Stopify

AI-Powered Impulse Purchase Prevention System

Built at Hack@Brown 2026

2nd Place - Marshal Wace Intelligent Systems Prize

Overview

Stopify is an intelligent browser extension and backend system that helps users make better purchasing decisions by detecting and preventing impulse purchases. The system combines real-time biometric analysis with AI-powered contextual reasoning to identify when users are making impulsive decisions and provides timely interventions.

Key Features

• Dual-Brain Analysis System

  • Fast Brain: Real-time Bayesian inference using biometric data (heart rate, facial expressions)
  • Slow Brain: RAG-based reasoning with Google Vertex AI for contextual analysis

• Biometric Monitoring (via Presage SDK)

  • Heart rate variability detection
  • Facial expression analysis
  • Stress level indicators

• Smart Intervention

  • Real-time purchase decision analysis
  • Personalized warnings based on user goals and budget
  • Self-refining memory system that learns from user behavior

• Wide E-Commerce Support

  • Amazon, eBay, Walmart, Target, Best Buy
  • Etsy, AliExpress, Temu, Wayfair
  • Nike, Adidas, Shein, IKEA, and more
  • Gambling site detection and prevention

Architecture

┌─────────────────┐
│ Browser         │
│ Extension       │
│ (Stopify)       │
└────────┬────────┘
         │
         │ Biometric Data + Product Info
         ▼
┌─────────────────────────────────────────────────┐
│ FastAPI Backend                                 │
│                                                 │
│  ┌──────────────────────────────────────────┐   │
│  │ Fast Brain (Bayesian Inference)          │   │
│  │  └─> Real-time impulse score             │   │
│  └──────────────────────────────────────────┘   │
│                                                 │
│  ┌──────────────────────────────────────────┐   │
│  │ Slow Brain (RAG + Vertex AI)             │   │
│  │  ├─> RAG Retrieval (ChromaDB)            │   │
│  │  ├─> Context Analysis                    │   │
│  │  └─> Gemini AI Reasoning                 │   │
│  └──────────────────────────────────────────┘   │
│                                                 │
│  ┌──────────────────────────────────────────┐   │
│  │ Memory System (Self-Refining)            │   │
│  │  ├─> Goals.md                            │   │
│  │  ├─> Budget.md                           │   │
│  │  ├─> Behavior.md                         │   │
│  │  └─> State.md                            │   │
│  └──────────────────────────────────────────┘   │
└────────┬────────────────────────────────────────┘
         │
         │ {impulse_score, reasoning, intervention}
         ▼
┌─────────────────┐
│ User Interface  │
│ (Popup/Overlay) │
└─────────────────┘

Getting Started

Prerequisites

• Python 3.10+
• Google Cloud account with Vertex AI API access
• Chrome/Chromium browser
• (Optional) Presage SmartSpectra SDK for real biometric data

Backend Setup

1. Clone the repository

   git clone https://github.com/calebellenberg/hackAtBrown2026.git
   cd hackAtBrown2026/hackAtBrown2026/backend

2. Create virtual environment

   python -m venv venv
   # Windows
   venv\Scripts\activate
   # macOS/Linux
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Configure environment variables

   Create a .env file in the backend directory:

   VERTEX_SERVICE_ACCOUNT_PATH=path/to/your/key.json
   GOOGLE_CLOUD_PROJECT=your-project-id

5. Set up Google Cloud credentials

   • Create a service account in Google Cloud Console
   • Enable Vertex AI API
   • Download the JSON key file
   • Place it in the backend directory and update .env

6. Run the backend server

   uvicorn app:app --host 0.0.0.0 --port 8000 --reload

   The API will be available at http://localhost:8000

Extension Setup

1. Navigate to extension directory

   cd ../extension

2. Configure the backend URL

   Update the API endpoint in background.js if needed (default: http://localhost:8000)

3. Load extension in Chrome

   • Open Chrome and go to chrome://extensions/
   • Enable "Developer mode" (top right)
   • Click "Load unpacked"
   • Select the extension directory

4. Grant permissions

   • Allow camera access when prompted (for biometric monitoring)
   • The extension will request access to e-commerce websites

Presage Vitals Setup (Optional)

For real biometric monitoring using the Presage SmartSpectra SDK:

Windows/Linux

cd backend/persage
./build.sh
python broker.py

macOS

cd backend/persage
python mac_broker.py

The vitals broker will run on http://localhost:8766

See persage/SETUP.md for detailed setup instructions.

Usage

1. Set Your Goals and Budget

   • Click the extension icon
   • Configure your financial goals and spending limits
   • The system will learn from your preferences

2. Browse Shopping Sites

   • Navigate to any supported e-commerce website
   • The extension monitors your browsing behavior
   • Biometric data is collected in real-time

3. Receive Intelligent Warnings

   • When an impulse purchase is detected, you'll see a warning
   • The system considers:
     • Your current emotional state (via biometrics)
     • Your stated goals and budget
     • Past purchasing patterns
     • Product price and context

4. Make Informed Decisions

   • Review the AI's reasoning
   • Proceed with purchase or reconsider
   • The system learns from your decisions

Technology Stack

Backend

• FastAPI - High-performance API framework
• Google Vertex AI - Gemini AI for contextual reasoning
• ChromaDB - Vector database for RAG retrieval
• NumPy/SciPy - Bayesian inference calculations
• OpenCV - Camera/biometric processing

Frontend

• Chrome Extension API - Browser integration
• WebSockets - Real-time data streaming
• Vanilla JavaScript - Lightweight and fast

Biometric Processing

• Presage SmartSpectra SDK - Heart rate and vitals detection
• C++ with OpenCV - High-performance video processing

Project Structure

hackAtBrown2026/
├── backend/
│   ├── app.py                    # Main FastAPI application
│   ├── memory.py                 # Memory engine with RAG
│   ├── inference_engine.py       # Bayesian inference (Fast Brain)
│   ├── analyze_with_gemini.py    # Vertex AI integration
│   ├── requirements.txt          # Python dependencies
│   ├── memory_store/             # Self-refining markdown memory
│   │   ├── Goals.md
│   │   ├── Budget.md
│   │   ├── Behavior.md
│   │   └── State.md
│   ├── persage/                  # Biometric monitoring system
│   │   ├── broker.py             # Vitals data broker
│   │   ├── headless_vitals.cpp   # C++ vitals processor
│   │   ├── mac_broker.py         # macOS compatibility
│   │   └── build/                # Compiled binaries
│   └── tests/                    # Test suite
│
└── extension/
    ├── manifest.json             # Extension configuration
    ├── background.js             # Service worker
    ├── content.js                # Page interaction
    ├── popup.html/js             # User interface
    ├── tracker.js                # Behavior tracking
    ├── camera.html/js            # Biometric data capture
    ├── icon.png                  # Extension icon
    └── gambling_sites.json       # Gambling site blocklist

API Endpoints

Main Analysis Endpoint

POST /analyze

Analyzes a purchase decision combining biometric data and contextual reasoning.

{
  "p_impulse_fast": 0.75,
  "product": "Wireless Headphones",
  "cost": 199.99,
  "website": "amazon.com",
  "user_id": "user123"
}

Response:

{
  "impulse_score": 0.82,
  "confidence": 0.91,
  "reasoning": "High impulse likelihood detected...",
  "intervention_action": "warn",
  "memory_update": true
}

Other Endpoints

• POST /sync-memory - Force re-indexing of all memory files
• POST /update-preferences - Update user goals and budget preferences
• POST /reset-memory - Reset memory to defaults
• GET /health - Health check endpoint
• GET /vitals - Get current biometric data
• POST /pipeline-analyze - Full pipeline analysis with vitals
• POST /gemini-analyze - Direct Gemini AI analysis

See API_DOCUMENTATION.md for complete API reference.

Testing

cd backend

# Run all tests
pytest

# Test API endpoints
python tests/test_api.py

# Test memory system
python tests/test_memory.py

# Test Vertex AI integration
python tests/test_vertex_ai.py

# Run with extension data
python tests/test_with_extension_data.py

Contributing

This project was built for Hack@Brown 2026. Contributions, issues, and feature requests are welcome!

Team

Built at Hack@Brown 2026

Winner - 2nd Place in Marshal Wace Intelligent Systems Prize

Acknowledgments

• Marshal Wace - For recognizing innovative approaches to intelligent systems
• Presage Technologies - SmartSpectra SDK for biometric monitoring
• Google Cloud - Vertex AI and Gemini API
• Hack@Brown - For the amazing hackathon experience

---

Made at Hack@Brown 2026