but keep in mind that it won't work in serverless platforms such as Vercel, Cloudflare and so on. it needs to be hosted on long lasting instances such as AWS, GCP and so on.

As SSE is supposed to run for long time as OpenAI API does.

Or,

you are going to impl a ChatGPT like app, use Next.js as a proxy to OpenAI, just use Vercel AI SDK, which converts SSE to streaming nicely. sample codes are aloso available in that repo above.