v1.2.0 — Clearer token rates

Highlights

• The footer shows this reply as now · in 12,000 tok/s · out 98 tok/s.
• in is prompt tokens per second until the first generated token. Cached prompt tokens are not counted.
• out is generated tokens per second after that, including reasoning.
• Alt+Shift+T switches to the session average: session · in 8,400 tok/s · out 71 tok/s. That is all measured tokens divided by all measured time, not a mean of each reply.

Validation

• npm run build, then npm test: 204 tests passed (165 + 39).

npm publication remains separate and maintainer-owned. This GitHub release does not publish the npm package.