Why kalosm_language::chat::Chat consumes LLM?

[InAnYan]

What if I want to have 2 chats? I have to load 2 LLMs?

In other LLM crates/(libraries from other programming languages), typically Chat is just a structure that holds messages and resembles and API request. It exists separately from LLM. And then LLM will consume that request and do what it needs to do. IMHO this is a better approach, as Chat doesn't depend on LLM and it's easy to modularize libraries and test them.

[ealmloff]

The Llama struct is a reference counted pointer to the language model. It is cheaply clonable so you can use it in many different places

[InAnYan]

Thanks for the clarification