Chat Completion with multiple user messages: how to distinguish between them in the response?

[tight breach]

Currently I'm working on a project using Chat Completion and I supply two ChatMessages in my request, the first one asking to generate a certain text and the second one asking to summarize that text. However, both the generated text and the generated summary seem to be combined into one response. How can I separate between these? I don't want to do two separate Chat Completion requests as this will increase token usage by a lot.

[elfin steppe]

This will not increase token usage if it returns the same data either way. However, to answer your question, you'll have to find a way to delineate both responses and then split them using that delineation.

[tight breach]

Well to get a summary of the generated text, I have to pass that generated text as an input for the second request which increases token usage right?

Since there's no "conversation memory" or anything

[elfin steppe]

I see what you're saying now.

You have it summarize the text as it writes.

So yeah, you'd need to create a delineation and instruct it to use that delineation in the system message.

[tight breach]

Alright, that's the 'workaround' I'm currently working on but I guess there's no better way then unfortunately.