AI-Powered Text-to-SQL Generator

This project is an AI-powered Text-to-SQL generator that converts natural language queries into SQL queries using GPT-4 and FAISS vector databases. By leveraging Retrieval-Augmented Generation (RAG), the tool intelligently retrieves relevant database schema information to generate accurate SQL queries.

Features

• GPT-4 Integration: Utilizes OpenAI's GPT-4 model to generate SQL queries from natural language input.
• FAISS Vector Database: Stores and retrieves database schema information using FAISS, enabling efficient semantic search.
• Retrieval-Augmented Generation (RAG): Combines the power of semantic search and GPT-4 to improve query generation.
• Customizable: Easily configurable to work with different SQL database schemas.
• Future Web Tool: Plans to create a web-based interface for ease of use.

Prerequisites

• Python 3.7+
• OpenAI API Key (for GPT-4 access)
• FAISS library for vector database handling

How It Works

1. Schema Embeddings: The tool connects to your SQL database, retrieves the schema, and generates embeddings using the Sentence-Transformer model or OpenAI GPT-4.

2. Query Generation: The user provides a natural language query, which is processed using RAG and FAISS to retrieve relevant schema details, and then GPT-4 generates the corresponding SQL query.

3. Vector Database: FAISS is used for storing and retrieving embeddings of the schema for efficient search.

Example Usage

To generate a SQL query for "Find users with orders above a certain total", the system retrieves relevant schema information and uses GPT-4 to generate the query:

SELECT ProductID, ProductName
FROM Products
WHERE ProductID NOT IN (SELECT ProductID FROM OrderDetails)