Overview
Online chatbot via LINE platform which can automatically respond to users' chat and requests to make a purchase of a prepaid voucher for all cellular operators within Indonesia. The chatbot uses NLP build with TensorFlow (Python).
This project is no longer active, cause the price to maintain EC2 + RDS instance in AWS is too expensive. It is impossible to migrate to the serverless function because the latency difference is too high for a chatbot and it needs high computing power to train and run the ML model itself. Also, there are no market for this since people just can use m-banking or other e-commerce to buy prepaid voucher or "pulsa" anyway.
Tech Stacks
Backend - Python • Flask • MySQL + SQL-Alchemy • Docker • EC2 + RDS • TensorFlow • LINE API • Midtrans + Pulsa Provider
Frontend - React + MaterialUI • Line Liff
Tools - Github • Travis CI
Lessons Learned & Notes
Here are some notes and lessons that I learned in the process of developing this project:
-
Integrate midtrans and pulsa provider
-
Creating an NLP model and integrating it with a chatbot
-
EC2 + RDS is too expensive and hard to maintain
-
You can develop a fully functional web app that runs on LINE chat and it's very powerful. The framework is called LINE Liff
-
You can customize chat responses in LINE using Flex Message API to have a beautiful interface in a LINE chat. Here is the documentation