I'm open for any feedback or code-review. I used Tornado and Tornado-Redis for non-blocking redis client. I'm using redis' pub/sub channels for real-time updates. In frontend side, I choose the React.js for building interface.

You have a few hardcodes that assume where the service is being run: https://github.com/fatiherikli/klassify/search?utf8=%E2%9C%9...

The first one is for development environment. But you're right, I should make it configurable with command line arguments. So user can give any hostname or api. Currently, only the server port is configurable.

Thank you for feedback.

what does it used as the backend? The actual classification system used?

Looks like it uses Python on the backend, with nltk. http://www.nltk.org/

Using nltk for tokenizing and stemming text. There is no classification backend, I implemented a Naive Bayes classifier.