Sometimes, even when the status is 'success', the responses from the GPT-3.5-Turbo model are empty. Additionally, I'm encountering an error message stating 'Response body too large'. Has anyone else experienced this, and if so, could you share any insights or solutions to resolve this issue?
Hi ! The message "Response body too large" is generated when the size of the data we're attempting to send to our frontend exceeds the capacity limits set by our current serverless infrastructure. This message doesn't reflect the actual request or response exchanged with the language model; it's more about the constraints we have on the data we can transmit in one go.
We will be addressing these limitations very soon as we are switching to dedicated servers!
Hi, so I took a look at one of the requests where the body is missing. This is due to the response body size. Since you're streaming, the stream data is quite large. We only support the large bodies for enterprise users at this time, however we're in the middle of adding support for all users.