I have configured the test so we can import modules via the esm without any errors using babel and I have written test for generateVideoTag util function.

don't mind the redlines, that's eslint acting up on my editor but the test runs.

Create the tweet endpoint

This is a post endpoint. When this endpoint is hit with a JSON payload as shown below, it would post the tweet and save it to the DB. If the text is more than 208 characters, an error message is thrown. The response would be the tweet data from Twitter.

{
    "tweet": "tweet text" 
}

For now, the authentication for tweeting is going to be a generated token hard coded as an env variable AUTH_TOKEN

In this PR

• Express was added to the project to enable it to bind with heroku's HTTP port.
• The Procfile was added to enable deployment to Heroku.

It closes #10

Till Further notice, this Project would be hosted on Heroku. With that in mind, it is required that it is bounded to a port to keep running, therefore a server would be set up. This server would also serve as the foundation for any other server-based functionality that would be enabled on this project.

A Heroku Procfile is also required.

This PR integrates the Twitter bot with audd.io's matching APIs. It also expanded the media converter function to cater for different scenarios like the length of the audio snippet to match and the expected output format.

Abstract a mention reply function and write a cronjob that would run every 5 minutes to get all new mentions and call the mention reply function to operate on them.

Problem description

By default, the bot parses the middle 3 seconds of the video to search for a song. But, it is possible for a single video to have more than one song in it at different time stamps, or for some parts of a song to be audibly more distinct than others. Create a functionality that allows a user to add a timestamp when calling the bot. As such: @whatsongishere in mm:ss-mm2:ss2.

• It is important for the time stamp range to be valid and within the duration of the video.
• The timestamp range should be a minimum of 3 seconds, but a maximum of 6 seconds.
• If a start timestamp is passed without an end timestamp @whatsongishere in mm:ss, the end timestamp is taken to be 3 seconds after the start
• If an end timestamp is passed without an end timestamp @whatsongishere in -mm:ss, the start timestamp is taken to be 3 seconds before the end timestamp.
• All other input after what has been parsed as the end timestamp can be ignored.
• Try not to install any new package

Recommended Approach

• Create a utility function that parses timestamps if available in the bot tag
• Move the logic that gets the duration of the tagged video and parses the default time stamps from the convertToRaw function in rawConverter to the utility function just created.
• The output of the utility function should either be an object in the form:

{
  "start": "mm:ss",
  "end": "mm2:ss2"
}

or throw an error due to an invalid timestamp. The error thrown should cause the mention to be saved to the DB(This is important as we don't want the particular mention to be handled again), and a response tweet in the form Sorry @name, it seems you have passed an invalid timestamp.

• The convertToRaw function should take two new parameters start and end which should be valid timestamps from the utility function you have created.

An audio fingerprint is a unique characteristic of any piece of audio that can be used by programs to identify the song from a given snippet.

With that said, I hypothesize that all that we consider music (instrumentals, rap, songs, hymns, etc) have common characteristics in their fingerprints that an ML classifier model can be trained on to be able to differentiate between music and "sound". As I intend to have the entire project written in javascript, I want a function/module that can generate the fingerprint of given audio, which would be plugged into whatever system/module makes use of it.