LLM/AI-agent based use cases could be served by a special flag/option that uses a tokenizer or token-usage estimator for the character budget. in other would, allow the user to request a preview in N tokens instead of N characters