Hi there, I want to make sure. I use gpt-4o-mini and RAG for my document. In the playground, everytime I update the system prompt it’s instantly change the style of the answer output. But, I found that when I use the API on my corporate website chatbot, the answer style not instantly updated (I need more minutes, before the answer match with the new system instructions).
My question is, is that normal? or it should be change real time?
Which API is your chatbot using?
It may relate the way the system prompt is maintained for a current run.
You will need to dig into your code to find out.
If you were using Chat Completions, though, the change would be instant and applied on the very next call as there is no memory.
I use the thread API. I already check anything on my code and still have no clue about this. Until now, I must wait about 1 hour and more to check is the system prompt already implement or not