SQL Server RAG Application

This application implements a Retrieval-Augmented Generation (RAG) system that connects to a SQL Server database and provides answers based on the data stored in it.

Prerequisites

• Python 3.8 or higher
• SQL Server
• OpenAI API key
• ODBC Driver for SQL Server

Installation

1. Clone this repository
2. Install the required dependencies:

   pip install -r requirements.txt

3. Install the SQL Server ODBC driver for your operating system

Configuration

1. Copy the .env.example file to .env
2. Update the .env file with your configuration:
   • Database connection details
   • OpenAI API key
   • Vector store path

Usage

1. Update the table_name variable in app.py with your target table name
2. Run the application:

   python app.py

The application will:

1. Connect to your SQL Server database
2. Create a vector store from your table data
3. Allow you to query the data using natural language

How it Works

1. The application connects to your SQL Server database
2. It fetches data from the specified table
3. The data is converted into text chunks and embedded using sentence transformers
4. A vector store is created using Chroma
5. When you ask a question, the system:
   • Retrieves relevant context from the vector store
   • Uses OpenAI's GPT model to generate an answer based on the context
   • Returns the answer along with the source documents

Customization

You can customize the following aspects:

• Embedding model by changing the model_name in the HuggingFaceEmbeddings initialization
• Chunk size and overlap in the RecursiveCharacterTextSplitter
• Prompt template in the get_qa_chain method
• LLM model and parameters in the ChatOpenAI initialization

License

MIT