How to avoid 'conversational' responses when using API?

By conversational responses I am referring to those along the lines of ‘Certainly! Here is what you asked for’. Or ‘Apologies for the oversight.’ Or ‘Apologies, but as an AI language model …’

In the last case, a simple “No data/answer available” would suffice.

Is there an instruction, special flag or something for obtaining answers without these additions? Should I append special instructions to each question that I am asking. I am including answers in automatically created reports, it would be nice if I wouldn’t need to go through each report and just delete those sentences.

Thanks in advance!

I’m pretty new to prompting but I was able to bypass the conversational responses by adding a line in my prompt for GPT to only return to the point/ concise answers. Hope that helps!

write instructions using SYSTEM message, to guide the model to generate responses that meet your specific requirements

such as
Provide concise answers without additional explanations or apologies.
Give me the information directly without any introductory sentences.
Exclude any extra wording and just provide the essential answer.

1 Like