server/AudioTranscriber.py

import io
import threading
from multiprocessing import Queue
from queue import Empty
from faster_whisper import WhisperModel

import logging
import sys

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
    handlers=[logging.StreamHandler(sys.stdout)],
)

# Get a logger for your app
logger = logging.getLogger(__name__)


class AudioTranscriber(threading.Thread):
    def __init__(self, audio_queue: "Queue[io.BytesIO]", text_queue: "Queue[str]"):
        super().__init__()
        self.audio_queue = audio_queue
        self.action_queue = text_queue
        self.daemon = True  # Thread will exit when main program exits

        self.transcriber = WhisperModel(
            "large",
            device="cuda",
            compute_type="int8",
        )

    def run(self):
        while True:
            try:
                # Wait for 1 second before timing out and checking again
                audio_chunk = self.audio_queue.get(timeout=1)

                # Process the audio chunk using the faster-whisper implementation
                segments, info = self.transcriber.transcribe(audio_chunk, language="fr")

                # Put the transcription results in the output queue
                for segment in segments:
                    self.action_queue.put(segment.text)
                    # Still print for debugging
                    logger.info(
                        f"[{segment.start:.2f}s -> {segment.end:.2f}s] {segment.text}"
                    )

            except Empty:
                continue  # If queue is empty, continue waiting
            except Exception as e:
                logger.error(f"Error processing audio chunk: {e}")