I am still simply doing plain fetch requests to LLMs APIs, and it works great, 10/10 would recommend.