Skip to content

v1.3.0 — Token rates stay on screen

Latest

Choose a tag to compare

@rustytensor rustytensor released this 16 Nov 19:56

Highlights

  • Each model reply ends with one line in the message area, such as in 12,000 tok/s · out 98 tok/s. The line is saved in the session and is not sent to the model.
  • The footer keeps the last finished reply as last · in 12,000 tok/s · out 98 tok/s. It is not cleared when the next reply starts.
  • While a reply is running, the footer shows now only after that reply has numbers.
  • Alt+Shift+T switches the footer between that last reply and the session average.

Validation

  • npm test: 206 tests passed (167 + 39).

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