innolux/api/whisper.py

# import os, sys
# from typing import List, Any
from fastapi import Request, APIRouter, UploadFile, File
from fastapi.responses import FileResponse, PlainTextResponse
# from fastapi import FastAPI, HTTPException, status
# from fastapi.middleware.cors import CORSMiddleware
# import uvicorn
# from fastapi.exceptions import HTTPException
# from fastapi.encoders import jsonable_encoder
# from urllib.parse import urlparse, urljoin
from pathlib import Path
from api.openai_scripts_tai_gi.main_script import main
from api.openai_scripts_chinese.main_script import main as main2
# from api.openai_scripts_new.main_script import main as main3
from datetime import datetime
import random
import string

router = APIRouter()

# router = FastAPI()
# router.add_middleware(
#     CORSMiddleware,
#     allow_origins=["*"],
#     allow_credentials=True,
#     allow_methods=["*"],
#     allow_headers=["*"],
# )


@router.post('/tai_gi')
async def whisper_auto(file: UploadFile = File()):
    if file == None:
        return {'message': '請上傳檔案'}
    extension = file.filename.split(".")[-1] 
    if extension not in ("mp3", "wav", "webm"):
        return PlainTextResponse("Audio must be mp3, wav or webm format!", 400)
    filename = Path(__file__).parent.parent/'speech_audio'/datetime.now().strftime(f"%Y%m%d%H%M%S_{''.join(random.sample(string.ascii_lowercase, 3))}.{extension}")
    with open(filename, 'wb') as f:
        f.write(await file.read())
    with open(filename, 'rb') as f:
        raw_transcript, corrected_transcript = main(f)
    # if raw_transcript and corrected_transcript:
        # os.remove(filename)
        # return {'message': corrected_transcript}
    # else:
        # os.remove(filename)
        # return {"message": "Audio processing failed."}
    if raw_transcript and corrected_transcript:
        # os.remove(save_path)
        # return {'message': {"Raw transcript": raw_transcript, "Corrected transcript": corrected_transcript}}
        return {'message': corrected_transcript}
    else:
        # os.remove(save_path)
        return {"message": "Audio processing failed."}
        
        
@router.post('/chinese')
async def whisper_auto(file: UploadFile = File()):
    if file == None:
        return {'message': '請上傳檔案'}
    extension = file.filename.split(".")[-1] 
    if extension not in ("mp3", "wav", "webm"):
        return PlainTextResponse("Audio must be mp3, wav or webm format!", 400)
    filename = Path(__file__).parent.parent/'speech_audio'/datetime.now().strftime(f"%Y%m%d%H%M%S_{''.join(random.sample(string.ascii_lowercase, 3))}.{extension}")
    with open(filename, 'wb') as f:
        f.write(await file.read())
    with open(filename, 'rb') as f:
        raw_transcript, corrected_transcript = main2(f)
    if raw_transcript and corrected_transcript:
        # return {'message': {"Raw transcript": raw_transcript, "Corrected transcript": corrected_transcript}}
        return {'message': corrected_transcript}
    else:
        return {"message": "Audio processing failed."}
        

# import numpy as np
# from transformers import pipeline
# import gradio as gr

# @router.post('/tai_gi_new')
# async def whisper_auto(file: UploadFile = File()):
#     if file == None:
#         return {'message': '請上傳檔案'}
#     extension = file.filename.split(".")[-1] 
#     if extension not in ("mp3", "wav", "webm"):
#         return PlainTextResponse("Audio must be mp3, wav or webm format!", 400)
#     filename = Path(__file__).parent/'speech_audio'/datetime.now().strftime(f"%Y%m%d%H%M%S_{''.join(random.sample(string.ascii_lowercase, 3))}.{extension}")
#     with open(filename, 'wb') as f:
#         f.write(await file.read())
#     # with open(filename, 'rb') as f:
#         # raw_transcript, corrected_transcript = main3(f)
#     # audio_data, sample_rate = librosa.load(filename, sr=None)
#     # 讀取音頻文件
#     audio_input = gr.Audio(type="filepath")
#     raw_transcript, corrected_transcript = main3(audio_input(filename))
#     # if raw_transcript and corrected_transcript:
#         # os.remove(filename)
#         # return {'message': corrected_transcript}
#     # else:
#         # os.remove(filename)
#         # return {"message": "Audio processing failed."}
#     if raw_transcript and corrected_transcript:
#         # os.remove(save_path)
#         return {'message': {"Raw transcript": raw_transcript, "Corrected transcript": corrected_transcript}}
#     else:
#         # os.remove(save_path)
#         return {"message": "Audio processing failed."}

        
# if __name__ == "__main__":
#     uvicorn.run("whisper:router", reload=False, port=8086, host='cmm.ai', ssl_keyfile="/etc/letsencrypt/live/cmm.ai/privkey.pem", ssl_certfile="/etc/letsencrypt/live/cmm.ai/fullchain.pem")