Closed chobberoni closed 8 months ago
Thanks, @chobberoni. Could you please attach your client-side logs from the session? Additionally, did this happen locally or with the deployed application? If locally, please provide the server logs as well. If deployed, approximately when did you reproduce this? I can try to find the logs in EB.
Additionally, if this was reproduced with the deployed version of the application, can you confirm whether you received the error less than 60 seconds after making the query? (I believe the Vercel request timeout is 60 seconds, so the timing will help inform if it's a Vercel timeout or a server error).
Thanks, @chobberoni. Could you please attach your client-side logs from the session? Additionally, did this happen locally or with the deployed application? If locally, please provide the server logs as well. If deployed, approximately when did you reproduce this? I can try to find the logs in EB.
The server logs from last week indicate one error: attempting to query a session that does not exist.
After checking the session in question, I confirmed that the bot was ejected as expected when the room expired. The default room expiry is 15 minutes and participants are set to be ejected by default. The bot was ejected as expected at this time. This would mean anyone trying to query this session after this time would fail, because the session no longer exists (as designed).
But the problem is that somehow participants were able to query this session despite it having been expired, and I suspect you did not ping the assistant API directly to do so (which would have explained this), right? In the dashboard, I saw some strange activity in human participants' session events for this session and will start a thread in Slack to follow up, as that is more of an underlying API question.
@lazeratops Got it. I did extend the expiry of the room during the call so it could have messed things up for the bot.
Feedback