app.py

import gradio as gr
from main import index, run
from gtts import gTTS
import os, time

from transformers import pipeline

p = pipeline("automatic-speech-recognition")

"""Use text to call chat method from main.py"""

models = ["GPT-3.5", "Flan UL2", "GPT-4", "Flan T5"]

def add_text(history, text, model):
    print("Question asked: " + text)
    response = run_model(text, model)
    history = history + [(text, response)]
    print(history)
    return history, ""


def run_model(text, model):
    start_time = time.time()
    print("start time:" + str(start_time))
    response = run(text, model)
    end_time = time.time()
    # If response contains string `SOURCES:`, then add a \n before `SOURCES`
    if "SOURCES:" in response:
        response = response.replace("SOURCES:", "\nSOURCES:")
        # response = response + "\n\n" + "Time taken: " + str(end_time - start_time)
    print(response)
    print("Time taken: " + str(end_time - start_time))
    return response



def get_output(history, audio, model):

    txt = p(audio)["text"]
    # history.append(( (audio, ) , txt))
    audio_path = 'response.wav'
    response = run_model(txt, model)
    # Remove all text from SOURCES: to the end of the string
    trimmed_response = response.split("SOURCES:")[0]
    myobj = gTTS(text=trimmed_response, lang='en', slow=False)
    myobj.save(audio_path)
    # split audio by / and keep the last element
    # audio = audio.split("/")[-1]
    # audio = audio + ".wav"
    history.append(( (audio, ) , (audio_path, )))
    print(history)
    return history

def set_model(history, model):
    print("Model selected: " + model)
    history = get_first_message(history)
    index(model)
    return history


def get_first_message(history):
    history = [(None,
                '''Hi!! I AM GRADY!! I am a grading assistant to help you grade assignments based on a rubric!! <br>
                Today, I will be grading <a href="https://hbsp.harvard.edu/product/908D01-PDF-ENG"> Paediatric Orthopaedic Quiz. </a> <br>
                Use the format as given in the example below to get an accurate grade. <br> 
                WARNING! I might get things wrong, so double check before your final grading. All the best. ''')]
    return history


def bot(history):
    return history

with gr.Blocks() as demo:
    gr.HTML("<h1 style='text-align: center;color: darkblue'>Grady - Your helpful Grading Assistant</h1>")
    chatbot = gr.Chatbot(get_first_message([]), elem_id="chatbot", interactive=True).style(height=500)

    with gr.Row():
        # Create radio button to select model
        radio = gr.Radio(models, label="Choose a model", value="GPT-3.5", type="value", visible=False)
    with gr.Row():
        with gr.Column(scale=0.75):
            txt = gr.Textbox(
                label="Student Response",
                placeholder="Enter text and press enter", lines=1, interactive=True
            ).style(container=False)

        with gr.Column(scale=0.25):
            audio = gr.Audio(source="microphone", type="filepath").style(container=False)
    with gr.Row():
        gr.Examples(examples=["""11: Currently the process is not very efficient as each patient goes through the same steps at the front desk and the radiology department although the sub-activities and processes are different. Also, the staff is doing multiple activities based on patient requirements.  

One solution is to have a streamlined and differentiated process for each sub-type with dedicated staff. For example, at the front desk, all new patient cases can be handled by one nurse while all follow-up cases by a second nurse. 

Similarly, in the radiology department, all upper extremity cases can be handled by 2 technicians while lower extremity cases by the other 2 technicians with dedicated X-ray machines. The 3rd nurse will be responsible for handling the hand-off of X-rays and inserting them into the patient's files.  

By having staff do a single type of task on a particular day, and by having the patients go through differentiated workflows, it should be possible to improve overall efficiency. """], inputs=[txt], label="Answers")

    txt.submit(add_text, [chatbot, txt, radio], [chatbot, txt], postprocess=False).then(
        bot, chatbot, chatbot
    )

    audio.change(fn=get_output, inputs=[chatbot, audio, radio], outputs=[chatbot]).then(
        bot, chatbot, chatbot
    )

    radio.change(fn=set_model, inputs=[chatbot, radio], outputs=[chatbot]).then(bot, chatbot, chatbot)

    audio.change(lambda:None, None, audio)

    set_model(chatbot, radio.value)



if __name__ == "__main__":
    demo.queue()
    demo.queue(concurrency_count=5)
    demo.launch(debug=True)