Please supply a secrets.py
When you run this script it gets the Tracks audio features and puts them in a txt file for that track.
Obviously have to have a track playing to do this. Although static data should be available as a later addition.
I'm pretty sure these audio features can be used as part of the seeding when getting recommendations.
So things like danceability, tempo, key, energy etc. can be used so as to get similar tracks that maybe we haven't heard.
The next bit is to find out how to use them in the search. 

Watch This space 

Ollie