Randomly occurring: The app encountered an error processing your request.

[ID925 (IainD925)]

Error type: <class 'openai.BadRequestError'>
Happening in both Chat & Ask. I ask the same bunch of questions several times and randomly get this error on a previously asked question. Refreshing doesn't fix but appears to resolve itself after a period of time.

Logs don't show anything of note.

Anyone else seeing this behaviour?

[ID925 (IainD925)]

Found the answer in Application Insights - Errors.

This model's maximum context length is 4096 tokens. However, your messages resulted in 14821 tokens. Please reduce the length of the messages.

I increased the chunking size and when you ask for multiple sources, the response is additive. Reduced to 1 source and issue resolved.

[Pamela Fox (pamelafox)]

Interesting, we have logic that tries to truncate the old conversation to fit in the message history, so it should have just truncated to 4096. However, it will fail if even a single question's sources exceed the token window. I made a lil video showing that message truncation logic- https://youtu.be/GIuj_w2HlT0
Glad you figured out a workaround!