token-budget-advisor

To control AI response length, a pre-answer depth selection intercepts your prompt to estimate input tokens and present four depth options with token estimates before the model answers. This allows you to choose the exact response detail level upfront.

Input tokens are estimated heuristically by differentiating between prose-dominant and code-dominant text. This heuristic estimation approximates the input size to calculate token windows and present accurate depth choices for your specific prompt.

Yes, you can set a persistent response depth level across multiple prompts using session-aware depth persistence. Once you select a level, the system honors it silently for subsequent prompts, and you can use shortcuts like tldr or 50% depth for quick changes.

The best way to get a concise checklist instead of an exhaustive guide is to request token-based depth choices before the refactor plan is generated. This presents options ranging from brief summaries to exhaustive details, letting you select the exact output length.

Yes, token budget estimation works without external dependencies because it uses internal heuristic logic. It estimates input size and calculates token windows by classifying prompt complexity directly, functioning as a standalone pre-answer depth selection mechanism.