import librosa
from transformers import pipeline
import logging

# Set up logging
logging.basicConfig(level=logging.DEBUG)

ASR_SAMPLING_RATE = 16_000

try:
    pipe = pipeline("automatic-speech-recognition", model="facebook/mms-1b-all")
    logging.info("ASR pipeline loaded successfully.")
except Exception as e:
    logging.error(f"Error loading ASR pipeline: {e}")

def transcribe(audio):
    try:
        if audio is None:
            logging.error("No audio file provided")
            return "ERROR: You have to either use the microphone or upload an audio file"
        
        logging.info(f"Loading audio file: {audio}")

        # Try loading the audio file with librosa
        try:
            audio_samples, _ = librosa.load(audio, sr=ASR_SAMPLING_RATE, mono=True)
        except FileNotFoundError:
            logging.error("Audio file not found")
            return "ERROR: Audio file not found"
        except Exception as e:
            logging.error(f"Error loading audio file with librosa: {e}")
            return f"ERROR: Unable to load audio file - {e}"
        
        # Process the audio with the pipeline
        try:
            transcription = pipe(audio_samples)["text"]
        except Exception as e:
            logging.error(f"Error during transcription with pipeline: {e}")
            return f"ERROR: Transcription failed - {e}"
        
        logging.info("Transcription completed successfully.")
        return transcription
    except Exception as e:
        logging.error(f"Error during transcription: {e}")
        return "ERROR"