I’m using the OpenAI Realtime Transcriptions API via WebRTC. I expected there to be some sort of event logging token usage for each transcription event, but there is none. Additionally, although the Usage API allows you to programmatically get token usage for the model I’m using, I can’t filter by batch because there’s no way to specify the batch when creating the realtime transcription session. I think this was a huge oversight - there’s no way to track token usage on a per-session basis. I hope that this gets corrected at some point, but if not, does anyone have a workaround?