Hey! Have you tried out Edge Streaming yet? It uses the Edge Runtime which is a fraction of the cost of serverless functions and lets you stream responses for much longer than 10 seconds, giving you the "chatting" effect that you see on ChatGPT.

Docs: http://vercel.fyi/streaming Example: https://vercel.com/blog/gpt-3-app-next-js-vercel-edge-functi...