Price a code review bot that runs a reasoning model

A reasoning model writes better reviews and bills for every token it thinks. Price both models, and the thinking.

Your task

Declare the review bot's pipeline, then write the function that prices it per month.

• declare PIPELINE: one entry per step, with an eval and a fallback on every model step
• write monthly_cost(pipeline, prices, items_per_day, days) and get the arithmetic right
• prices are dollars per million tokens; the cache discount applies to INPUT tokens only
• reasoning_tokens bill at the OUTPUT rate on every run and are never cached; a step without the field has none
• runs_per_item is fan-out: a step that runs on 30 percent of diffs is 0.3

design.md and cost_math.md are read by the critique, not by the suite.