How can I use "Hosted inference API"?

I alse love the idea of being able to run a quick demo so I was trying to implement it in my own model repo, but I don’t know how to configure it, can you help me?

Here are the files in my repo: