Live version here.
Note: As the project is now completed, we have taken down the backend database so AWS RDS won't start charging us.
WhatBot is a student-tutor support chatbot developed for Capstone Project of Master of IT at UNSW. This repository contains all the packages for the WhatBot software and it is managed as a mono-repo.
WhatBot uses a client-server architecture. The frontend is developed using React/Redux with semantic-UI as the CSS engine. The backend composes of a Python Flask server, AWS RDS running PostgreSQL, AWS S3 and Google's Dialogflow to make the whole system work. The PostgreSQL database is used as the knowledge base while AWS S3 is used to store AI training file contents. Google's Dialogflow API is the backbone of WhatBot's chatbot capability.
The key difference between WhatBot and other chatbots are that it can be easily configured to train for different purposes. It provides AI traning capability such that user can simply upload text files of questions and answers that the Dialogflow agent can be quickly trained through WhatBot's interfaces.
Read more about WhatBot's capabilities and architectures in Project_FinalReport.pdf
Running locally
To run the whole system locally, a Makefile has being created. In this folder run:
Make
and you should see all the dependencies being installed and the whole app running afterwards.
Alternatively, you can see the deployed version which is live at https://whatbot9900.herokuapp.com/. The deployed version is the most production ready and stable one. To see how to do deployment or how it is done, please see the Deployment section below.
WhatBot can be found and used https://whatbot9900.herokuapp.com/.
It is deployed on Heroku using Docker. As a client-server architecture is used, the frontend and backend are each running in their own Docker container. In production build, both frontend and backend are running on their own nginx server inside their respective Docker container.
Continuous integration and continous deployment using CircleCI has being setup to automatically deploy both frontend and backend to Heroku when changes to master branch is detected and the changes passes the tests.
Deploying to Heroku
To manually deploy to Heroku, you need to have Heroku CLI and Docker installed.
Install Heroku CLI here.
Install Docker here.
After you have installed both of them:
Log into Heroku CLI:
heroku login
Log into Heroku container registry:
heroku container:login
Deploy frontend
cd frontend
heroku container:push web -a whatbot9900
heroku container:release web -a whatbot9900
Deploy backend
cd backend
heroku container:push web -a whatbot9900backend
heroku container:release web -a whatbot9900backend
More information on Heroku container registry here