Handle Local/Provider LiteLLM

[nanni00]

Handle LiteLLM local VS remote parameters (like n_ctx, keep_alive, ...) from config yaml file to llm client. Invalid parameters cause errors when launching a completion.

llm_client.py:274

        completion_args = {
            "model": self.model,
            "messages": messages,
            "temperature": temp,
            "keep_alive": self.keep_alive,
            "num_ctx": self.num_ctx,
            **kwargs,
        }

Here, should be considered as some generic kwargs? Parsed considering the model type string?