MusicFlow API

MusicFlow is an API that use the Spotify's API to predict the number of song streams using the Spotify audio features and track metadata.

The MusicFlow API uses the personal Spotify streaming history to train a machine learning model that predicts the number of streams for a given song. The training dataset uses number of stream values between 0 and 30.

The MusicFlow API is built using the FastAPI framework, which makes it fast, easy to use, and well-documented.

Available Endpoints

The MusicFlow API provides the following endpoints:

• musicflow.link/ - The root page of the MusicFlow API.
• musicflow.link/api/health - Returns the health status of the API.
• musicflow.link/api/raw_features/?song={song}&artist={artist} - Returns the unprocessed audio features of a song using the Spotify API.
• musicflow.link/api/features/?song={song}&artist={artist} - Returns the unprocessed audio features of a song.
• musicflow.link/api/prediction/?song={song}&artist={artist} - Predicts the number of streams for a given song based on the Spotify audio features and track metadata.

Examples

root

https://musicflow.link/

/prediction

Let's predict the number of streams for "The Less I know the better" from "Tame Impala":

https://musicflow.link/api/prediction/?song=The Less I know the better&artist=Tame Impala

Response:

{
    "song": "The Less I Know The Better",
    "artist": "Tame Impala",
    "prediction": 5.35,
    "description": (
        "The predicted number of future streams of the song based on the Spotify"
        " API audio features."
    ),
    "song_metadata": {
        "song": "The Less I Know The Better",
        "artist": ["Tame Impala"],
        "album": "Currents",
    },
    "message": {"emoji": "😍", "text": "What a banger!"},
}

/features

https://musicflow.link/api/features/?song=The Less I know the better&artist=Tame Impala

Response:

{
  "track":{
    "track_name":"The less I know the better",
    "artist_name":"Tame Impala",
    "number_of_available_markets":183,
    "num_artists":1,
    "duration_ms":216320,
    "explicit":true,
    "popularity":83,
    "isrc":"AUUM71500303"
  },
  "album":{
    "release_date_precision":"day",
    "release_year":2015,
    "release_month":7,
    "release_day":17,
    "date_is_complete":true,
    "album":"Currents"
  },
  "audio_features":{
    "danceability":0.64,
    "energy":0.74,
    "key":4,
    "loudness":-4.083,
    "mode":1,
    "speechiness":0.0284,
    "acousticness":0.0115,
    "instrumentalness":0.00678,
    "liveness":0.167,
    "valence":0.785,
    "tempo":116.879,
    "id":"6K4t31amVTZDgR3sKmwUJJ",
    "time_signature":4
  }
}

Get Spotify API Access Client ID and Client Secret

To access the Spotify API, you need to create a Spotify Developer account and obtain a Client ID and Client Secret. Follow these steps:

• Go to the Spotify Developer Dashboard and log in with your Spotify account.
• Click on "Create an app" and fill out the required information.
• Once you have created the app, you will see your Client ID and Client Secret on the app dashboard.
• Add them to your env file

Model Features derived from the Spotify API

Used Spotify API endpoints

The model features are created based on the responses from the endpoints below.

• tracks
• audio_features
• audio_anaysis (currenly not used)

Model Features

Column Name	Description	Data Type	Allowed Value Ranges
number_of_available_markets	The number of markets in which the track can be played	integer	0-79
release_year	The year the track was released	integer	0 <= year <= current year
release_month	The month the track was released	integer	1 <= month <= 12
release_day	The day the track was released	integer	1 <= day <= 31
date_is_complete	Indicates if the release date is completely known	boolean	true/false
num_artists	The number of artists credited on the track	integer	n/a
duration_ms	The duration of the track in milliseconds	integer	n/a
explicit	Indicates if the track contains explicit content	boolean	true/false
popularity	The popularity of the track	integer	0 <= popularity <= 100
danceability	The danceability of the track	float	0 <= danceability <= 1
energy	The energy of the track	float	0 <= energy <= 1
loudness	The loudness of the track in decibels (dB)	float	-60 <= loudness <= 0
mode	The modality (major or minor) of the track	integer	0 = minor, 1 = major
speechiness	The presence of spoken words in the track	float	0 <= speechiness <= 1
acousticness	The acousticness of the track	float	0 <= acousticness <= 1
instrumentalness	The instrumentalness of the track	float	0 <= instrumentalness <= 1
liveness	The liveness of the track	float	0 <= liveness <= 1
valence	The valence of the track	float	0 <= valence <= 1
tempo	The tempo of the track in beats per minute (BPM)	float	0 <= tempo <= 500
time_signature	The time signature of the track	integer	0 <= time_signature <= 7
A	The key of the track with pitch class A	integer	0 = not present, 1 = present
A#/Bb	The key of the track with pitch class A#/Bb	integer	0 = not present, 1 = present
B	The key of the track with pitch class B	integer	0 = not present, 1 = present
C	The key of the track with pitch class C	integer	0 = not present, 1 = present
C#/Db	The key of the track with pitch class C#/Db	integer	0 = not present, 1 = present
D	The key of the track with pitch class D	integer	0 = not present, 1 = present
D#/Eb	The key of the track with pitch class D#/Eb	integer	0 = not present, 1 = present
E	The key of the track with pitch class E	integer	0 = not present, 1 = present
F	The key of the track with pitch class F	integer	0 = not present, 1 = present
F#/Gb	The key of the track with pitch class F#/Gb	integer	0 = not present, 1 = present
G	The key of the track with pitch class G	integer	0 = not present, 1 = present
G#/Ab	The key of the track with pitch class G#/Ab	integer	0 = not present, 1 = present
Unknown	The key of the track is unknown	integer	0 = not present, 1 = present

Next steps and ongoing work

• Version model
• Add logging
• Allow for continous training

• Improve feature processing pipeline

• upload the request to S3 (add metadata from request IP, etc.)
• [] Move all configrations outside of code into a yaml file and settings classes
• Add Deep learning models
• [] add model baseline
• [] version the data
• [] add model versioning