Trello Assistant

Bringing artificial intelligence to your boards and cards.

A proof of concept assistant that suggests labels based on card titles.

Notice that this is a work in progress, but is functional.

Usage:

This application can be deployed to Heroku.

Before building service, make sure to update .env with the developer key to be used.

The key must also be updated in public/auth.html:13

There's a live server that can be used for testing by creating a power-up with url:

https://pure-wave-30477.herokuapp.com/index.html

Implementation:

The current proof of concept implementation takes a card id, which is used to retrieve the necessary card information, along with the cards in the same bard.

Then it calculates the term frequency/inverse document frequency vectors for the titles and descriptions of all cards. These are used to calculate the cosine distance between each card and the one we want to tag.

Finally, each card get to vote in their label set, weighted by the similarity, and the one that achieves the highest score is applied to the card.

Contents:

• server.py Web service implementation.

• recommender.py Label recommendation implementation.

• requirements.txt List of python libraries needed.

• settings.py Configuration.

• tests.py Unit testing.

• public/index.html Trello power-up main point of entry.

• public/auth.html Power-up authentication.

Requirements:

• Python, tested on v3.5.2.
• Bottle, for web service.
• requests, for querying Trello API.
• scikit-learn, for text vectorizing and similarity metric.

Make sure to check the requirements.txt file for further information.