Applying the content filter

I see this requirement, unless there are exceptions: “Implement the Content Filter to prevent Unsafe (CF=2) completions.”

I’m a non-technical founder. Maybe I’m misunderstanding something…

At this link: OpenAI API it says, among other things, to apply the content filter, max tokens should be set to 1, and temperature to 0.

If max tokens are set to one, then the AI can only return 1 word, right? I need to return multiple sentences.

And my model needs a temperature above zero.

Does this mean I should apply for exceptions, and not apply the content filter?

Or am I misunderstanding something here?

Using 1 for max tokens doesn’t sound right

1 Like

Works as a charm for me. I followed the steps in the documentation, including max_tokens=1 for whatever reason, and I get a nice response object back :+1: