Track Recommender

A simple recommendation engine that recommends tracks based on the user input. The user is able to communicate with it via simple API endpoint.
Currently, the project can recommend tracks based on user input, which should be the track IDs, it is planned to implement broader search functionality to allow to search by album or artist.
The recommendation engine runs on KDTree algorithm, which establishes the nearest neighbours using tree decisions and then stores them, so it can be computationally efficient, since it does not have to calculate the distances on each API call.

How to Install and Use

In order to interact with the project, you need to install Docker, then it is enough to just clone this repository

git clone https://github.com/rmnskb/track-recommender.git
cd track-recommender

Create a .env file with your desired database credentials:

echo "POSTGRES_USER=YOUR_POSTGRES_USER" >> .env
echo "POSTGRES_PASSWORD=YOUR_POSTGRES_PASSWORD" >> .env
echo "POSTGRES_DB=YOUR_POSTGRES_DB" >> .env

And run the docker compose:

# verify that the docker is installed correctly
docker run hello-world 

# build and run the project
docker-compose up -d --build

Please ensure that the default ports 5432 and 5000 are not used by any other application, please adjust the docker-compose.yaml and Dockerfile file accordingly otherwise.
At this point, the API should be up and running. At this time, it supports only the POST call that accepts track IDs, an example call may look like following:

curl -X POST http://127.0.0.1:5000/api/v1/recommend \
 -H "Content-Type: application/json" \
 -d '{"ids": ["5SuOikwiRyPMVoIQDJUgSV", "1iJBSr7s7jYXzM8EGcbK5b"], "n_recs": 7}'
 
 # or alternatively, a GET request for the autocomplete endpoint
 curl -X GET http://127.0.0.1:5000/api/v1/autocomplete?q=Come

As you can see, the first call has two arguments, the track IDs and number of recommendations per song to return.

Frontend

You can also access the web app using port :3000 with you browser.

If you have a Spotify API id and secret, you can do following steps to try out the frontend:

cd app 
echo SPOTIFY_ID=YOUR_SPOTIFY_ID >> .env
echo SPOTIFY_SECRET=YOUR_SPOTIFY_SECRET >> .env

After setting up these variables, the application will handle overall communication with the Spotify API.

You can type in your search, choose the desired track from the suggested autocompletes and receive the recommendations. All cards are clickable as they lead to Spotify tracks.