rnit-vanna 1.0.0

Enhanced wrapper for Vanna SQL generation library with custom utilities

RNIT Vanna - Enhanced SQL Generation

RNIT Vanna is an enhanced wrapper around the official Vanna library that provides additional utilities while always using the latest Vanna version.

Key Features

✨ Always Up-to-Date: Automatically uses the latest official Vanna release 🔧 Enhanced Utilities: Additional tools for database inspection and training generation 🚀 Quick Start: Pre-configured classes with sensible defaults 📦 100% Compatible: Works exactly like Vanna, but with extras 🎯 Smart Defaults: Optimized settings for SQL generation

Installation

pip install rnit-vanna

This automatically installs the latest Vanna with all required dependencies.

Quick Start

Option 1: Drop-in Replacement (Works exactly like Vanna)

from rnit_vanna import OpenAI_Chat, ChromaDB_VectorStore

class MyVanna(ChromaDB_VectorStore, OpenAI_Chat):
    def __init__(self, config=None):
        ChromaDB_VectorStore.__init__(self, config=config)
        OpenAI_Chat.__init__(self, config=config)

vn = MyVanna(config={'api_key': 'your-openai-key'})
vn.connect_to_sqlite('my_database.sqlite')
vn.train(ddl="CREATE TABLE users (id INT, name TEXT)")
sql = vn.generate_sql("Show all users")

Option 2: Use RNIT Enhanced Features

from rnit_vanna import RNITVanna, VannaQuickStart

# Quick setup with auto-training
vn = VannaQuickStart.for_sqlite('my_database.sqlite', auto_train=True)

# Or use RNITVanna with auto API key detection
vn = RNITVanna()  # Reads OPENAI_API_KEY from environment
vn.connect_to_sqlite('my_database.sqlite')

# Batch training
vn.batch_train({
    'ddl': ["CREATE TABLE users ...", "CREATE TABLE orders ..."],
    'queries': [
        {"question": "Show all users", "sql": "SELECT * FROM users"},
        {"question": "Count orders", "sql": "SELECT COUNT(*) FROM orders"}
    ]
})

Option 3: Database Inspection and Auto-Training

from rnit_vanna import DatabaseInspector, TrainingGenerator, RNITVanna

# Inspect your database
inspector = DatabaseInspector()
db_info = inspector.inspect_sqlite('my_database.sqlite')
print(inspector.generate_summary(db_info))

# Generate training queries automatically
generator = TrainingGenerator()
for table_name, info in db_info['tables'].items():
    queries = generator.generate_basic_queries(table_name, info['columns'])
    # Now you have training queries for each table!

# Train Vanna with generated queries
vn = RNITVanna()
vn.connect_to_sqlite('my_database.sqlite')
for query in queries:
    vn.train(question=query['question'], sql=query['sql'])

Enhanced Features

🎯 RNITVanna Class

• Auto-detects OpenAI API key from environment
• Optimized defaults (gpt-4o-mini, low temperature for consistent SQL)
• Training history tracking
• Batch training support
• Test accuracy methods

🚀 VannaQuickStart

• for_sqlite() - Auto-setup and train from SQLite
• for_postgres() - Quick PostgreSQL setup
• for_mysql() - Quick MySQL setup
• create_sample_project() - Generate complete project structure

🔍 DatabaseInspector

• Inspect database structure
• Extract relationships and foreign keys
• Generate human-readable summaries
• Count rows and analyze tables

🎓 TrainingGenerator

• Auto-generate training queries from schema
• Create JOIN queries from relationships
• Generate aggregation queries for numeric columns
• Save training sets to JSON

⚡ QueryOptimizer

• Validate SQL against database
• Get execution plans
• Suggest indexes for performance
• Format SQL for readability

Complete Example

from rnit_vanna import VannaQuickStart

# Create a complete sample project
VannaQuickStart.create_sample_project('my_sql_assistant', db_type='sqlite')

# This creates:
# my_sql_assistant/
#   ├── .env.example       # Environment variables template
#   ├── main.py           # Starter code
#   ├── requirements.txt  # Dependencies
#   └── README.md        # Documentation

Why Use RNIT Vanna?

Feature	Official Vanna	RNIT Vanna
Latest Vanna version	✅	✅ Automatic
Basic SQL generation	✅	✅
Auto API key detection	❌	✅
Database inspection	❌	✅
Training generation	❌	✅
Batch training	❌	✅
Quick start utilities	❌	✅
Project scaffolding	❌	✅

Environment Variables

# .env file
OPENAI_API_KEY=your_openai_api_key_here
RNIT_VANNA_VERBOSE=true  # Optional: Enable verbose logging

Advanced Usage

Custom Configuration

vn = RNITVanna(config={
    'model': 'gpt-4',        # Use GPT-4 instead of default
    'temperature': 0.5,      # Higher creativity
    'max_tokens': 1000,      # Longer responses
})

Training from Files

# Train from JSON file
vn.train_from_queries_file('training_queries.json')

# Train from folder of DDL files
vn.train_from_ddl_folder('schema/')

Testing Accuracy

test_queries = [
    {"question": "Show all users", "expected_sql": "SELECT * FROM users"},
    {"question": "Count orders", "expected_sql": "SELECT COUNT(*) FROM orders"}
]

results = vn.test_accuracy(test_queries)
print(f"Accuracy: {results['accuracy']:.2%}")

Compatibility

• Python 3.7+
• Works with all databases supported by Vanna (SQLite, PostgreSQL, MySQL, Snowflake, etc.)
• Compatible with all Vanna versions >= 0.5.0

License

MIT License - see LICENSE file for details.

Contributing

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

Support

For issues, questions, or suggestions, please open an issue on GitHub.

Acknowledgments

This project wraps the excellent Vanna.AI library. All core SQL generation capabilities come from Vanna.