshold?
- Behavioral Triggers - Seasonal preferences, health choices, new varieties
- Urgency Factors - Limited quantity, time sensitivity, expiration proximity
Example reasoning: "Perfect for meal prep Sunday - matches weekly buying pattern, 30% discount exceeds usual 20% threshold, last purchased 8 days ago (usual cycle: 10 days)"
Store dashboard provides actionable insights:
- "Based on current decay rates and customer buying patterns, reduce next banana order by 30%"
- "Your avocado customers also shop at Whole Foods - stock more organic"
- "50 customers notified about strawberries but only 10 bought - discount wasn't deep enough"
For items that won't sell even with targeted discounts:
- System auto-schedules food bank donations 12 hours before critical decay
- Customers who referred the food bank earn "waste warrior" points
- Community impact tracking shows pounds of food saved
Everything is queryable:
- Stores ask: "Which products have the worst sell-through rate at 60% freshness?"
- Customers ask: "How much could I save monthly based on my shopping patterns?"
- Instant AI-powered insights with context
graph TB
subgraph "Edge Layer"
CAM[📹 Smart Cameras<br/>YOLO + ResNet + Gemini]
end
subgraph "Backend Services"
API[🔧 Flask API<br/>REST + WebSocket]
DB[(🗄️ SQLite DB<br/>Inventory + Customers)]
AI[🤖 xAI Grok<br/>Recommendation Engine]
end
subgraph "External APIs"
KNOT[🔗 Knot API<br/>Purchase History]
end
subgraph "User Interfaces"
ADMIN[👨💼 Admin Dashboard<br/>React + TypeScript]
CUSTOMER[👤 Customer Portal<br/>React + TypeScript]
end
CAM -->|Real-time Detection| API
API <-->|Store/Query| DB
API <-->|Get Recommendations| AI
API <-->|Sync Purchase Data| KNOT
API <-->|WebSocket Updates| ADMIN
API <-->|WebSocket Notifications| CUSTOMER
style CAM fill:#4ade80,stroke:#22c55e,color:#000
style API fill:#60a5fa,stroke:#3b82f6,color:#000
style DB fill:#a78bfa,stroke:#8b5cf6,color:#000
style AI fill:#fbbf24,stroke:#f59e0b,color:#000
style KNOT fill:#fb923c,stroke:#f97316,color:#000
style ADMIN fill:#ec4899,stroke:#db2777,color:#000
style CUSTOMER fill:#06b6d4,stroke:#0891b2,color:#000
sequenceDiagram
participant Camera as 📹 Smart Camera
participant API as 🔧 Backend API
participant DB as 🗄️ Database
participant Grok as 🤖 xAI Grok
participant Knot as 🔗 Knot API
participant Customer as 👤 Customer App
Camera->>API: Detect banana @ 65% freshness
API->>DB: Update freshness score
API->>DB: Calculate discount (35% off)
alt Discount >= 20%
API->>DB: Query customers
API->>Knot: Fetch purchase patterns
Knot-->>API: Sarah buys bananas weekly
API->>Grok: Analyze match quality
Grok-->>API: Priority: 92 (Perfect match)
API->>DB: Create recommendation
API->>Customer: 🔔 Push notification
Customer-->>API: View notification
API->>DB: Mark as viewed
end
Note over Camera,Customer: Real-time processing: ~2-3 seconds
flowchart LR
subgraph "Computer Vision Pipeline"
IMG[📸 Camera Frame] --> YOLO[YOLO v8<br/>Object Detection]
YOLO --> CROP[Crop Objects]
CROP --> RESNET[ResNet18<br/>Freshness 0-100]
CROP --> GEMINI[Gemini Robotics<br/>Blemish Detection]
end
subgraph "Decision Engine"
RESNET --> SCORE[Freshness Score]
GEMINI --> BLEMISH[Blemish Count]
SCORE --> DISCOUNT[Calculate<br/>Dynamic Discount]
BLEMISH --> DISCOUNT
end
subgraph "Recommendation AI"
DISCOUNT --> FILTER{Discount<br/>>= 20%?}
FILTER -->|Yes| GROK[xAI Grok<br/>Match Customers]
FILTER -->|No| SKIP[Skip]
GROK --> NOTIFY[Push Notification]
end
style IMG fill:#4ade80,stroke:#22c55e
style YOLO fill:#60a5fa,stroke:#3b82f6
style RESNET fill:#a78bfa,stroke:#8b5cf6
style GEMINI fill:#fbbf24,stroke:#f59e0b
style GROK fill:#fb923c,stroke:#f97316
style NOTIFY fill:#ec4899,stroke:#db2777
journey
title Customer Experience with EdgeCart
section Discovery
Download app: 5: Customer
Connect Knot account: 4: Customer
Sync purchase history: 5: System
section First Deal
Camera detects ripe avocado: 5: System
AI matches with customer: 5: System
Push notification sent: 5: System
Customer sees deal: 4: Customer
section Purchase
Customer visits store: 4: Customer
Buys discounted avocado: 5: Customer
Saves 40%: 5: Customer
section Impact
Feel good about waste: 5: Customer
See savings dashboard: 5: Customer
Get another deal: 5: System
mindmap
root((EdgeCart))
Backend
Flask REST API
WebSocket Server
SQLite Database
Python 3.11
AI/ML
YOLOv8
Object Detection
Real-time Inference
ResNet18
Transfer Learning
Freshness Classification
Gemini Robotics
Blemish Segmentation
Quality Assessment
xAI Grok
NLP Recommendations
Context Understanding
Frontend
React + TypeScript
Framer Motion
WebSocket Client
Custom Terminal UI
External APIs
Knot API
Purchase History
Customer Profiles
Multi-merchant Data
- ✅ Reduce food waste by 30-50%
- ✅ Recover revenue from inventory that would expire
- ✅ Better inventory forecasting based on actual customer demand
- ✅ Turn waste into profit
- ✅ Save 30-50% on groceries they already buy
- ✅ Get notified only about items they actually want
- ✅ Buy produce at optimal ripeness for their needs
- ✅ Feel good about reducing waste
- 🌍 Less food in landfills = reduced methane emissions
- 💧 Conserve water used to grow wasted food
- 🌱 More efficient food supply chain
- ♻️ Circular economy approach to grocery retail
- Flask - REST API & WebSocket server
- Flask-SQLA - Database ORM
- OpenCV + YOLO - Real-time object detection
- PyTorch + ResNet - Freshness classification model
- Google Gemini - Blemish segmentation & detection
- Knot API - Customer purchase data integration
- xAI Grok - AI-powered recommendation engine
- SQLite/PostgreSQL - Database
- React - Modern component-based UI
- WebSocket - Real-time updates
- Framer Motion - Smooth animations
- Custom Terminal UI - Cyberpunk aesthetic
- YOLOv8 - Object detection for identifying produce
- ResNet18 - Transfer learning for fresh/rotten classification
- Gemini Robotics - High-precision blemish detection
- xAI Grok - Natural language recommendations
# Python 3.11+
python --version
# Node.js 18+
node --versioncd backend
# Create virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Set up environment variables
cp .env.example .env
# Edit .env with your API keys:
# - KNOT_CLIENT_ID (from Knot)
# - KNOT_SECRET (from Knot)
# - XAI_API_KEY (from xAI)
# - GEMINI_API_KEY (from Google)
# Initialize database with sample data
POPULATE=true python main.pycd frontend
# Install dependencies
npm install
# Start development server
npm run dev- Admin Dashboard: http://localhost:5173
- Customer Portal: http://localhost:5173/#/customer
- Backend API: http://localhost:3000
Try these pre-loaded customer profiles (no Knot API key needed):
Health-conscious mom who loves organic berries
- Buys strawberries, blueberries, spinach regularly
- Shops at Instacart and Whole Foods
- Average spend: $45-50 per trip
Fitness enthusiast buying citrus for smoothies
- Buys oranges, grapefruits, kale 2x per week
- Shops at Walmart and Target
- Average spend: $22-35 per trip
Foodie exploring exotic fruits
- Buys grapes, dragon fruit, kiwi occasionally
- Shops at specialty stores
- Average spend: $30-40 per trip
- 📹 Live Camera Feed with real-time fruit detection
- 📊 Inventory Management with freshness scores
- 🎨 Blemish Detection with visual highlighting
- 📈 Analytics - Waste prevention metrics
- 🔔 WebSocket Updates - Real-time inventory changes
- 📉 Predictive Ordering - AI-suggested inventory adjustments
- 🛒 Personalized Deals based on Knot purchase history
- 🔔 Smart Notifications via WebSocket (only relevant items)
- 📜 Purchase History from both Knot and EdgeCart
- 💰 Savings Tracker showing total money saved
- 🌍 Impact Dashboard showing food waste prevented
- 🎯 AI Reasoning explains why each deal was recommended
- 🧠 Multi-dimensional Analysis (timing, value, behavior, urgency)
- 🎯 Priority Scoring (80-100 for perfect matches)
- 📖 Natural Language Reasoning for each recommendation
- ⚡ Rate-Limited to optimize API costs (10s between calls)
- 🔄 Feedback Loop learns from customer responses
stores fruit_inventory freshness_status
- id - id - id
- name - store_id - inventory_id
- location - fruit_type - freshness_score (0-1.0)
- quantity - predicted_expiry_date
- original_price - discount_percentage
- current_price - status (fresh/ripe/clearance)
- thumbnail_path - last_checked
customers purchase_history recommendations
- id - id - id
- knot_customer_id - customer_id - customer_id
- name - inventory_id - inventory_id
- email - quantity - priority_score
- preferences (JSON) - price_paid - reason (JSON)
- discount_applied - viewed
- knot_transaction_id - purchased
quantity_change_log waste_log
- id - id
- inventory_id - inventory_id
- old_quantity - quantity_wasted
- new_quantity - estimated_value_loss
- change_type - reason
- freshness_score - logged_at
Create a .env file in the backend/ directory:
# Knot API (for customer data)
KNOT_CLIENT_ID=your_knot_client_id
KNOT_SECRET=your_knot_secret
KNOT_ENV=tunnel # tunnel, dev, or prod
KNOT_USE_REAL=true
# xAI (for AI recommendations)
XAI_API_KEY=your_xai_api_key
# Google Gemini (for blemish detection)
GEMINI_API_KEY=your_gemini_api_key
# Database
DATABASE_URL=sqlite:///edgecart.db
# Server
PORT=3000
POPULATE=true # Seed with sample data on first runMIT License - See LICENSE file for details