Chatbot app using GPT with Retrieval Augmented Generation

This repository contains a basic adaptable Python application designed to work with custom data & OpenAI's GPT models. Utilizing the LangChain library for LLM access & processing PDF files and, Streamlit for creating a self-hosted web app chat interface, this project enables enhanced interaction with AI models by supporting the retrieval-augmented generation concept.

Retrieval Augmented Generation (RAG) is a technique that combines the generative capabilities of models like GPT with a retrieval component. This approach allows the model to pull in relevant information from a dataset (in this case, PDF documents) at runtime, leading to outputs that are informed by a broader context than the model's pre-trained knowledge.

Features

• Custom Data Processing: Process PDF files to use as custom data for GPT model inference.
• Retrieval-Augmented Generation: Enhance GPT model responses by integrating retrieval from custom data sources.
• Streamlit Web App: Interact with the OpenAI's GPT model through a user-friendly chat interface hosted on your own server.
• OpenAI API Integration: Seamlessly use OpenAI's API with your own API key for model access.

Prerequisites

• Python 3.10 or later.
• An active OpenAI API key.

Installation

Clone the repository and install the dependencies:

git clone https://github.com/ASHISRAVINDRAN/custom-chatbot-app.git
cd custom-chatbot-app
pip install -r requirements.txt

Usage

First put any of your data in PDF format in the data folder. Update your OPENAI API key in the secretkey.py file. Then, run the Streamlit web app using the command:

streamlit run chatbot.py

Note

Currently, the code uses the MITK's documentation as custom data for demonstration purposes.

Disclaimer

This project is only a proof-of-concept for experimental purpose. It could lack a lot of features including prompt engineering and might have bugs.

Feel free to fork &/or contribute to this repo.