Feature Request: Character Example Messages/Dialog

[ZeriLeandra]

Is your feature request related to a problem? Please describe.
One thing I've noticed in Lettuce is the characters I know the best don't have their 'voice' and 'vibe' necessarily. Not unless you give them a long starting message that focuses more on presenting their speech style which doesn't necessarily work with whatever scenario or template you're trying to setup.

Describe the solution you'd like
Most other chat front ends, when defining a character, allow you to create an "Example Dialog" section that the LLM understands is not part of current events or roleplay but trains the model on the character's mannerisms and speech in a way that is difficult to capture in a character definition. LLMs love to fall into stereotypical tropes so if you describe a character as analytical, it'll have the character start rambling data and falling into a 'geek' stereotype when if they're more like batman, they analyze everything and say nothing. In short; character personality and character presentation can be different things that some models need more help with.

Describe alternatives you've considered
Adding the example messages to the character definition, usually trying to put them into something to help the LLM realizes is separate using some kind of markup or coding tags like html. Not so great. Re-writing some character cards as less definition and more narrative to find a way to sneak in more of their nuance. That makes the Definition box harder to read and troubleshoot character behavior.

[MegalithOfficial]

I understand the request, but I do not think a separate Example Dialog field would make much meaningful difference in output quality on its own.

In practice, character behavior is driven much more by the main definition, prompt structure, recent context, and model choice than by a small block of example lines. In many cases, this would just become another text field with limited actual influence unless the entire prompt pipeline were redesigned around it.

So I am not against the idea in principle, but I see it as a low-priority authoring convenience rather than a feature that would noticeably improve responses. Right now I would rather focus on changes that have clearer impact on generation quality and overall usability.