Implementation doubt

[lmbelo]

**** | DUMB QUESTION WARNING | ****

I’ve got a dumb question, but I’m curious: why doesn’t this library use LlamaCpp’s standard chat templates (llama_chat_apply_template) and instead go for a completely new implementation? I see it adds a lot of extra features, like function calling, but aren’t most (or at least a lot) of the chat formats already covered in LlamaCpp? Same goes for grammar and speculative stuff, maybe?

[sriramsowmithri9807]

Great question!! While LlamaCpp’s llama_chat_apply_template covers a lot of basics, this library offers more flexibility with features like function calling, grammar constraints, and speculative sampling things not fully addressed by the standard templates. Think of it as a multi-tool for more complex use cases.

Cheers,
sriram.