Inspiring example, thank you!
I had a look at the source to your aiotg package for writing such Telegram bots. IIUC, you run a single Python process using asyncio, and that uses coroutines to handle all requests (in and out).
What is your experience w.r.t. scalability in practice, for example with your music bot? (iow have you run into any issues with large numbers of users interacting with the bot?)