Temporal recently put out a wonderful demo of using Temporal to make prod-ready OpenAI agents, and I’ve seen many around asking a particular question about Temporal itself:

How do I stream responses from Temporal?

It’s a natural question, given how often SSE responses are used with LLMs. There are a few ways to approach this problem, but I will focus on a singular solution. I implemented my own Worker(s) -> SSE stream implementation about a year ago to handle fan-out notifications, and will share it with you here. I do not prescribe that this is the ideal solution, instead, this is meant to tease out some creative solutions from you, the reader.