Summary
I recorded this screencast to showcase the Rose Email Local AI Brain I built over two days, demonstrating how to set it up, index my Gmail, query it, and understand the associated costs. I start by running the auth command to complete the Gmail OAuth flow and store a token file, then use the generate command which indexes mail in phases—phase 1 completes in about 2.5 minutes, after which the brain is usable even though full indexing takes ~20 minutes. Each paid stage prints its cost, with the entire build costing roughly $1.40 and subsequent rebuilds being free unless data changes. I then query the brain with common and abstract prompts, showing how tool‑call clamping prevents endless loops and how the model returns citations from the last two years of email. Finally, I highlight cost per tool call (average ~6 cents) and the two‑year data window, confirming the system works as intended.