I would kindly suggest to have the filter check as a parameter setting in the
openai.Completion.create function. When the filter check fails we have to retry and in such a way we spend for tokens that we don’t use. And in cases when we need to generate more than one completion, through the settings
best_of, the filter checking may reduce the advantage of the gain in speed that one gets through the fact that there is a setting such as
n. Maybe there’s a better way than the one I found here [OpenAI API](https://content filter OpenAI) and I just don’t know.
Thank you very much!
Building content filtering setting for incoming prompts and generated results for all the apis that take and generate text would be really helpful.
Right now I need 3 round trips for every call which really kills latency for my interactive app:
# Round trip 1
if ClassifyContent(user_prompt) < 2:
# Round trip 2
result = openai.Completion.create(... prompt=user_prompt ...)
# Round trip 3
if ClassifyContent(result) < 2:
# Show result to user
# show them a message saying was potentially unsafe
# show them a message saying their text was potentially unsafe
Maybe there is a better way to do this that I’m missing?