I have been using the API for well over a year now without any issues and I am now consistently getting responses of “Refusal” null.
My prompts have not changed, I am simply scanning documents and providing feedback as to the quality of the document. I am thinking the new safety feature has been turned up a few degrees to warm and is causing false positives. Has anyone else experienced this?
It has gone from bad to worse. Now every response is the “refusal” null. This is not good folks!!!
1 Like
Any News? I just updated my filemaker scripts, from gpt-3.5 to gpt-4o-mini, and I get after every answer, the, ‘refusal: null’ string
EDIT: Got the solution: The json-file contains in the new model 4o-mini one line more, the refusal-field!
Solution: Just cut of the parsing at the end of the content field! This works
1 Like
This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.