Different API requests in the same context

Hello, we use an api request to generate product descriptions. How can we setup a “Chatroom” or ID with the same system information once to not always repeat the same system setup, just sending the content?

System:
"You will be provided with a product name, product type, brand and keywords. Your task is to generate a product description for a web shop using headlines starting with

tags and the following headlines with

tags.

This send only once.The following requests should referee to this (as in the same chat room)

The content of the future requests should only be:
Product typ: Watchbox
Product name: Berlin
Brand: Benson
Keywords: xxx, xxx, xxx

Thanks for help,
Michael

Welcome to the forum. The API has no “state,” so you need to send the system message each time for it to be used.

1 Like

Oh really, thanks. I thought OpenAI could set an “virtual” chat room as we have an account and the room could have some sort of ChatID or RoomID. Would be the counterpart for the chat room in ChatGPT in our account. Should technically not that problem. Is the same sort of data storage in our account.

I hope there will be a solution from OpenAI soon.

Could be “fine-tuning” a way to achieve this?

Maintaining per/room/chat conversation state is really expensive and OpenAI would need to bill you for that storage separately. They’re not in the storage provider business so they make you do all your own state storage.

It’s true that they do this storage for the public facing GPT Service but they either make up for those costs in subscription revenue or they just eat it.