Indicating text type in data for open-ended text generation

I’ll start with a specific case. The OpenAI documentation describes how to use open-ended text generations to create haikus.

So far so good. But let’s say I want to to make it generate either haikus or and another kind of poetry (for example limericks). Obviously I’d need to include both haikus and limericks in the training data. But I’ll need to indicate which is a haiku and which is a limerick.

Should I include this information in the completion? e.g.
{“prompt”:“”, “completion”:" This is a haiku: END"}
{“prompt”:“”, “completion”:" This is a limerick: END"}

OR

Should I include this information in the prompt? e.g.

{“prompt”:“Generate a haiku.”, “completion”:” END"}
{“prompt”:“Generate a limerick.”, “completion”:” END"}

The documentation seems to indicate the former, for any kind of open-ended text generation.

Am I right?

I imagine this question applies to any use case when I want an AI that can generate different types of text (open-endedly) on demand.

Sorry if this is a noob question, I’m taking my first baby steps with this.

Put it in the prompt

Thanks for the reply.