Serverless functions for calling LLM?

I’m building a Web app that calls an LLM and I noticed the limitations for the serverless functions. I’ve already started hitting the 10 second limit and I’m not even doing the complex stuff yet.

Am I approaching this wrong or deploying to the wrong place or need to consider different?
I see people have increased their timeout setting some. I upgraded to the pro plan to see if I could do it through the website but I could not. I’m not even sure if 26 seconds if that is the true upper limit, will be enough time. I do want the user experience to be real time not delayed. I could make the response start sending information back immediately, but it won’t finish quickly either.

You need to stream the response to the client. You can use Edge Functions for that for streaming without a time limit.