drarvindsathi / katana-assistant

Text based assistant powered by Machine Learning and NLP

Katana Assistant

Machine Learning based agent, helps to enable business automation.

Technology: TensorFlow, Keras, Flask, Python, Node.js, JavaScript

Author: Katana, Red Samurai Consulting, Andrejus Baranovskis

Instructions

- Machine Learning

Install TensorFlow

pip install tensorflow

Install Keras

pip install keras

Model code is located in mlmodels folder.

Sample set of intents is available in the file mlmodels/intents.json. There is pre-built model in mlmodels/katana-assistant-model.pkl. If you want to rebuild model - run Jupyter notebook katana-assistant-model.ipynb

To start Katana assistant model endpoint in the background process run it with PM2 manager:

pm2 start katana-assistant-endpoint.py

This will start endpoint on port 5001

- Node.js Backend

Backend code is located in mlbackend folder.

Run backend with PM2 manager on port 3000:

PORT=3000 pm2 start -l 0 ./bin/www

Socket.IO endpoint will be started on port 8000. Check mlbackend/routes/assistant.js

- JavaScript Frontend

Frontend code is located in mlfrontend folder.

UI client is implemented with Oracle JET. Follow instructions to install Oracle JET.

Navigate to folder mlfrontend/socketiojet and run this command to setup required libraries:

ojet restore

Run UI client:

ojet --server-port=8010 serve

This will start frontend on port 8010.

License

Licensed under the Apache License, Version 2.0. Copyright 2019 Red Samurai Consulting. Copy of the license.