🛒 ShopSmart AI Agent

AI-powered shopping assistant that helps you find the best deals, compare prices, and track product prices across the web.

Features

• Multi-user accounts — Register/login with JWT auth; every user's tracked products, chats and notifications are private to them
• AI Chat Interface — Chat naturally to search products and compare prices
• Web Search Tool — Search across the internet for products and deals
• Price Scraper — Extract prices from product pages automatically (site-specific selectors + sane price-range validation)
• Price Tracking — Track products and monitor price changes over time
• Auto price monitor — Background job re-scrapes tracked products on a schedule and records history
• Price Analysis — AI-powered trend analysis and buying recommendations
• Real notifications — In-app alerts plus optional Telegram / email push when a price drops or hits your target

Tech Stack

Quick Start

1. Clone & Install

git clone https://github.com/YOUR_USERNAME/ShopSmart.git
cd ShopSmart
pip install -r requirements.txt

2. Set API Key

Get a free Gemini API key from Google AI Studio.

# Windows PowerShell
$env:GEMINI_API_KEY='your_api_key_here'

# Windows CMD
set GEMINI_API_KEY=your_api_key_here

# Linux/Mac
export GEMINI_API_KEY='your_api_key_here'

3. Run

python app.py

Open http://127.0.0.1:5000 in your browser, then register an account to get started.

app.py is the development server (Werkzeug). For production, see Deployment.

Configuration

Copy .env.example to .env and fill in what you need. Only GEMINI_API_KEY is required.

Deployment

app.py is the Flask dev server — don't use it in production. Use a real WSGI server instead.

Single process (waitress, cross-platform incl. Windows)

python serve.py
# or:  waitress-serve --host=0.0.0.0 --port=5000 wsgi:app

The background price monitor starts in-process — fine for a single process.

Multiple workers (e.g. gunicorn on Linux)

Running N workers would spawn N monitors and scrape every product N times. To avoid
that, disable the in-process monitor and run one dedicated monitor process:

# .env: PRICE_MONITOR_IN_PROCESS=false
gunicorn -w 4 -b 0.0.0.0:5000 wsgi:app   # web workers
python monitor.py                        # single monitor process

Always set JWT_SECRET to a fixed value in production so tokens survive restarts.

Testing

pytest

Tests run against an isolated temporary SQLite database — they never touch shopsmart.db.

Architecture

ShopSmart/
├── app.py                  # Flask server & API endpoints
├── config.py               # Configuration
├── requirements.txt        # Python dependencies
├── agent/
│   ├── core.py             # AI Agent with ReAct loop
│   ├── prompts.py          # System prompts
│   └── tools/
│       ├── web_search.py   # DuckDuckGo product search
│       ├── price_scraper.py# Price extraction from URLs
│       ├── price_tracker.py# Database price tracking
│       ├── price_analyzer.py# Price trend analysis
│       └── notifier.py     # Notification system
├── database/
│   ├── db.py               # SQLite connection & schema
│   └── models.py           # Data models (CRUD)
└── frontend/
    ├── index.html           # Single-page application
    ├── css/                 # Dark theme UI styles
    └── js/                  # Chat, dashboard, notifications

Agent Tools

License

MIT License