SpeechKit v0.57.0
·
39 commits
to main
since this release
Highlights
- Talk to the assistant on your phone: the Android app now holds a real
spoken conversation — press to talk, hear the answer, watch the transcript
build as you speak — instead of promising one for later. - A conversation without leaving the text field: the SpeechKit keyboard
can open the assistant in place, so you can think something through and
then go back to typing. Dictation still writes what you say into the
field; the conversation stays out of it on purpose. - One assistant, one face: the animation on Android is now the same one
the Windows overlay and the web app draw, from a single shared source, so
a change to the visual language reaches every screen at once. - The keyboard setup finds the right keyboard: enabling and selecting
SpeechKit as your keyboard is now detected correctly, so onboarding stops
claiming a step is done before it is.
Added
- android: add the realtime Voice Agent client the platform was missing (ecd04c4)
- android: fold Voice Agent sessions into one render state (8115caa)
- android: hold a conversation from inside the keyboard (afcd2e0)
- android: let the assistant speak one visual language throughout (bce2ed8)
- android: make the Voice Agent reachable in the app (359ee48)
- android: publish the shared orb so other apps can stop copying it (60496d4)
Fixed
- android: detect our own keyboard instead of a stranger's (288723a)
- ci: let dependency updates past the standards gate too (f7000f1)
- delivery: unblock security updates stuck behind an unavailable token (6722ad3)
- release: actually merge the installer mirror instead of warning about it (9db0160)
Full Changelog: v0.56.0...v0.57.0