What to Watch?

This is a demo of a basic web app that helps people decide what to watch next. The goal was to allow people to quickly narrow down a long list of shows and come up with a few suggestions.

The app revolves around a Tinder-like 'yup' / 'nope' experience, with some additional filtering on top. The user can either 'save' a movie or 'shelve' it:

Any item can be opened to see additional details and get linked to other informational sites.

A small visualization is included in the top-left corner which shows the distribution of audience and critic ratings, as well as the current saved/shelved state. The audience and critic minimum ratings can be adjusted concurrently using the draggable handle in the chart.

Requirements & Setup

Tested with Python 3.5.1 and Django 1.9.

pip install -r requirements.txt   # requires Django 1.9
python manage.py migrate
python manage.py loaddata movie_fixture.json  # populate DB
python manage.py runserver

Implementation Notes

Data extraction and pre-processing is performed using a Jupyter Python notebook in the data-utils directory.
Poster thumbnails are delivered in sprite sheets for efficiency.
High-level data is included with the initial request by the server so all filtering can happen client-side. Additional movie meta-data is retrieved on-demand.
Front-end is implemented in riot.js.
Chart built with D3.js
For the sake of simplicity, this Django application conflates the concept of 'projects' and 'apps', putting both into a hybrid project/app module called starter.

Missing Features

Ability to handle an arbitrary number of movies. Current limit is unknown.
Check if it's available to stream on Netflix
Per-user customization, including saving app state
Better layout and scrolling behaviors
Convert Jupyter notebook to command-line Python scripts
Ensemble RNN-based recommendation engine
...

Design

The initial sketch, based on the use case: "what should I watch next?"

Copyright

All code is freely usable under the MIT license.