Sales Data Question-Answering with GPT-J and ChromaDB

This project demonstrates a retrieval-augmented question-answering (QA) system for sales data stored in Excel files. The system combines ChromaDB for efficient document retrieval, HuggingFace embeddings for text encoding, and GPT-J-6B for natural language responses. This setup allows users to query sales data in plain English and receive accurate, context-aware answers.

---

Features

• Load and preprocess Excel files for sales data analysis.
• Create a vector store using ChromaDB for fast document retrieval.
• Generate human-like responses with GPT-J-6B.
• Modular and scalable Python architecture.
• Example interactions to demonstrate capabilities.

---

Getting Started

1. Clone the Repository

git clone https://github.com/abdulvahapmutlu/sales-data-qa.git
cd sales-data-qa

2. Install Dependencies

Ensure you have Python 3.8+ installed. Then, install the required libraries:

pip install -r requirements.txt

3. Add Your Sales Data

Place your Excel files in the data/ directory.

---

Usage

Run the Application

Execute the following command to start the application:

python src/app.py

Ask Questions

You can interact with the chatbot by asking questions about your sales data, such as:

• "What is the total sales in the sales data?"
• "Can you provide a breakdown of sales by region?"

---

Components

Data Preprocessing

• The data_loader.py script loads and preprocesses Excel files to create a unified text column for retrieval.

ChromaDB Integration

• chromadb_manager.py manages the creation of vector stores and the addition of documents in batches.

Language Model Initialization

• model_initializer.py initializes the GPT-J-6B model for generating responses.

Retrieval QA

• retrieval_qa.py defines the question-answering chain using LangChain components.

Main Application

• app.py orchestrates all components and provides a user-friendly interface.

---

License

This project is licensed under the MIT License.

---

Contributing

Feel free to submit pull requests or open issues for any suggestions or improvements.

---

Acknowledgments

• ChromaDB: For vector storage.
• HuggingFace: For providing state-of-the-art language models.
• LangChain: For seamless integration of retrieval-augmented generation pipelines.