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

.profile

@@ -33,6 +33,6 @@ fi
 
 if [ -z "$WAYLAND_DISPLAY" ] && [ "$XDG_VTNR" = 1 ]; then
     export WLR_NO_HARDWARE_CURSORS=1
-    exec tmux new-session -A -s debug sway --unsupported-gpu
+    exec sway --unsupported-gpu
 fi
 

bin/command-list

@@ -168,11 +168,15 @@ if result.returncode != 0:
     exit(result.returncode)
 
 choice = result.stdout.strip()
-val = vals[choice]
-
-if val["kind"] == "bin":
-    subprocess.run([val["path"]])
-elif val["kind"] == "app":
-    subprocess.run(["gtk-launch", val["name"]])
-elif val["kind"] == "url":
-    subprocess.run(["xdg-open", val["url"]])
+if choice[0] == '@':
+    choice = choice[1:]
+val = vals.get(choice)
+if val is not None:
+    if val["kind"] == "bin":
+        subprocess.run([val["path"]])
+    elif val["kind"] == "app":
+        subprocess.run(["gtk-launch", val["name"]])
+    elif val["kind"] == "url":
+        subprocess.run(["xdg-open", val["url"]])
+else:
+    subprocess.run(["xdg-open", "https://www.google.com/search?q=" + choice])

bin/pi-voice-query

@@ -0,0 +1,95 @@
+#!/bin/sh
+set -eu
+
+SESSION="pi-voice"
+APP_ID="pi-voice"
+TITLE="Pi Voice"
+OUT="${XDG_RUNTIME_DIR:-/tmp}/pi-voice-query.txt"
+STATUS_SCRIPT="$HOME/.config/sway/status.sh"
+
+refresh_bar() {
+    pkill -USR1 -f "$STATUS_SCRIPT" 2>/dev/null || true
+}
+
+position_pi_window() {
+    # Size scratchpad window to 60% of the focused output using absolute pixels.
+    # Doing this here is more reliable than a for_window percentage resize.
+    set -- $(swaymsg -t get_outputs 2>/dev/null | python3 -c '
+import json, sys
+outs = json.load(sys.stdin)
+out = next((o for o in outs if o.get("focused")), None) or next((o for o in outs if o.get("active")), None)
+rect = (out or {}).get("rect", {})
+w = int(rect.get("width", 1200) * 0.6)
+h = int(rect.get("height", 800) * 0.6)
+print(w, h)
+')
+    w=${1:-1200}
+    h=${2:-800}
+    swaymsg '[app_id="'"$APP_ID"'"] border pixel 2, resize set width '"$w"' px height '"$h"' px, move position center' >/dev/null 2>&1 || true
+}
+
+ensure_pi_terminal() {
+    if ! tmux has-session -t "$SESSION" 2>/dev/null; then
+        tmux new-session -d -s "$SESSION" 'pi'
+    fi
+
+    if command -v swaymsg >/dev/null 2>&1; then
+        if ! swaymsg -t get_tree 2>/dev/null | grep -q '"app_id": "'"$APP_ID"'"'; then
+            foot --app-id="$APP_ID" --title="$TITLE" \
+                --override=main.resize-by-cells=no \
+                --override=colors.alpha=1.0 \
+                --override=main.pad=0x0 \
+                tmux attach-session -t "$SESSION" >/dev/null 2>&1 &
+            # Give the window a moment to appear before trying to show/focus it.
+            sleep 0.5
+        fi
+        swaymsg '[app_id="'"$APP_ID"'"] scratchpad show, focus' >/dev/null 2>&1 || true
+        position_pi_window
+    fi
+}
+
+send_to_pi() {
+    text=$1
+    [ -n "$text" ] || exit 0
+
+    ensure_pi_terminal
+
+    # Paste literally, then press Enter to submit to pi.
+    printf '%s' "$text" | tmux load-buffer -
+    tmux paste-buffer -t "$SESSION"
+    tmux send-keys -t "$SESSION" Enter
+}
+
+case "${1:-}" in
+    start)
+        ensure_pi_terminal
+        : > "$OUT"
+        voxtype record start --file="$OUT"
+        refresh_bar
+        ;;
+    stop)
+        voxtype record stop
+        refresh_bar
+
+        # Wait for transcription to land in OUT. Stop early once voxtype is idle
+        # and the file has text, otherwise allow slower large-model runs.
+        i=0
+        while [ "$i" -lt 180 ]; do
+            if [ -s "$OUT" ]; then
+                break
+            fi
+            state=$(voxtype status 2>/dev/null || true)
+            [ "$state" = idle ] && [ "$i" -gt 2 ] && break
+            sleep 0.5
+            i=$((i + 1))
+        done
+        refresh_bar
+
+        text=$(cat "$OUT" 2>/dev/null || true)
+        send_to_pi "$text"
+        ;;
+    *)
+        echo "usage: $0 start|stop" >&2
+        exit 2
+        ;;
+esac

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