app.py

import gradio as gr
from gtts import gTTS
from groq import Groq
import whisper
import tempfile
import os

# Initialize Groq client
api = os.environ.get('GroqApi')
client = Groq(api_key=api)

# Load the Whisper model locally
whisper_model = whisper.load_model("base")  # Options: "tiny", "base", "small", "medium", "large"

# Function to handle transcription, LLM response, and audio synthesis
def voice_to_voice(audio_file):
    try:
        # 1. Transcribe the audio using the local Whisper model
        result = whisper_model.transcribe(audio_file)
        user_input = result["text"]

        # 2. Interact with the LLM via Groq API
        chat_completion = client.chat.completions.create(
            messages=[
                {
                    "role": "user",
                    "content": user_input,
                }
            ],
            model="llama3-8b-8192",
            stream=False,
        )
        response_text = chat_completion.choices[0].message.content

        # 3. Convert text response to speech using GTTS
        tts = gTTS(response_text)
        temp_audio_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
        tts.save(temp_audio_file.name)

        return temp_audio_file.name, response_text

    except Exception as e:
        return None, f"Error: {str(e)}"

# Build Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("## Real-Time Voice-to-Voice Chatbot")
    audio_input = gr.Audio(type="filepath", label="Speak Something")
    audio_output = gr.Audio(label="Bot Response")
    text_output = gr.Textbox(label="Transcription & Response")

    btn = gr.Button("Process")
    btn.click(voice_to_voice, inputs=audio_input, outputs=[audio_output, text_output])

# Launch the interface
demo.launch()