domovoy_bot/services/voice_service.py

72 lines
2.6 KiB
Python

import os
import logging
from faster_whisper import WhisperModel
import sqlite3
from datetime import datetime
from pydub import AudioSegment
logger = logging.getLogger(__name__)
class VoiceService:
def __init__(self, db_path: str, model_size: str = "small"):
self.db_path = db_path
self.model_size = model_size
# CPU-only optimization: int8 is fast on standard CPUs
self.model = WhisperModel(model_size, device="cpu", compute_type="int8")
self._init_db()
def _init_db(self):
conn = sqlite3.connect(self.db_path)
cursor = conn.cursor()
cursor.execute('''
CREATE TABLE IF NOT EXISTS voice_recognition_logs (
id INTEGER PRIMARY KEY AUTOINCREMENT,
user_id INTEGER,
username TEXT,
message_type TEXT,
duration INTEGER,
text TEXT,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
)
''')
conn.commit()
conn.close()
def transcribe(self, file_path: str, user_id: int, username: str, msg_type: str) -> str:
try:
start_time = datetime.now()
# Convert to wav if needed (Telegram uses .oga/.ogg)
wav_path = file_path.rsplit('.', 1)[0] + ".wav"
audio = AudioSegment.from_file(file_path)
duration = int(len(audio) / 1000)
audio.export(wav_path, format="wav")
segments, info = self.model.transcribe(wav_path, beam_size=5)
text = " ".join([segment.text for segment in segments]).strip()
# Clean up temp files
if os.path.exists(wav_path): os.remove(wav_path)
# Log to DB
self._log_to_db(user_id, username, msg_type, duration, text)
return text
except Exception as e:
logger.error(f"Transcription error: {e}")
return f"[Ошибка распознавания]: {e}"
def _log_to_db(self, user_id, username, msg_type, duration, text):
try:
conn = sqlite3.connect(self.db_path)
cursor = conn.cursor()
cursor.execute('''
INSERT INTO voice_recognition_logs (user_id, username, message_type, duration, text)
VALUES (?, ?, ?, ?, ?)
''', (user_id, username, msg_type, duration, text))
conn.commit()
conn.close()
except Exception as e:
logger.error(f"DB Log error: {e}")
# Global instance will be initialized in main.py or handlers