I've the impression that using the GPT 4o Realtime API the Knowledge Base files are not getting considered?

@simonpfeifhofer I understand your concern. The GPT-4 Realtime API should indeed consider the Knowledge Base files if they are configured correctly. If you're noticing that the Knowledge Base doesn't seem to be taken into account, there might be an issue with the configuration or how the API is being called. To assist you better, could you provide more details about how you're implementing the Knowledge Base with your API calls? This will help me understand the situation and provide you with the most accurate assistance.

Just using the UI in a first step

@simonpfeifhofer as of now we are migrating our KB to new providers which will be completed by end of this week. So, I suggest to wait for your answer till end of this week with new updates.

@Shubham Bajaj thanks a lot. How will the KB have impact on the pricing? If a part of the prompt gets moved to a text-file which gets added to the KB. What will be the price difference between the two options (with- and without KB).

For KB previously we were not charging anything, as it was included as token cost so now it will follow the same model the number of tokens consumed in the call.

How will the KB work with GPT-4 Realtime API? How will the files be added to an assistant? The question targets the following issue: An a bit larger prompt will get extremely expensive and will not be an option for many use-cases.

Hey @simonpfeifhofer , as of right now we do not support KB with the real time api, but we will look to support it in the near future