app.py

import os
os.environ["GROQ_API_KEY"] = "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"
import gradio as gr
import whisper
from gtts import gTTS
import io
from groq import Groq

# Initialize the Whisper model
model = whisper.load_model("base")

# Initialize the Groq client with the API key from environment variables
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))

def process_audio(file_path):
    try:
        if file_path is None:
            raise ValueError("No audio file received. Please upload or record your voice.")

        # Load the audio file and transcribe using Whisper
        audio = whisper.load_audio(file_path)
        result = model.transcribe(audio)
        text = result["text"]

        # Generate a response using Groq
        chat_completion = client.chat.completions.create(
            messages=[{"role": "user", "content": text}],
            model="llama3-8b-8192",
        )
        response_message = chat_completion.choices[0].message.content.strip()

        # Convert the response text to speech
        tts = gTTS(response_message)
        response_audio_io = io.BytesIO()
        tts.write_to_fp(response_audio_io)
        response_audio_io.seek(0)

        # Save the audio to a file
        audio_file_path = "response.mp3"
        with open(audio_file_path, "wb") as audio_file:
            audio_file.write(response_audio_io.getvalue())

        # Return the response text and path to the saved audio file
        return response_message, audio_file_path

    except Exception as e:
        return f"An error occurred: {e}", None

# Gradio interface
iface = gr.Interface(
    fn=process_audio,
    inputs=gr.Audio(type="filepath"),
    outputs=[gr.Textbox(label="Response Text"), gr.Audio(label="Response Audio")],
    live=True,  # Enables real-time processing
    title="Voice-to-Voice Chatbot",
    description="A real-time voice-to-voice chatbot using OpenAI Whisper, Groq LLaMA model, and GTTS."
)

# Launch the Gradio interface
iface.launch()