🌱 ReNova - GenAI-Powered Waste Intelligence Platform

Smart waste management through multimodal AI: Vision, Voice, and Dual-RAG-powered personalized recycling guidance

ReNova transforms waste disposal from a mundane chore into an engaging, rewarding experience. Using cutting-edge AI (CLIP vision, Whisper voice, Llama reasoning), we provide hyper-personalized recycling guidance that adapts to your city's regulations, your past behavior, and real-time recycler availability.

---

Team Members - Team - Minus (Ramaiah Institute Of Technology)

• Saksham Yadav - 7678369133
• Sagar S R - 9482209148
• Samrudh P - 7676296599
• Kushal L - 8660179391

🎯 Problem Statement

India generates 150,000 tonnes of waste daily, but only 60% is collected and 15% is processed. Key challenges:

• ❌ Lack of awareness: Citizens don't know how to segregate waste correctly
• ❌ No standardization: Each municipality has different rules (Patiala ≠ Mumbai)
• ❌ No incentives: Why should users bother recycling?
• ❌ Broken last-mile: Recyclers are hard to find, unverified, offer low rates
• ❌ Compliance burden: Businesses struggle with waste audit trails for ESG reporting

Result: Recyclables end up in landfills, contaminating soil and water, releasing methane (28x worse than CO₂).

---

💡 Our Solution

ReNova is a multimodal GenAI platform that makes waste management:

✅ Effortless - Scan item → Get instant guidance (vision AI)
✅ Accessible - Speak your question → Get answers (voice AI in 12 languages)
✅ Personalized - Learns your behavior, adapts to your location
✅ Rewarding - Gamified tokens redeemable for pickups, discounts
✅ Compliant - Auto-logs waste for ESG/CSR reporting
✅ Connected - Matches users with verified recyclers via smart marketplace

🎬 How It Works

1. USER → Scan waste (image) or ask question (voice)
2. CLIP → Identifies material, cleanliness, hazards
3. DualRAG → GlobalRAG : Retrieves city regulations + user history
             PersonalRAG : personal patterns 
4. LLM → Generates personalized disposal advice
5. GEO → Finds nearest verified recyclers
6. REWARD → User earns tokens based on material value

---

🏗️ Architecture

High-Level System Design

┌─────────────────────────────────────────────────────────────┐
│                     FRONTEND (React)                        │
│  ┌──────────┐  ┌───────────┐  ┌────────────┐  ┌─────────┐ │
│  │  Camera  │  │   Voice   │  │  Recycler  │  │  Stats  │ │
│  │   Scan   │  │   Query   │  │    Map     │  │ Dashboard│ │
│  └─────┬────┘  └─────┬─────┘  └──────┬─────┘  └────┬────┘ │
└────────┼─────────────┼────────────────┼─────────────┼──────┘
         │             │                │             │
         └─────────────┴────────────────┴─────────────┘
                              │
                    ┌─────────▼──────────┐
                    │   FastAPI Backend  │
                    │   (API Gateway)    │
                    └─────────┬──────────┘
                              │
         ┌────────────────────┼────────────────────┐
         │                    │                    │
    ┌────▼─────┐      ┌──────▼──────┐     ┌──────▼──────┐
    │  Vision  │      │    Voice    │     │  Reasoning  │
    │  Service │      │   Service   │     │   Service   │
    │  (CLIP)  │      │  (Whisper)  │     │  (Llama)    │
    └────┬─────┘      └──────┬──────┘     └──────┬──────┘
         │                   │                    │
         └───────────────────┴────────────────────┘
                              │
              ┌───────────────┼───────────────┐
              │               │               │
         ┌────▼────┐    ┌─────▼─────┐   ┌────▼─────┐
         │ MongoDB │    │  Faiss    │   │   OSM    │
         │ (Users, │    │  (Vector  │   │ (Geo +   │
         │ Scans)  │    │   RAG)    │   │ Routing) │
         └─────────┘    └───────────┘   └──────────┘

Why Dual RAG?

• ReNova uses two separate vector databases:

  Global RAG — shared knowledge base for all users Contains general waste management guidelines, recycling best practices, material classifications, disposal instructions, and environmental regulations. Built from curated documents and maintained by the platform. Read-only for users; updated by admins or automated ingestion pipelines.

  Personal RAG — user-specific knowledge base Stores documents uploaded by individual users (e.g., local recycling center guidelines, personal notes, community-specific rules). Each user has their own isolated vector space. Allows personalization: users in different cities get context relevant to their local recycling infrastructure.

---

Tech Stack

Backend (Python 3.10+)

• FastAPI: Async REST API framework
• MongoDB: User profiles, scans, transactions
• Faiss: Vector database for RAG (global + personal knowledge)
• CLIP (ViT-B/32): Zero-shot image classification (local inference)
• Whisper (Small): Multilingual speech-to-text (local inference)
• Groq (Llama 3.3 70B): LLM reasoning (FREE API)
• OSRM: Open-source routing for recycler navigation

Frontend (React 18)

• Vite: Fast build tool
• React Router: Client-side routing
• Zustand: Lightweight state management
• Leaflet: Interactive maps
• Axios: HTTP client

---

🚀 Quick Start

Prerequisites

• Python 3.10+ (for backend)
• Node.js 18+ (for frontend)
• MongoDB 5.0+ (local or Atlas)
• Git

1. Clone Repository

git clone https://github.com/Samrudhp/sathack.git
cd sathack

2. Backend Setup

cd backend

# Create virtual environment
python3.10 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Create .env file
cat > .env << EOF
MONGODB_URI=mongodb://localhost:27017/
GROQ_API_KEY=your_groq_api_key_here

EOF

# Run server
uvicorn app.main:app --reload --port 8000

Backend will be available at: http://localhost:8000
API Docs: http://localhost:8000/docs

3. Frontend Setup

cd frontend-v2

# Install dependencies
npm install

# Run dev server
npm run dev

Frontend will be available at: http://localhost:5174

4. Seed Database (Optional)

cd backend
python scripts/seed_recyclers.py  # Adds sample recyclers in Patiala
python scripts/seed_rag_docs.py   # Adds waste management guidelines

---

📸 Features Showcase

1. Smart Image Scanning

• CLIP-powered zero-shot classification (no training data needed!)
• Detects material type (PET, aluminum, e-waste, etc.)
• Estimates weight using object detection
• Assesses cleanliness score (affects token rewards)
• Identifies hazards (sharp objects, chemicals, biohazards)

2. Voice Assistant

• Whisper transcription in 12 Indian languages
• Natural language queries: "Where do I recycle batteries in Patiala?"
• RAG-powered responses using city-specific guidelines
• Bhashini translation for multilingual responses

3. Smart Recycler Marketplace

• Geospatial ranking (distance, material acceptance, pricing)
• Real-time availability and capacity tracking
• Route optimization using OSRM (save time & fuel)
• Verified recyclers with ratings and reviews

4. Token Economy

• Earn tokens based on material value × weight × cleanliness
• Redeem for: pickups, premium analytics, partner discounts
• Leaderboards and challenges for engagement
• Referral system (both parties get bonus tokens)

5. Personalized RAG

• Dual-context retrieval: Global rules + your past behavior
• "You recycled PET well last time - try combining with HDPE pickup!"
• Learns your preferred recyclers, optimal days, material patterns

6. Environmental Impact

• Real-time tracking: CO₂ saved, water conserved, landfill avoided
• Verified calculations (not random numbers!)
• ESG-ready reports for businesses

---

🔐 Security & Privacy

For Users:

✅ No facial recognition - Only waste images stored
✅ Location privacy - Coordinates never shared with recyclers until pickup confirmed
✅ Encrypted storage - MongoDB encrypted at rest
✅ GDPR-compliant - Right to deletion, data export
✅ Anonymous analytics - Personal data never sold

For Recyclers:

✅ Verification required - Business license + GST checks
✅ Escrow payments - Tokens held until pickup confirmed
✅ Fraud detection - AI flags suspicious patterns (fake scans, rating manipulation)
✅ Dispute resolution - Built-in mediation system

Security Measures Implemented:

• Rate limiting (10 scans/min, 30 voice queries/min)
• Input validation (file type checks, size limits)
• SQL injection prevention (parameterized queries)
• CORS configuration (whitelist trusted origins)
• API key rotation (monthly automated)
• Audit logs (all transactions tracked)

---

📊 API Documentation

Core Endpoints

POST /api/scan

Scan waste image and get disposal advice

curl -X POST http://localhost:8000/api/scan \
  -F "user_id=USER_ID" \
  -F "image=@bottle.jpg" \
  -F "latitude=30.34" \
  -F "longitude=76.38" \
  -F "language=en"

Response:

{
  "material": "PET",
  "confidence": 0.87,
  "weight_estimate_kg": 0.03,
  "cleanliness_score": 70,
  "hazard_class": null,
  "estimated_credits": 8,
  "disposal_instruction": "♻️ Clean, crush, recycle at nearest PET center...",
  "recycler_ranking": [
    {
      "name": "Green Recyclers",
      "distance_km": 2.3,
      "phone": "+91-1234567890"
    }
  ],
  "environmental_impact": {
    "co2_saved_kg": 0.075,
    "water_saved_liters": 1.5
  }
}

POST /api/voice_input

Voice query with transcription + reasoning

curl -X POST http://localhost:8000/api/voice_input \
  -F "user_id=USER_ID" \
  -F "audio=@query.webm" \
  -F "language=hi"

GET /api/marketplace/recyclers

Get nearby verified recyclers

curl "http://localhost:8000/api/marketplace/recyclers?lat=30.34&lon=76.38&material=PET"

Full API docs: http://localhost:8000/docs (Swagger UI)

---

🎮 Token Economics

How Tokens are Calculated

estimated_credits = weight_kg × material_rate × (cleanliness_score / 100)

# Example: 30g PET bottle, 70% clean
= 0.03 × 400 × 0.7
= 8.4 tokens (rounded to 8)

Material Rates (credits per kg)

Material	Rate	Real-World Value
PET	400 credits/kg	₹12/kg
Aluminum	600 credits/kg	₹18/kg
E-Waste	700 credits/kg	₹20/kg
Paper	200 credits/kg	₹6/kg
Plastic (generic)	250 credits/kg	₹7.50/kg
Glass	150 credits/kg	₹4.50/kg

Redemption Options

• 100 tokens = Free doorstep pickup
• 500 tokens = ₹50 voucher (Swiggy, Zomato, Amazon)
• 1,000 tokens = Premium analytics (waste trends, comparisons)
• 2,000 tokens = Tree planted in your name (verified via Grow-Trees.com)

---

💰 Business Model & Monetization

Revenue Streams:

1. Recycler Commissions (15% on transactions)

   • User books pickup → Recycler pays 15% platform fee

2. Premium Subscriptions (₹99/month)

   • Priority pickups, advanced analytics, zero ads

3. B2G (Business-to-Government) Licensing

   • Sell white-label solution to municipalities (₹5-10L per city)

4. Carbon Credit Aggregation

   • Sell verified credits to corporates (₹1000-5000/ton CO₂)

5. Data Licensing (anonymized)

   • Insights to waste management companies (₹50K-2L/month)

---

🌍 Impact & Sustainability

Current Impact (MVP Stage):

• 500+ scans processed
• 50kg waste diverted from landfills
• 125kg CO₂ emissions avoided
• 2,500L water conserved

Projected Impact @ 100K Users:

• 5M kg/year waste diverted
• 12,500 tons CO₂/year avoided (= planting 570,000 trees!)
• 250M liters/year water saved

UN SDG Alignment:

• SDG 11: Sustainable Cities and Communities
• SDG 12: Responsible Consumption and Production
• SDG 13: Climate Action
• SDG 17: Partnerships for Goals

---

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

How to Contribute:

1. Fork the repo
2. Create feature branch: git checkout -b feature/amazing-feature
3. Commit changes: git commit -m 'Add amazing feature'
4. Push to branch: git push origin feature/amazing-feature
5. Open Pull Request

Areas We Need Help:

• 🐛 Bug reports and fixes
• 🌐 Translations (add more languages)
• 📱 Mobile app development
• 🎨 UI/UX improvements
• 📚 Documentation
• 🧪 Test coverage

---

📞 Contact & Support

• Email: samrudhprakash3084@gmail.com
• GitHub Issues: Report bugs here

---

🙏 Acknowledgments

• Groq - Free LLM API (Llama 3.3 70B)
• OpenAI - CLIP and Whisper models
• OpenStreetMap - Geospatial data and routing
• MongoDB - Database platform
• Qdrant - Vector database for RAG

---

📊 Project Stats

---

⭐ Star this repo if you found it helpful!

Made with 💚 for a sustainable future

Report Bug · Request Feature · Contribute