Is it by design that Prompt Section's renderAsText and renderAsMessage will return result with different length for same content?

[blackchoey]

Take following conversation history as example:

[
    { role: 'user', content: 'Hello' },
    { role: 'assistant', content: 'Hi! How can I help you?' },
    { role: 'user', content: "I'd like to book a flight" },
    { role: 'assistant', content: 'Sure, where would you like to go?' }
]

The renderAsText function will return result with token length 36, but renderAsMessage will return result with token length 25.
Seems the major difference is renderAsMessage does not count the length of prefixes (like "user: ", etc.). But since the message will be rendered as text finally, items from renderAsMessage may be truncated again if the length is approaching the limit. I'm not sure whether this will cause possible confusions.

[corinagum]

This seems like a bug to me. Ideally having notes to explain differences like this in our documentation would help prevent confusion in the future.

[corinagum]

@Stevenic Do you have any input?