Question: How can I utilize categories and other features for new users in the Sequential Retriever

I implemented the Sequential Retriever using the tutorial and was hoping to expand it for new users by taking in an additional parameter which is the categories the user prefers. I'm hoping to make the predictions like so

movies_watched = ["1", "2", "3"]
preferred_categories = ["Action", "Comedy"]

recommendations = predict(movies_watched, preferred_categories)

The goal is to provide recommendations for new users by passing an empty array for movies_watched and passing the collected categories.

Really appreciate some guidance on how I have achieve this.

Start with the Sequential Retriever tutorial and the proposed predict(movies_watched, preferred_categories) entry point, then inspect how user and item features are represented. Clarify the intended cold-start behavior when movies_watched is empty and define an example or test showing recommendations based on preferred categories.