app.py

import os
import gradio as gr
from fastrtc import Stream, ReplyOnPause, AdditionalOutputs

# Import your custom models
from tts import tortoise_tts, TortoiseOptions
from stt import whisper_stt
import cohereAPI

# Try to import HumAware-VAD, install if not available
try:
    from humaware_vad import HumAwareVADModel
except ImportError:
    print("Installing humaware-vad...")
    subprocess.check_call([sys.executable, "-m", "pip", "install", "humaware-vad"])
    from humaware_vad import HumAwareVADModel

# Environment variables
COHERE_API_KEY = os.getenv("COHERE_API_KEY")
system_message = "You respond concisely, in about 15 words or less"

# Initialize conversation history
conversation_history = []

# Initialize the HumAware-VAD model
vad_model = HumAwareVADModel()

# Create a handler function that uses both your custom models
def response(audio):
    global conversation_history
    
    # Convert speech to text using your Whisper model
    user_message = whisper_stt.stt(audio)
    
    # Yield the transcription
    yield AdditionalOutputs(user_message)
    
    # Send text to Cohere API
    response_text, updated_history = cohereAPI.send_message(
        system_message, 
        user_message, 
        conversation_history,
        COHERE_API_KEY
    )
    
    # Update conversation history
    conversation_history = updated_history
    
    # Print the response for logging
    print(f"Assistant: {response_text}")
    
    # Use your TTS model to generate audio
    tts_options = TortoiseOptions(voice_preset="random")
    
    # Stream the audio response in chunks
    for chunk in tortoise_tts.stream_tts_sync(response_text, tts_options):
        yield chunk

# Create the FastRTC stream with HumAware-VAD for better pause detection
stream = Stream(
    handler=ReplyOnPause(response, model=vad_model),  # Use HumAware-VAD model
    modality="audio",
    mode="send-receive",
    additional_outputs=[gr.Textbox(label="Transcription")],
    additional_outputs_handler=lambda old, new: new if old is None else f"{old}\nUser: {new}"
)

# Launch the Gradio UI
if __name__ == "__main__":
    # Update your requirements.txt to include humaware-vad
    stream.ui.launch(
        server_name="0.0.0.0",
        share=False,
        show_error=True
    )