Gabriel Vidal da23d22519 feat(scripts): usage-calibration — fit /usage % against transcript spend ...

Lines the plan windows' official percentages (HA sensors from the homelab's
claude-usage-notifier) up against the account's archived token spend, priced
at list price, and fits dollars-per-1% for the 5h and week windows plus a
Fable multiplier. docs/usage-calibration.md records the method and the
2026-10 calibration (1% week ≈ $22, 1% 5h ≈ $5.6, Fable ×1.75, no separate
Fable bucket) and notes that pricing.py overstates the newest models.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

2026-10-12 22:40:14 +02:00

12 KiB

Executable File

Raw Permalink Blame History

View Raw