zedarvates cost_estimator — tokens · model · money · time
from skills.cost_estimator import estimate, estimate_fix
estimate("code_review", 4000).human() # "2000 tok · Haiku/Flash (cloud) · $0.0001 · ~16s"
estimate_fix("dead_code", count=11) # local · free · ~34s
Reuses tiered_router's token+$ model and adds a throughput model for time.
Local tiers are free (slower/token); cloud tiers are billed (faster). Powers the
fix plan and any "what will this cost?" question. Exposed via [[llm_mcp]] as
estimate_cost. Related: [[fix]], [[auto_router]], [[metrics]].