main.py

import uvicorn
from fastapi import FastAPI, UploadFile, File, HTTPException
import whisper
import io
import soundfile as sf
import logging
import torch
import os

# Configure logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(name)s - %(message)s')
logger = logging.getLogger(__name__)

app = FastAPI()

# Global variable to store the model
whisper_model = None

@app.on_event("startup")
async def startup_event():
    global whisper_model
    logger.info("Starting up FastAPI application...")

    device = os.environ.get("WHISPER_DEVICE", "cuda" if torch.cuda.is_available() else "cpu")

    logger.info(f"Loading Whisper model (tiny) on {device}...")
    try:
        whisper_model = whisper.load_model("tiny", device=device)
        logger.info("Whisper model (tiny) loaded successfully.")
    except Exception as e:
        logger.error(f"Failed to load Whisper model: {e}", exc_info=True)
        raise RuntimeError(f"Could not load Whisper model: {e}. Check torch/whisper installation or device.")

@app.post("/transcribe_audio/")
async def transcribe_audio(audio_file: UploadFile = File(...)):
    if whisper_model is None:
        logger.error("Whisper model is not loaded.")
        raise HTTPException(status_code=503, detail="Whisper model is not loaded yet. Please try again in a moment.")

    logger.info(f"Received audio file: {audio_file.filename}, type: {audio_file.content_type}")

    try:
        audio_bytes = await audio_file.read()
        audio_io = io.BytesIO(audio_bytes)

        data, samplerate = sf.read(audio_io)

        if data.dtype != 'float32':
            data = data.astype('float32')

        if data.ndim > 1:
            data = data.mean(axis=1)

        result = whisper_model.transcribe(data)
        transcription = result["text"]
        logger.info(f"Transcription complete: '{transcription[:100]}...'")

        return {"transcription": transcription}
    except Exception as e:
        logger.error(f"Error during transcription: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail=f"Error processing audio: {e}. Please check server logs.")

if __name__ == "__main__":
    logger.info("Starting FastAPI Whisper server on port 1990...")
    uvicorn.run(app, host="0.0.0.0", port=1990)