13 Commits

Author SHA1 Message Date
nico
06d2bd5ebb Match lock screen date color to the clock
The date label used color8 (muted grey) while the clock used foreground,
making them visually inconsistent. Both now use foreground.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-11-06 00:08:13 +01:00
nico
6373ed38ec Add a palette-themed hyprlock lock screen
Themed through the same template pipeline as everything else: clock,
date, username and password-field colors come from theme/colors.toml,
background is a re-encoded, size-capped copy of the active wallpaper
(handles both an oversized original and a GIF saved under the .png-
named active wallpaper file). Bound to Mod+L, guarded against double-
invocation. Animations are disabled - this GPU/driver combo crashes
hyprlock's fade-in otherwise, regardless of what background is set.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-11-05 17:38:26 +01:00
nico
231a6a6c94 Include gtk_selection_color/opacity in the built-in Default preset
DEFAULT_PALETTE didn't list these keys, so loading "Default" after a
preset that did set them left the selection color stuck on whatever
it was before.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-10-11 01:35:28 +02:00
nico
5a1e4ca3ad Default GTK selection color to accent instead of fixed red
gtk_selection_color's comment already claimed it defaults to accent,
but the value was hardcoded to a leftover red instead.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-10-11 01:31:14 +02:00
nico
4547c27aa7 Move mouse widget standalone, add RGB control widget, align peripherals
The mouse battery pill lived in Bar.qml, duplicated per-screen (it
renders once per monitor) and invisible whenever taskbar mode hid the
stats row entirely. Pulled into its own MouseWidget.qml, same
standalone pattern as the theme widget - single instance, click opens
DPI/poll-rate/idle-time/low-battery-threshold controls. mouse_battery.py
folded into the new mouse.py (adds set-dpi-stage/set-poll-rate/
set-idle-time/set-low-battery-threshold on top of the old status-only
script).

New RGBWidget.qml + scripts/rgb.py: per-device mode/color/speed/
brightness for RAM/GPU/mainboard RGB via OpenRGB's SDK server
(autostarted headless in hyprland.lua), plus a one-click "sync all to
accent". Both RAM sticks are grouped into one card/one set of controls
since they're the same model - rgb.py's commands take a comma-separated
device-id list for exactly this. Found and worked around two real
issues along the way: this server build mishandles openrgb-python's
default protocol version (pinned to 3 instead), and set_color() doesn't
reliably force Direct mode first on its own - a device left on a
cycling effect can keep animating right through a color-set call
despite it reporting success, worst on the GPU's own i2c adapter.
Mitigated with an explicit mode switch + verify-and-retry.

Also: hyprland.lua was missing accel_profile entirely, so libinput's
default adaptive curve was always active - flat now, for the 1:1
response a gaming mouse is normally run with (found while chasing an
unrelated OpenRazer DPI reset making it suddenly noticeable).

Widget layout: all three corner widgets now align on a shared row
height instead of each guessing the previous one's width.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-10-04 02:12:09 +02:00
nico
0c4a1307a9 Fix hardcoded /home/nico paths breaking theming for other users
ThemeWidget.qml's themeDir and five paths in Bar.qml (theme dump,
stats, notifications, volume) were literal "/home/nico/..." strings.
Harmless for a single-user machine, but a second user with their own
clone of this repo would have their bar and theme widget silently
pointed at this user's files instead of their own - which they can't
even read, since $HOME is 700. Now resolved per-user via
Quickshell.env("HOME") / Quickshell.shellDir.

qt6ct/qt6ct.conf (tracked in this repo) had the same problem in its
color_scheme_path, baked in from whoever last edited it rather than
generated - now a proper render.py template using a new {home}
placeholder (not a palette color, always available) instead of one of
the existing color substitutions, since this is the first template
that needs an absolute path rather than a color value.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-30 02:28:07 +02:00
nico
3ae03bed4f Add a custom palette-driven SDDM login theme
New sddm/aether-rice/: matches the rest of the rice (Bar/mako palette,
JetBrainsMono, popup-glass login card), synced live into
/usr/share/sddm/themes/aether-rice by render.py's sync_sddm_theme() -
no sudo after the one-time setup documented in the README, since the
greeter can't read anything under ~/.config ($HOME is 700).

Custom Rectangle+MouseArea buttons instead of SddmComponents' Button,
which can get its background stuck on the wrong color (drives it
through named states, but also assigns it as a plain property - once
something sets that outside the state machine, idle has nothing
correct to revert to). Session/layout row is last in the card, not
above the buttons, because ComboBox/LayoutBox open their dropdown as a
plain child Rectangle that paints under whatever comes later in the
same Column - it was invisible behind the buttons above it.

Xsetup runs `xrandr --output <biggest> --primary` before the greeter
starts, so the login card lands on the right monitor - `<biggest>` is
picked dynamically by resolution, not hardcoded, since Xorg (esp. with
nvidia) often names outputs differently than Wayland/DRM does.

Known, unfixable-from-here limitation documented in the README: the
keyboard layout shows/applies as the wrong one until the first
keypress, on every SDDM theme including its own bundled ones - a
confirmed open upstream bug (sddm/sddm#1845), not something a theme's
QML can work around.

Also fixes wallpaper.py never syncing the login screen's background:
it doesn't go through render.py at all normally, so it now calls
render.sync_sddm_theme() itself.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-30 00:39:41 +02:00
nico
15f5da4509 Support animated GIF wallpapers via awww
swaybg only ever renders one frame, so an animated GIF (detected by its
magic bytes, not the .gif extension - the active wallpaper file is
always literally named wallpaper.png regardless of real format) now
goes through awww instead, which was already installed. Found that
awww img's --transition-type none - the obvious flag for an instant
switch - silently kills the render loop that also drives GIF frame
advancement; --transition-step 255 gives the same instant switch
without that side effect.

New `wallpaper.py start` picks the right backend for whatever's
currently active and replaces the hardcoded swaybg autostart line in
hyprland.lua, so a GIF wallpaper survives a restart correctly too.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-29 01:33:03 +02:00
nico
823277f535 Switch file manager to Thunar, fix GTK3 theming from the ground up
Dolphin's selection color is stuck on stock Breeze blue outside a full
Plasma session (unfixed upstream regression, see README) - dropped it
for Thunar. Getting Thunar themed exposed a chain of GTK3-specific
issues fixed along the way: !important is silently rejected by GTK3's
CSS parser (not just deprioritized), the widget and icon themes resolve
via completely different mechanisms under Hyprland (no XSETTINGS
daemon), and Thunar's ExoIconView grid reads its selection color from a
named @define-color instead of normal :selected matching. Folder icon
colors are now generated to match the accent exactly (theme/icon_theme.py,
no sudo) instead of snapping to a papirus-folders preset.

Also adds the kde-colorscheme template (the missing piece
KColorSchemeManager actually resolves ColorScheme= against) and new
gtk_selection_color/gtk_selection_opacity palette keys with widget
controls, following the existing bar/pill/popup color+opacity pattern.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-29 01:19:42 +02:00
nico
1ca9812608 Theme qt6ct too - that's what actually colors Dolphin, not kdeglobals
kdeglobals's [Colors:*] sections (added last commit) turned out not to be
what colors a plain Qt6 widget like Dolphin: hyprland.lua sets
QT_QPA_PLATFORMTHEME=qt6ct, which makes Qt load its whole QPalette from
qt6ct's own color scheme file instead of asking KDE Frameworks for one.
kdeglobals still matters for KDE/Plasma-integration bits, just not this.

New theme/templates/qt6ct/ renders that file (a serialized QPalette, one
color per Qt::ColorRole - mapped against qpalette.h, not guessed) to
~/.config/qt6ct/colors/aether-rice.conf, the path qt6ct.conf already pointed
at. A few roles (Light/Midlight/Dark/Mid/Shadow/AlternateBase/NoRole/
ToolTipBase/HighlightedText) have no equivalent in colors.toml and reuse the
same static shades already left unmapped in kdeglobals.

Confirmed by restarting Dolphin: an already-open window keeps its old colors,
a fresh one immediately reflects the live palette. README now says plainly
which of the two files actually drives Dolphin's colors, and that a restart
is needed either way.

Also committing the current rendered outputs (colors.toml has been edited
live via the widget/presets since the last commit).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-28 23:15:02 +02:00
nico
cc37bcfe18 Theme KDE/Qt apps (Dolphin, etc.) via kdeglobals
New theme/templates/kde/ renders the KDE color scheme embedded in kdeglobals
(Colors:View/Window/Button/Selection/Tooltip/Complementary + WM), themeing
every KDE/Qt app that reads it, not just Dolphin - kdeglobals was previously a
static, hand-maintained file (still bearing its ColorScheme=AetherRice name
from when the old aether app generated it once). The rendered output is
byte-identical to what was already there, confirming every mapped value was
right.

Needed render.py to grow a {key.rgb} placeholder format ("r,g,b" decimal, no
alpha - what kdeglobals uses, vs. the hex/rgba() formats every other template
needed). A couple of BackgroundAlternate shades don't correspond to any role
in colors.toml and are left as static literals rather than mapped to
something that wouldn't mean the same thing - noted in the template.

reload_apps() now also sends a best-effort dbus-send KGlobalSettings ping;
not every KDE app listens for it, so an already-open Dolphin may still need a
restart to fully pick up a change, same as most apps here.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-28 22:54:45 +02:00
nico
63a56f4a01 Theme presets, and make the bar itself actually re-theme (colors + opacity)
Presets (theme/presets.py): named snapshots of the whole palette, saved/loaded/
deleted from a new bookmark button on the theme widget. Stored outside the
repo in ~/.local/share/quickshell/theme-presets/ (JSON, per-machine data, same
precedent as the wallpaper history). A built-in Default entry can't be
overwritten or deleted.

Wiring that up exposed a real gap: the bar's own colors were hand-matched
hardcoded QML literals, not actually driven by theme/colors.toml, and opacity
wasn't themeable anywhere yet. Fixed both:
- colors.toml gains bar_color/pill_color/popup_color (paired with the existing
  *_opacity floats) plus generic opacity support in render.py (--set now
  accepts a 0-1 float, not just #rrggbb)
- Bar.qml now fetches the palette via render.py --dump and rebinds every
  matching hardcoded hex literal (accent, foreground, color1/2/3/5/6/8/12,
  the bar/pill/popup backgrounds, the logo tint and the bar's bottom border)
  to it; kitty's background_opacity is template-driven instead of a second
  hardcoded line fighting the themed one
- both the bar and the widget listen on their own Quickshell IPC target
  (render.py/presets.py ping both after every change) so edits reload with no
  restart, matching the "instant" promise the color editor already had
- the theme widget's color popup gained a slider + matching color swatch per
  surface (Bar/Pills/Popups/Terminal) below the existing role list, and its
  role labels now say what each color actually paints instead of "Color N"
  (verified against the templates and Bar.qml's literals, not guessed) with
  unused roles honestly marked as such
- the CPU pill's icon color is deliberately hardcoded again after this -
  it coincidentally shared color12's hex value with the help icon/detail
  text, so theming color12 was recoloring it too

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-28 22:45:09 +02:00
nico
3c69367660 Replace aether with a homegrown theme system + desktop theme widget
aether (the third-party GUI app) is gone - it didn't work reliably. Replaced
with theme/: colors.toml (palette), templates/<app>/ (Hyprland, Kitty, Rofi,
mako; unused Waybar template dropped), render.py (renders + reloads apps,
supports --set/--get/--dump), and wallpaper.py (archives + swaps the
wallpaper, deduped by hash, via a native zenity picker).

New ThemeWidget.qml: a small standalone widget pinned to the top-left corner
of the main monitor, independent of the bar and sitting on the Bottom wlr
layer (so normal windows can cover it). Two buttons - wallpaper gallery/picker,
and a color-role picker with a preset palette + hex input - call straight into
the scripts above.

Also includes an unrelated hyprland.lua fix found while testing the widget:
new kitty windows were coming up in genuine Hyprland fullscreen regardless of
workspace; a fullscreen_state window rule forces it off (plain
'fullscreen = false' turned out to be a no-op).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Dujsk9jYfxUVLtb5B8eC94
2026-09-28 01:53:04 +02:00