How exactly does the chatgpt user interface retrieve data?

Using the browser’s developer tools we can “hijack” the fetch/xhr requests or observe any established websocket connection made from the a given platform to the server.

Listening to the outgoing traffic on twitter for example is pretty easy since its nothing but bunch of rest api calls. But chatgpt seems to be more complicated.

It does indeed send a request (“https://chat.openai.com/backend-api/conversation”) which contains the user question — but the response does not contain the answer chatgpt provides.

There is however a so called eventstream tab – does this imply the chatgpt response download is issued after the fetch?

So i make a fetch request to “backend-api/conversation” and then wait for any signal to download the response?

What technology is used specifically? I genuinely try to understand the mechanism behind this.