Settings · Praxis

AI-worded feedback

Rephrases feedback and session summaries. The checker decides what's right; the model only changes the wording. Off by default — nothing downloads or calls out until you turn it on.

Off

Word every resultNormally only wrong or undecided answers are sent to the model — a correct answer already has the right message, and wording it just adds latency. Turn this on to see the model on every path while authoring.

Provider

In browser (WebGPU)

Checking WebGPU…