How do I balance user control and policy compliance in a playlist generator plugin for ChatGPT and for YouTube Music?

I'm developing MeloGenAI, a free and open source playlist generator plugin for ChatGPT and YouTube Music. I'm seeking advice on balancing user control and compliance with YouTube Data API Services - Developer Policies.

Functionality

MeloGenAI uses the YouTube Data API to create a playlist based on a provided title, list of songs, and privacy setting. It searches YouTube for each song, takes the first result, and adds it to the playlist. The results can be unexpected if it doesn't find the expected version of a song.

Policy Considerations

I'm trying to reconcile two conflicting requirements:

1. Don't steer model behavior: OpenAI's plugin policies discourage steering the model's behavior. For instance, asking for user consent before generating the playlist might be seen as steering and could disrupt the user experience.

2. Give users control: The YouTube API Services - Developer Policies require users to be aware of and actively consent to the actions an API Client takes on their behalf.

I'm unsure about the level of control and transparency to provide users, and how to comply with YouTube's API guidelines. Should I break the plugin into multiple endpoints to give more control to users?

Here's an example where the plugin directly made the playlist without telling the user the songs it was going to search for: MeloGenAI example

Would this comply with YouTube's API developer policies? Any advice or feedback is welcome. Thanks!