Use the circular Web AI Agent microphone button in the center above in a push to talk style. That means when you click on the button and hold it will listen to your voice, and when you let go, it stops listening. Think what you want to say before you push the button so you speak fluently for best results. Let go when you have finished talking!
For example try asking something like: "Make a playlist for coding or deep work" or "Create a playist for 80s music"
Glad you asked - check out this YouTube course that goes from zero to hero!
Make a new app on Spotify developer and then be sure to set redirect URL to be:
https://spotify-web-ai-agent-dj.glitch.me/index.html
Then locate and enter your Client ID and Client Secret below for your app and click auth - you will just need to do this once
Agent thinking for the curious folk will show here
This demo, created by Jason Mayes at Google, uses Web AI large language models to power this experience to provide Agent like behaviours. That means the AI model is downloaded to your machine in the browser and run entirely locally - so no calls to the cloud - keeping costs to zero and your privacy is preserved.
Specifically the Google Gemma 2, 2B parameter model, is downloaded and executed powered by the MediaPipe Web LLM library.
While you wait for this to finish, watch this 5 min video (click play and expand full screen) to see the demo in action so you know how to use it once it has downloaded!