app.py

import gradio as gr
import requests
import os
import time

# Retrieve ASR API URL and Authorization Token from environment variables
ASR_API_URL = os.getenv('ASR_API_URL')
AUTH_TOKEN = os.getenv('AUTH_TOKEN')

def transcribe_audio(file_path):
    if not ASR_API_URL or not AUTH_TOKEN:
        return "Error: Missing ASR_API_URL or AUTH_TOKEN."

    # Prepare headers and data
    headers = {
        'accept': 'application/json',
        'Authorization': f'Bearer {AUTH_TOKEN}',
    }
    files = {
        'file': (file_path, open(file_path, 'rb'), 'audio/mpeg'),
    }
    start_time = time.time()
    # Send POST request
    response = requests.post(ASR_API_URL, headers=headers, files=files)
    inference_time = time.time() - start_time # in seconds

    # Check if response is successful
    if response.status_code == 200:
        return response.json().get("transcription", "No transcription returned.")
    else:
        return f"Error: {response.status_code}, {response.text}"


description_text = """
The Gooya model is one of the most [powerful](https://huggingface.co/spaces/navidved/open_persian_asr_leaderboard) and fastest Persian ASR models. 

Upload an audio file in Persian, and this Model will transcribe it.
"""


# Set up the Gradio interface
gr.Interface(
    fn=transcribe_audio,
    inputs=gr.Audio(type="filepath"),  # Updated here
    outputs="text",
    title="Gooya v1.4 Persian ASR",
    description="The Gooya model is one of the most powerful and fastest Persian ASR models. Upload an audio file in Persian, and this Model will transcribe it."
).launch()