AI Shopping Concierge for High-End Purchases

Mina researches premium products across retailers, analyzes reviews, and delivers confident recommendations with full reasoning transparency—so you can make high-end purchase decisions with clarity.

• 🎯 Intelligent Category Selection: Specialized support for laptops, furniture, and appliances
• 🌐 Multi-Retailer Research: Browser Use integration for live scraping across retailers
• 🤖 AI-Powered Analysis: Claude Sonnet 4 analyzes reviews, specs, and value propositions
• 🔒 Safe Code Execution: Daytona sandbox for secure data processing
• 📊 Confidence Scoring: Galileo observability with multi-factor confidence metrics
• 🚀 REST API & WebSocket: FastAPI server for frontend integration with real-time updates
• ⚡ Async Operations: Efficient parallel scraping and processing
• 💡 Clear Reasoning: Every recommendation comes with detailed pros, cons, and reasoning
• 💰 Budget-Aware: Focuses on purchases $500+ with budget filtering
• 🛡️ Graceful Fallbacks: Works even without API keys for testing and development

1. Category Selection: Choose your product category (laptop, furniture, or appliance)
2. Requirements Gathering: Specify your budget and priorities
3. Multi-Retailer Scraping: Browser Use cloud browser scrapes multiple retailers in parallel
4. Sandbox Processing: Daytona safely executes data analysis code in isolated environment
5. AI Analysis: Claude Sonnet 4 analyzes specifications and reviews against requirements
6. Confidence Scoring: Galileo-logged multi-factor confidence calculation
7. Transparent Recommendations: Receive ranked recommendations with full reasoning and metrics

• Claude (Anthropic): Claude Sonnet 4 for AI-powered product analysis
• Browser Use: Cloud browser automation for anti-bot protected scraping
• Daytona: Secure sandbox environment for code execution and data processing
• Galileo: Observability and confidence score tracing
• FastAPI: Modern web framework for REST API and WebSocket support
• Python 3.8+: Core implementation with async/await support
• Pandas & NumPy: Data processing and analysis

• Python 3.8 or higher
• pip (Python package manager)

1. Clone the repository:

git clone https://github.com/wildhash/Mina.git
cd Mina

2. Install dependencies:

pip install -r requirements.txt

3. Configure environment variables:

cp .env.example .env
# Edit .env and add your API keys

4. Set up your API keys (all optional for testing):
   • Anthropic API key from console.anthropic.com - For Claude AI
   • Browser Use API key from cloud.browser-use.com - For web scraping ($10 free credits)
   • Daytona API key from daytona.io - For sandbox execution ($200 free credits)
   • Galileo API key from galileo.ai - For observability
   • Add all keys to your .env file

Note: Mina works without API keys using mock data and fallback mechanisms for testing/development.

Run the FastAPI server to enable frontend integration:

python api_server.py

The server starts on http://localhost:8000 with the following endpoints:

• Health Check: GET /api/health - Check server status and integrations
• Product Search: POST /api/search - Search for products based on requirements
• WebSocket: ws://localhost:8000/ws - Real-time search updates

Example API Request:

curl -X POST http://localhost:8000/api/search \
  -H "Content-Type: application/json" \
  -d '{
    "category": "laptop",
    "budget_max": 3000,
    "priorities": ["Performance", "Battery Life"],
    "specific_needs": "For software development"
  }'

Interactive API Documentation:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

For detailed API documentation and configuration, see START_API.md.

Run the integration demo to see all features:

python demo_integrations.py

This demonstrates:

• Browser Use integration (with fallback to mock data)
• Daytona sandbox execution (with fallback to direct processing)
• Galileo observability (with @log decorators)
• Claude AI analysis (with fallback to rule-based)
• Async workflow for efficient parallel operations

Run Mina interactively from the command line:

python mina_cli.py

The agent will guide you through:

1. Selecting a product category
2. Specifying your budget and priorities
3. Reviewing research results
4. Understanding recommendations with confidence scores

Use Mina in your Python code:

import asyncio
from mina_agent import MinaAgent

async def find_laptop():
    agent = MinaAgent()
    
    # Async workflow (recommended)
    recommendations = await agent.research_product_async(
        category="laptop",
        budget=3000,
        priorities=["Performance", "Battery Life"]
    )
    
    # Display results
    agent.present_recommendations(recommendations)

asyncio.run(find_laptop())

Welcome to Mina - Your AI Shopping Concierge
============================================================

I help you make confident decisions on high-end purchases ($500+)
by researching products, analyzing reviews, and providing
transparent recommendations with confidence scores.

What type of product are you looking to purchase?
1. Laptop
2. Furniture
3. Appliance

Enter your choice (1-3): 1

============================================================
Great! Let's find the perfect laptop for you.
============================================================

What's your maximum budget? ($): 3000

What are your top priorities? (You can select multiple)
1. Performance
2. Battery Life
3. Display Quality
4. Portability
5. Build Quality

Enter priority numbers (comma-separated, e.g., 1,3,4): 1,2,5

...

• Performance comparison
• Battery life analysis
• Display quality assessment
• Build quality evaluation
• Price-to-performance ratios

• Durability and warranty analysis
• Comfort ratings
• Material quality assessment
• Design and aesthetics
• Space efficiency

• Energy efficiency ratings
• Capacity and features
• Reliability metrics
• Smart home integration
• Long-term cost analysis

Mina uses a multi-factor approach with Galileo observability to calculate confidence scores:

• Customer Ratings (25%): Verified buyer ratings and review sentiment
• Requirements Fit (40%): How well the product matches your stated priorities
• Review Confidence (20%): Volume and consistency of reviews
• Data Completeness (15%): Availability of comprehensive specifications

Scores range from 0-100%, with higher scores indicating greater confidence in the recommendation.

All scoring steps are logged via Galileo's @log decorators for full transparency and traceability.

Mina/
├── mina_agent.py           # Core agent with all integrations
├── mina_cli.py             # Command-line interface
├── api_server.py           # FastAPI server for frontend integration
├── demo_integrations.py    # Full integration demo
├── requirements.txt        # Python dependencies
├── .env.example            # Environment variable template
├── test_mina.py            # Core unit tests
├── test_api.py             # API endpoint tests
├── test_integrations.py    # Integration tests
├── examples/               # Usage examples
├── START_API.md            # API server documentation
├── .gitignore             # Git ignore rules
└── README.md              # This file

# Run core unit tests
python test_mina.py

# Run API tests
python test_api.py

# Run integration tests
python test_integrations.py

# Or use pytest for all
pip install pytest pytest-cov
pytest

# Run with coverage
pytest --cov=. tests/

Each integration can be tested independently:

# Test Browser Use
agent = MinaAgent()
assert agent.browser is not None or agent.browser is None  # Works with or without API key

# Test Daytona
result = agent.analyze_in_sandbox([{"test": "data"}])
assert "products" in result

# Test Galileo decorators
@log(span_type="workflow")
def my_function():
    pass  # Automatically traced when Galileo configured

• Browser Use integration for real-time retailer scraping
• Daytona sandbox for safe code execution
• Galileo observability and confidence tracing
• Claude Sonnet 4 AI analysis
• Async workflow support
• FastAPI REST API and WebSocket support for frontend integration
• Live scraping from major retailers (Best Buy, Amazon, B&H Photo)
• Enhanced parsing for product specifications
• Expanded product categories (TVs, cameras, watches)
• Price tracking and deal alerts
• Comparison view for side-by-side analysis
• Export recommendations to PDF
• Integration with price history APIs
• Multi-language support
• Mobile app interface

Contributions are welcome! Please feel free to submit a Pull Request.

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

For issues, questions, or suggestions:

• Open an issue on GitHub

• Built with Anthropic Claude
• Browser automation powered by Browser Use
• Confidence methodology inspired by Galileo.ai
• Developed with Daytona.io

Made with ❤️ for confident high-end shopping decisions