TableGPT

A conversational AI assistant that transforms structured data into insights using Large Language Models (LLMs). Inspired by a real deployment at Moneris using Azure OpenAI.

🧠 What It Does

This project demonstrates how to:

• Load and interpret structured business tables (e.g., sales, transactions, forecasts)
• Accept natural language queries like:
  “What was the top-performing product last quarter?”
  “Compare merchant volumes year-over-year.”
• Parse the question → construct a structured SQL-like query → extract answer → summarize with LLM

🛠 Features

• Modular LangChain-style pipeline
• Prompt-engineered context windows for structured data
• OpenAI (or Azure OpenAI) support
• Multi-table support (simulate sales, merchants, forecasts, fraud)

📁 Folder Structure

TableGPT/
├── data/                # Sample structured datasets (CSV)
├── notebooks/           # Test notebooks for question-answering
├── results/             # Query logs, responses, summaries
├── src/                 # Core logic (prompt builder, QA engine)
├── requirements.txt     # Python dependencies
└── README.md            # Project overview

🚀 Getting Started

1. Clone the repo:

git clone https://github.com/mageed-ghaleb/TableGPT.git
cd TableGPT

2. Install dependencies:

pip install -r requirements.txt

3. Add your OpenAI or Azure OpenAI key as environment variable:

export OPENAI_API_KEY=your-key-here

4. Run the pipeline (coming soon):

python src/query_interface.py

📊 Example Use Cases

• Conversational business dashboards
• Natural language BI queries
• GenAI layer on top of data warehouses

👨‍💻 Author

Developed by Mageed Ghaleb – Senior Data Scientist | AI Engineer | Co-Founder of MetaForge
Based on enterprise GenAI deployment for business intelligence.

📄 License

MIT License – Free to use with attribution.