Skip to content
/ gt-chat Public

Buzz AI, aka gt-chat, is a fast and intuitive question-answering chatbot for Georgia Tech. Powered by Next.js, FastAPI, and OpenAI, it sources from 14k GaTech websites for accurate answers. Fork, star, and join the buzz!

Notifications You must be signed in to change notification settings

hxu296/gt-chat

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Buzz AI

buzzai-badge

BuzzAI or gt-chat is a question-answering chatbot that is designed to answer any questions about GaTech. The chatbot is powered by Next.js, FastAPI, and OpenAI, and it provides a fast and intuitive interface for finding answers to commonly asked questions by sourcing from over 14k Georgia Tech websites. The webpages are collected, cleaned, and splitted into 49k 1024-character document chunks and embeded into 1536-dimension vectors using OpenAI's text-embedding-ada-002 for $4. The produced FAISS vector index enables the chatbot to quickly find the most relevant information to a question and use it to generate an answer. You can try it out at gt-chat.org!

Screenshot of BuzzAI

Getting Started

To use BuzzAI, you can simply go to the gt-chat.org or clone the repository and run the chatbot locally.

To run the chatbot locally, you will need to clone the repository and follow the instructions below.

Frontend Installation

To start the frontend you will need:

  • Node.js 12 or later installed
  • The URL of the backend server as an environment variable DOMAIN

Deployment

Deploy with Vercel

Installing

Clone the repository and then run the following command to install the required dependencies:

npm install

Running the Chatbot

To start the chatbot, run the following command:

npm run dev

This will start the Next.js development server and open the chatbot in your default browser at http://localhost:3000.

Backend Installation

The backend for the project is a python fastapi server that uses the LangChain + OpenAI API to generate answer for to /qa GET endpoint.

Supabase

You will need to set up a Supabase project and create a table called qa_log with the following schema:

Column Name Data Type
id uuid
created_at timestamp
question text
answer text
success boolean

You need the Supabase project URL and service key to set up the environment variables later.

Deployment

Change the Railway build command to bash build.sh and it should work out of the box.

Deploy on Railway

To run the server locally:

Step 1: Set up python environment and fetch OpenAI embeddings

python3 -m venv venv
bash build.sh

Step 2: Set up environment variables

export OPENAI_API_KEY=<your key>
export SUPABASE_URL=<supabase project url>
export SUPABASE_KEY=<supabase project *service* key, not annon key>

Step 3: Run Local Server

python main.py

Special Thanks

Risingwave for the chatgpt-plugin repo that I used as a starting point for the backend.

odysa for the BuckyAI repo that I used as a starting point for the frontend.

Contributing

Pull requests are always welcomed!

About

Buzz AI, aka gt-chat, is a fast and intuitive question-answering chatbot for Georgia Tech. Powered by Next.js, FastAPI, and OpenAI, it sources from 14k GaTech websites for accurate answers. Fork, star, and join the buzz!

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published