Feature Request: JSON Processing GPT

A common error I have encountered is when GPT cannot process its own JSON request because it gets too long and doesn’t terminate properly. This appears to the user as if the plugin is failing. OpenAI should add some error handling to this so that it gracefully deals with this error.

ApiSyntaxError: Could not parse API call kwargs as JSON: exception=Unterminated string starting at: line 3 column 18 (char 41)

2 Likes

Could also be a result of a plugin being poorly designed. I think if the plugin is returning JSON that’s too long it should truncate or abbreviate the response to ensure that it fits within a single message (and part of the response should indicate that this has been done).

What plugins do you see that come up in? I used to occasionally see it with complicated wolfram language queries, but not in a while I think they’ve fixed it.