137 lines
4.7 KiB
Python
137 lines
4.7 KiB
Python
"""
|
||
Smart Speaker - Main Application
|
||
Голосовой ассистент с wake word detection, STT, AI и TTS.
|
||
|
||
Flow:
|
||
1. Wait for wake word ("Alexandr")
|
||
2. Listen to user speech (STT)
|
||
3. Send query to AI (Perplexity)
|
||
4. Clean response from markdown
|
||
5. Speak response (TTS)
|
||
6. Loop back to step 1
|
||
"""
|
||
|
||
import signal
|
||
import sys
|
||
|
||
from wakeword import wait_for_wakeword, cleanup as cleanup_wakeword, check_wakeword_once
|
||
from stt import listen, cleanup as cleanup_stt, get_recognizer
|
||
from ai import ask_ai
|
||
from cleaner import clean_response
|
||
from tts import speak, initialize as init_tts
|
||
from sound_level import set_volume, parse_volume_text
|
||
|
||
|
||
def signal_handler(sig, frame):
|
||
"""Handle Ctrl+C gracefully."""
|
||
print("\n\n👋 Завершение работы...")
|
||
cleanup_wakeword()
|
||
cleanup_stt()
|
||
sys.exit(0)
|
||
|
||
|
||
def main():
|
||
"""Main application loop."""
|
||
print("=" * 50)
|
||
print("🔊 УМНАЯ КОЛОНКА")
|
||
print("=" * 50)
|
||
print("Скажите 'Alexandr' для активации")
|
||
print("Нажмите Ctrl+C для выхода")
|
||
print("=" * 50)
|
||
print()
|
||
|
||
# Setup signal handler for graceful exit
|
||
signal.signal(signal.SIGINT, signal_handler)
|
||
|
||
# Pre-initialize models (takes a few seconds)
|
||
print("⏳ Инициализация моделей...")
|
||
get_recognizer().initialize() # Initialize STT model first
|
||
init_tts() # Then initialize TTS model
|
||
print()
|
||
|
||
# Main loop
|
||
skip_wakeword = False
|
||
while True:
|
||
try:
|
||
# Step 1: Wait for wake word or Follow-up listen
|
||
if not skip_wakeword:
|
||
wait_for_wakeword()
|
||
# Standard listen after activation
|
||
user_text = listen(timeout_seconds=7.0)
|
||
else:
|
||
# Follow-up listen (wait 2.0s for start, then listen long)
|
||
print("👂 Слушаю продолжение диалога...")
|
||
user_text = listen(timeout_seconds=20.0, detection_timeout=2.0)
|
||
|
||
if not user_text:
|
||
# User didn't continue conversation, go back to sleep
|
||
skip_wakeword = False
|
||
continue
|
||
|
||
# Reset flag for now (will be set to True if we speak successfully)
|
||
skip_wakeword = False
|
||
|
||
# Step 2: Check if speech was recognized
|
||
if not user_text:
|
||
speak("Извините, я вас не расслышал. Попробуйте ещё раз.")
|
||
continue
|
||
|
||
# Check for volume command
|
||
if user_text.lower().startswith("громкость"):
|
||
try:
|
||
# Remove "громкость" prefix and strip whitespace
|
||
vol_str = user_text.lower().replace("громкость", "", 1).strip()
|
||
|
||
# Try to parse the number
|
||
level = parse_volume_text(vol_str)
|
||
|
||
if level is not None:
|
||
if set_volume(level):
|
||
speak(f"Громкость установлена на {level}")
|
||
else:
|
||
speak("Не удалось установить громкость.")
|
||
else:
|
||
speak(
|
||
"Я не понял число громкости. Скажите число от одного до десяти."
|
||
)
|
||
|
||
continue
|
||
except Exception as e:
|
||
print(f"❌ Ошибка громкости: {e}")
|
||
speak("Не удалось изменить громкость.")
|
||
continue
|
||
|
||
# Step 3: Send to AI
|
||
ai_response = ask_ai(user_text)
|
||
|
||
# Step 4: Clean response
|
||
clean_text = clean_response(ai_response)
|
||
|
||
# Step 5: Speak response (with wake word interrupt support)
|
||
completed = speak(clean_text, check_interrupt=check_wakeword_once)
|
||
|
||
# Enable follow-up mode for next iteration
|
||
skip_wakeword = True
|
||
|
||
# If interrupted by wake word, we still want to skip_wakeword (which is set above)
|
||
# but we can print a message
|
||
if not completed:
|
||
print("⏹️ Ответ прерван - слушаю следующий вопрос")
|
||
continue
|
||
|
||
print()
|
||
print("-" * 30)
|
||
print()
|
||
|
||
# Step 6: Loop continues with skip_wakeword=True
|
||
|
||
except KeyboardInterrupt:
|
||
signal_handler(None, None)
|
||
except Exception as e:
|
||
print(f"❌ Ошибка: {e}")
|
||
speak("Произошла ошибка. Попробуйте ещё раз.")
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|