app.py

import gradio as gr
import requests
import soundfile as sf
import tempfile
import os

# URL of your dedicated processing server (adjust port/path if needed)
SERVER_URL = "http://204.12.245.139:5000/process_audio"

def process_audio(audio):
    """
    Process the input audio:
      - Expects 'audio' as a tuple: (sample_rate, numpy_array)
      - Writes the audio data to a temporary WAV file
      - Sends the file via POST to your server endpoint
      - Returns the transcription (or server response) as text
    """
    if audio is None:
        return "No audio provided. Please record something."
    
    sample_rate, audio_data = audio
    # Write the audio data to a temporary WAV file.
    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp_file:
        wav_path = tmp_file.name
        sf.write(wav_path, audio_data, sample_rate)
    
    try:
        with open(wav_path, "rb") as f:
            files = {"file": f}
            response = requests.post(SERVER_URL, files=files, timeout=30)
        if response.status_code == 200:
            json_data = response.json()
            # Expect server to return a key 'transcription' or 'response'
            result = json_data.get("transcription") or json_data.get("response")
            if not result:
                result = "Server processed the audio, but did not return a result."
        else:
            result = f"Server error {response.status_code}: {response.text}"
    except Exception as e:
        result = f"Exception during processing: {e}"
    finally:
        os.remove(wav_path)
    
    return result

# Create a Gradio interface.
# Note: We removed "source" from gr.Audio to ensure compatibility with your Gradio version.
iface = gr.Interface(
    fn=process_audio,
    inputs=gr.Audio(type="numpy", label="Record Your Voice"),
    outputs=gr.Textbox(label="Server Response"),
    title="Live AI Call Agent – Browser Mic Frontend",
    description=(
        "Record audio using your browser microphone. The audio will be sent to our dedicated server "
        "for processing with GPU acceleration. Your server should return a transcription or generated response."
    )
)

if __name__ == "__main__":
    # Launch the Gradio app to listen on all interfaces. Hugging Face Spaces will assign a public URL.
    iface.launch(server_name="0.0.0.0", server_port=7860)