app.py

import os

import gradio as gr
import numpy as np
from groq import Groq
from transformers import pipeline

transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-base.en")

groq_client = Groq(api_key=os.getenv('GROQ_API_KEY'))

def transcribe(stream, new_chunk):
    """
    Transcribes using whisper
    """
    sr, y = new_chunk
    y = y.astype(np.float32)
    y /= np.max(np.abs(y))

    if stream is not None:
        stream = np.concatenate([stream, y])
    else:
        stream = y
    return stream, transcriber({"sampling_rate": sr, "raw": stream})["text"]

def autocomplete(text):
    """
    Autocomplete the text using Gemma.
    """
    if text != "":
        response = groq_client.chat.completions.create(
            model='gemma-7b-it',
            messages=[{"role": "system", "content": "You are a friendly assistant named Gemma."},
                      {"role": "user", "content": text}]
            )
            
        return response.choices[0].message.content

def process_audio(input_audio, new_chunk):
    """
    Process the audio input by transcribing and completing the sentences.
    Accumulate results to return to Gradio interface.
    """

    stream, transcription = transcribe(input_audio, new_chunk)
    text = autocomplete(transcription)

    print (transcription, text)
    return stream, text


demo = gr.Interface(
    fn = process_audio,
    inputs = ["state", gr.Audio(sources=["microphone"], streaming=True)],
    outputs = ["state", gr.Markdown()],
    title="Dear Gemma",
    description="Talk to the AI assistant. \n Powered by whisper-base-en, and gemma-7b-it (via Groq)",
    live=True,
    allow_flagging="never"
)

demo.launch()