This covers the case when a the response from the LLM already contains the total token count, thus it doesn't need to be calculated from the prompt or the completion again.
Especially useful when the underlying embeddings model is not available for correct calculation, while the response already has the counts.
This covers the case when a the response from the LLM already contains the total token count, thus it doesn't need to be calculated from the prompt or the completion again.
Especially useful when the underlying embeddings model is not available for correct calculation, while the response already has the counts.