I built a GPT4-Powered Song Recommendation engine by writing mostly prompts
Made a tik tok about it
So in Airkit we have a GPT “copilot” called Airkit Assistant, which will help you generate web pages and workflows by using just prompts. We also have an action that is connected to GPT4 for you to bring GPT4 into your apps. Check out this app I built in ~10 mins.
Prompts that I used:
Building the Web Page
/make a form with a header called “Song Recommendation Engine”. Have a single text area input that asks for a users favorite artist and genre. Have a button that says submit. Add a label underneath to display recommended artist and song
Building the Data Flow
/make a data flow to get the access token from spotify and searches Spotify. You will only need one input, which is the query string for the search API. For the access token, I have the client id and client secret. Do not add these as inputs, I will hardcode them in the HTTP Request. Use the client credentials grant_type Then after you get the access token, do a get request on the spotify search API and passes the input in as the q query parameter. Remember to pass in the access token from the previous operation. Then filter the response of the API request to get the URL of the first response and the image URL of the first response. Do this in one transformation so that it outputs an object like {“image” : <image url> “url”: <url>}
Recommending the Song
You are a song recommendation engine assistant. Given the following artists and/or genres:
text_area
Recommend a new track that the user would like based on the given artist and genre .
Then output a recommended track The format of this should look like <Recommended Track Name> by <Recommended Artist Name>. Do not wrap the output in quotes.