Native qs.Ui.NumberField (SpinBox) for lingerMs: mouse wheel, +/- buttons
and typing, 0-10000ms in 500ms steps, writing config.json like Mode and
Position so the display updates live. 0 selects the always-show behavior
from the previous commit.
The hide timer only starts (and only clears) when lingerMs > 0, so 0
means the last full combo stays visible until the next key replaces it
— the keyviz-video look, as an explicit opt-in rather than the previous
behavior where 0 hid instantly.
The three Position buttons (Bottom/Top/Center) overlapped the label in
the menu's width. Use qs.Ui.Dropdown with the full 3x3 grid — top /
middle / bottom by left / center / right — the display panel already
positions all of them.
The display now only updates on key-down (the combo grows as you press)
and when the last key is released (the empty payload starts the linger
with the last full combo). Intermediate releases no longer shrink the
display, so the order you let go of a shortcut doesn't matter: Ctrl+Shift+N
stays Ctrl+Shift+N whether Ctrl, Shift, or N goes up first, and the 1s
linger shows the full combo instead of a partial release state. Matches
the keybinding-tutorial use case better than keyviz, which shrinks
key-by-key on spaced releases.
The panel reads config.json from its plugin folder (seeded with defaults
on first run, hot-reloaded on save): mode (all | bindings — the bindings
mode only shows combos with a modifier, ideal for keybinding tutorials),
position (keyviz-style alignment, bottom-center by default) and margin,
plus the existing lingerMs. Replaces the hardcoded bottom anchors with
x/y computed from position and margin.
Previously the chips cleared the instant the keys were released, so the
120ms hide delay was invisible. Keep the last combo rendered for the
whole linger window and only clear it when the timer fires, so a quick
tap stays readable (keyviz keeps keys for lingerDurationMs, 5s by
default). Default 1000ms, tunable via lingerMs.
felixzsh