app.py

import os
import streamlit as st
from groq import Groq
from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq, pipeline  # Import pipeline
from espnet2.bin.tts_inference import Text2Speech
import soundfile as sf
from pydub import AudioSegment
import io
from dotenv import load_dotenv

# Load environment variables from .env file
load_dotenv()

# Load Groq API key from .env file
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
if not GROQ_API_KEY:
    st.error("Groq API key not found. Please add it to the .env file.")
    st.stop()

# Initialize Groq client
groq_client = Groq(api_key=GROQ_API_KEY)

# Load models
@st.cache_resource  # Use st.cache_resource for caching models
def load_models():
    # Speech-to-Text
    processor = AutoProcessor.from_pretrained("openai/whisper-small")
    stt_model = AutoModelForSpeechSeq2Seq.from_pretrained("openai/whisper-small")
    stt_pipe = pipeline(
        "automatic-speech-recognition",
        model=stt_model,
        tokenizer=processor.tokenizer,
        feature_extractor=processor.feature_extractor
    )

    # Text-to-Speech
    tts_model = Text2Speech.from_pretrained("espnet/espnet_tts_vctk_espnet_spk_voxceleb12_rawnet")

    return stt_pipe, tts_model

stt_pipe, tts_model = load_models()

# Streamlit app
st.title("Voice-Enabled Chatbot")

# Audio input
audio_file = st.file_uploader("Upload your voice input", type=['wav'])
if audio_file is not None:
    audio_bytes = audio_file.read()
    audio = AudioSegment.from_file(io.BytesIO(audio_bytes))
    audio.export("temp.wav", format="wav")
    speech, _ = sf.read("temp.wav")
    text = stt_pipe(speech)['text']
    st.write("Transcribed Text:", text)

    # Generate response using Groq API
    try:
        chat_completion = groq_client.chat.completions.create(
            messages=[{"role": "user", "content": text}],
            model="mixtral-8x7b-32768",
            temperature=0.5,
            max_tokens=1024
        )
        response = chat_completion.choices[0].message.content
        st.write("Generated Response:", response)

        # Convert response to speech
        speech, *_ = tts_model(response)
        sf.write("response.wav", speech, 22050)
        st.audio("response.wav")
    except Exception as e:
        st.error(f"Error generating response: {e}")