Wine Semantic Search MCP Server

A Model Context Protocol (MCP) server that enables semantic search capabilities over a PostgreSQL database containing wine listings and descriptions. The system uses vector embeddings to perform intelligent searches that understand the meaning and context of wine-related queries, returning relevant wine recommendations based on semantic similarity rather than just keyword matching.

Features

• Semantic Search: Uses OpenAI's text-embedding-3-small model for intelligent wine recommendations
• Vector Database: PostgreSQL with pgvector extension for efficient similarity search
• MCP Protocol: Full compliance with Model Context Protocol for AI assistant integration
• Configurable: Flexible search parameters and result limits
• Production Ready: Comprehensive error handling, logging, and connection management
• Type Safe: Full type annotations and validation using Pydantic

Requirements

• Python: 3.8 or higher
• PostgreSQL: 12+ with pgvector extension installed
• OpenAI API Key: For generating embeddings

Quick Start

1. Install Dependencies

# Clone the repository
git clone <repository-url>
cd wine-semantic-search

# Install the package
pip install -e .

# For development
pip install -e ".[dev]"

2. Database Setup

Install PostgreSQL and pgvector

Ubuntu/Debian:

sudo apt-get install postgresql postgresql-contrib
sudo apt-get install postgresql-14-pgvector  # Adjust version as needed

macOS (using Homebrew):

brew install postgresql
brew install pgvector

Docker (Alternative):

docker run --name wine-postgres -e POSTGRES_PASSWORD=password -p 5432:5432 -d pgvector/pgvector:pg16

Create Database and Schema

# Connect to PostgreSQL
psql -U postgres -h localhost

# Create database
CREATE DATABASE wine_database;

# Connect to the new database
\c wine_database

# Run the schema setup
\i database_schema.sql

Or use the provided schema file:

psql -U postgres -h localhost -d wine_database -f database_schema.sql

3. Configuration

Copy the example configuration file and customize it:

cp .env.example .env

Edit .env with your settings:

# Required settings
DATABASE_URL=postgresql://username:password@localhost:5432/wine_database
OPENAI_API_KEY=your_openai_api_key_here

# Optional settings
MAX_RESULTS=50
SIMILARITY_THRESHOLD=0.7

4. Load Wine Data (Optional)

If you have wine data to import, you can use the following format:

INSERT INTO wine (country, description, designation, points, price, province, region_1, region_2, variety, winery)
VALUES 
('France', 'A rich, full-bodied Bordeaux with notes of blackcurrant and oak', 'Château Example 2020', 92, 45.99, 'Bordeaux', 'Left Bank', 'Pauillac', 'Cabernet Sauvignon', 'Château Example'),
-- Add more wine records...

Note: Embeddings will be generated automatically when you perform searches.

5. Run the Server

# Start the MCP server
wine-search-server

# Or run directly with Python
python -m wine_semantic_search.main

Usage

As an MCP Server

The server exposes a single tool called search_wines that can be used by any MCP-compatible client:

Tool Schema:

{
  "name": "search_wines",
  "description": "Search for wines using natural language descriptions",
  "inputSchema": {
    "type": "object",
    "properties": {
      "query": {
        "type": "string",
        "description": "Natural language description of desired wine characteristics"
      },
      "limit": {
        "type": "integer",
        "description": "Maximum number of results to return (default: 10, max: 50)",
        "minimum": 1,
        "maximum": 50,
        "default": 10
      }
    },
    "required": ["query"]
  }
}

Example Queries:

• "I want a full-bodied red wine from France under $30"
• "Light, crisp white wine perfect for seafood"
• "Bold Cabernet Sauvignon with high ratings"
• "Sweet dessert wine from Germany"

Testing the Installation

You can test your setup using the built-in health check:

# Test database connection and configuration
python -c "
from wine_semantic_search.main import test_setup
import asyncio
asyncio.run(test_setup())
"

Database Schema

The system uses two main tables:

wine table

Stores wine information including:

• Basic details (name, country, region, variety, winery)
• Ratings and pricing (points, price)
• Descriptive text (description)

wine_embeddings table

Stores vector embeddings for semantic search:

• 512-dimensional vectors generated by OpenAI
• Linked to wine records via foreign key
• Indexed for efficient similarity search

See database_schema.sql for the complete schema definition.

Configuration Options

Environment Variable	Required	Default	Description
DATABASE_URL	Yes	-	PostgreSQL connection string
OPENAI_API_KEY	Yes	-	OpenAI API key for embeddings
MAX_RESULTS	No	50	Maximum search results (1-50)
SIMILARITY_THRESHOLD	No	0.7	Minimum similarity score (0.0-1.0)

Development

Running Tests

# Run all tests
pytest

# Run with coverage
pytest --cov=wine_semantic_search

# Run specific test categories
pytest tests/test_database.py  # Database tests
pytest tests/test_embedding.py  # Embedding tests
pytest tests/test_integration.py  # Integration tests

Code Quality

# Format code
black wine_semantic_search/
isort wine_semantic_search/

# Type checking
mypy wine_semantic_search/

# Linting
flake8 wine_semantic_search/

Property-Based Testing

The project uses Hypothesis for property-based testing to ensure correctness across a wide range of inputs:

# Run property-based tests with verbose output
pytest tests/ -v --hypothesis-show-statistics

Troubleshooting

Common Issues

1. pgvector extension not found

ERROR: extension "vector" is not available

Solution: Install pgvector extension for your PostgreSQL version.

2. Database connection failed

ERROR: could not connect to server

Solution: Check your DATABASE_URL and ensure PostgreSQL is running.

3. OpenAI API errors

ERROR: Invalid API key

Solution: Verify your OPENAI_API_KEY is correct and has sufficient credits.

4. No search results

INFO: No wines found matching query

Solution: Check that your database contains wine records and try adjusting the SIMILARITY_THRESHOLD.

Logging

The server provides detailed logging for troubleshooting:

# Enable debug logging
export LOG_LEVEL=DEBUG
wine-search-server

Performance Tuning

For large datasets, consider:

1. Database Indexing: The schema includes optimized indexes for common queries
2. Connection Pooling: Adjust DB_POOL_MAX_SIZE for high-concurrency scenarios
3. Similarity Threshold: Higher thresholds reduce computation but may miss relevant results
4. Result Limits: Lower limits improve response times

Architecture

The system follows a layered architecture:

• MCP Server Layer: Handles protocol communication and tool registration
• Service Layer: Business logic for embeddings and database operations
• Data Layer: PostgreSQL with pgvector for efficient vector operations

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes with tests
4. Run the test suite
5. Submit a pull request

License

[Add your license information here]

Support

For issues and questions:

1. Check the troubleshooting section above
2. Review the test files for usage examples
3. Open an issue on the repository