Add voice control

.config/sway/config

@@ -70,6 +70,7 @@ exec swayidle -w \
 
 input * {
     xkb_layout pl
+    xkb_options caps:none
     repeat_rate 80
     repeat_delay 300
 }
@@ -86,14 +87,6 @@ input "1267:12734:ELAN06FA:00_04F3:31BE_Touchpad" {
     scroll_factor 0.8
 }
 
-# Export important Wayland/Sway session variables to D-Bus and the systemd
-# user manager. This lets apps and user services launched outside the shell
-# know which Wayland display to use and that the current desktop is Sway.
-# Helps with portals, screen sharing, notifications, and D-Bus activated apps.
-exec dbus-update-activation-environment --systemd WAYLAND_DISPLAY XDG_CURRENT_DESKTOP=sway
-exec systemctl --user import-environment WAYLAND_DISPLAY XDG_CURRENT_DESKTOP
-
-
 
 ### Key bindings
 #
@@ -210,6 +203,9 @@ exec systemctl --user import-environment WAYLAND_DISPLAY XDG_CURRENT_DESKTOP
     # Show the next scratchpad window or hide the focused scratchpad window.
     # If there are multiple scratchpad windows, this command cycles through them.
     bindsym $mod+minus scratchpad show
+
+    # Dedicated scratchpad terminal for voice queries to pi.
+    for_window [app_id="pi-voice"] border pixel 2, move scratchpad
 #
 # Resizing containers:
 #
@@ -232,7 +228,9 @@ mode "resize" {
     # Return to default mode
     bindsym Return mode "default"
     bindsym Escape mode "default"
+    bindsym $mod+r mode "default"
 }
+
 bindsym $mod+r mode "resize"
 #
 # Utilities:
@@ -249,6 +247,16 @@ bindsym $mod+r mode "resize"
     bindsym Print exec sh -c 'mkdir -p ~/screenshots; f=~/screenshots/$(date +%F-%H%M%S).png; grim "$f"; wl-copy < "$f"'
     bindsym Shift+Print exec sh -c 'mkdir -p ~/screenshots; f=~/screenshots/$(date +%F-%H%M%S).png; grim -g "$(slurp)" "$f"; wl-copy < "$f"'
 
+    # Push-to-talk: hold Caps Lock to record, release to transcribe.
+    # xkb_options caps:none above disables the normal Caps Lock toggle, so bind by keycode.
+    bindcode 66 exec sh -c 'voxtype record start; pkill -USR1 -f "$HOME/.config/sway/status.sh"'
+    bindcode --release 66 exec sh -c 'voxtype record stop; pkill -USR1 -f "$HOME/.config/sway/status.sh"'
+
+    # Voice query to pi in a dedicated tmux session/scratchpad terminal.
+    # Hold $mod+Shift+x, speak, release; transcription is submitted to pi.
+    bindsym $mod+Shift+x exec ~/bin/pi-voice-query start
+    bindsym --release $mod+Shift+x exec ~/bin/pi-voice-query stop
+
 
     # Pick an entry from clipboard history and copy it back to the clipboard
     bindsym $mod+c exec sh -c 'cliphist list | wmenu -l 10 | cliphist decode | wl-copy'

.config/sway/status.sh

@@ -7,14 +7,21 @@ print_status() {
 
     brightness=$(brightnessctl -m 2>/dev/null | awk -F, '{print $4}')
 
-    printf 'VOL %s | BRI %s | %s\n' "${volume:-n/a}" "${brightness:-n/a}" "$(date +'%Y-%m-%d %H:%M:%S')"
+    voxtype_state=$(voxtype status 2>/dev/null)
+    case "$voxtype_state" in
+        recording) voxtype=' | VOX REC' ;;
+        transcribing) voxtype=' | VOX ...' ;;
+        *) voxtype='' ;;
+    esac
+
+    printf 'VOL %s | BRI %s%s | %s\n' "${volume:-n/a}" "${brightness:-n/a}" "$voxtype" "$(date +'%Y-%m-%d %H:%M:%S')"
 }
 
 trap print_status USR1
 
 print_status
 while :; do
-    sleep 60 &
+    sleep 1 &
     wait $!
     print_status
 done

install_the_dotfiles.sh

@@ -82,3 +82,24 @@ ssh-keygen -t ed25519 -f ~/.ssh/id_ed25519 -N ""
 # Extra
 ## Clipboard history
 # sudo apt install cliphist
+
+# Install
+curl -LO https://github.com/peteonrails/voxtype/releases/download/v0.6.0/voxtype_0.6.0-1_amd64.deb
+sudo dpkg -i voxtype_0.6.0-1_amd64.deb
+rm voxtype_0.6.0-1_amd64.deb
+sudo apt install wtype wl-clipboard libnotify-bin pipewire-alsa playerctl
+
+sudo usermod -aG input $USER
+
+read -r -p "Enable GPU support? [y/N] " answer
+case "$answer" in
+    [yY]|[yY][eE][sS])
+        sudo voxtype setup gpu --enable
+        ;;
+esac
+
+
+# Interactive model selection and systemd setup
+voxtype setup
+voxtype setup systemd
+voxtype setup model