feat(voice): détection deux niveaux + service systemd

- VAD mode : Whisper small (~0.3s) pour détecter "ok hermes", large-v3 uniquement pour transcrire la commande
- PTT mode : large-v3 direct (inchangé)
- install-service.sh : installe hermes-voice comme service utilisateur systemd (auto-start à la session)
- doc mise à jour

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

tools/hermes-voice/install-service.sh

@@ -0,0 +1,53 @@
+#!/usr/bin/env bash
+# Installe hermes-voice comme service utilisateur systemd
+# Lance automatiquement au démarrage de session, écoute "Ok Hermes, ..."
+
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+VENV_PYTHON="$SCRIPT_DIR/.venv/bin/python"
+SERVICE_DIR="$HOME/.config/systemd/user"
+SERVICE_FILE="$SERVICE_DIR/hermes-voice.service"
+
+# Vérifications
+if [ ! -f "$VENV_PYTHON" ]; then
+  echo "❌ venv introuvable — lancez d'abord :"
+  echo "   python3 -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt"
+  exit 1
+fi
+
+mkdir -p "$SERVICE_DIR"
+
+cat > "$SERVICE_FILE" << EOF
+[Unit]
+Description=Hermes Voice Client
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+ExecStart=$VENV_PYTHON $SCRIPT_DIR/hermes-voice.py
+WorkingDirectory=$SCRIPT_DIR
+Restart=on-failure
+RestartSec=10s
+Environment=PYTHONUNBUFFERED=1
+
+StandardOutput=journal
+StandardError=journal
+SyslogIdentifier=hermes-voice
+
+[Install]
+WantedBy=default.target
+EOF
+
+systemctl --user daemon-reload
+systemctl --user enable hermes-voice
+systemctl --user start hermes-voice
+
+echo ""
+echo "✓ Service installé et démarré"
+echo ""
+echo "Commandes utiles :"
+echo "  systemctl --user status hermes-voice"
+echo "  journalctl --user -u hermes-voice -f"
+echo "  systemctl --user stop hermes-voice"
+echo "  systemctl --user restart hermes-voice"