Files
dotfiles/bin/pi-voice-query
2026-06-18 15:12:10 +02:00

96 lines
2.8 KiB
Bash
Executable File

#!/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