I’m experiencing something strange (or that i’ve simply misunderstood) on requests counting with Assistants API. Each time i start a run it was counted more then one request. Is it normal? If yes, how are request counted with Assistants?
I think you’d get a lot of requests embedded in that run. For example, there’s getting the status of the run that’s done via polling, so there are N number of status requests. Also, if you defined custom functions there is the request involved in supplying the answer to the function calls. Furthermore, there’s the request to get the resultant message out of the thread.
I thought so, so i tryied to make a very simple request adding a message to a thread (threads.messages.create) but it’s not counted any request.
Only running the assistants seems to increase the count.
One thing is sure: my context could be very large becouse i’m using assistant for retrieval purpose (i’ve some file attached).