When providing a custom LLM provider to the chat, we want to be able to view ussage metrics and identify conversations consuming high amount of tokens/credits