app.py

# app.py
import gradio as gr
import openai
import os
import RadinMas  # Importing the RadinMas module
import base64

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY") 
openai.api_key = OPENAI_API_KEY

def image_to_base64(img_path):
    with open(img_path, "rb") as img_file:
        return base64.b64encode(img_file.read()).decode('utf-8')

img_base64 = image_to_base64("RadinMasSBC.JPG")
img_html = f'<img src="data:image/jpg;base64,{img_base64}" alt="SBC6" width="300" style="display: block; margin: auto;"/>'

def predict(question_choice, feedback_level, audio):
    # Transcribe the audio using Whisper
    with open(audio, "rb") as audio_file:
        transcript = openai.Audio.transcribe("whisper-1", audio_file)
    message = transcript["text"]  # This is the transcribed message from the audio input

    # Generate the prompt based on the feedback level
    feedback_prompt = RadinMas.generate_prompt(feedback_level) 
    
    # Determine question number based on question_choice
    question_number = RadinMas.questions.index(question_choice) + 1  # New line

    # Generate the system message based on the question number
    system_message = RadinMas.generate_system_message(question_number, feedback_level)  # Updated line to include feedback_level

    
    # Reference to the picture description from RadinMas.py
    picture_description = RadinMas.description

    # Determine whether to include the picture description based on the question choice
    picture_description_inclusion = f"""
    For the first question, ensure your feedback refers to the picture description provided:
    {picture_description}
    """ if question_choice == RadinMas.questions[0] else ""

    # Construct the conversation with the system and user's message
    conversation = [
        {
            "role": "system",
            "content": f"""
            You are an expert English Language Teacher in a Singapore Primary school, directly guiding a Primary 6 student in Singapore. 
            The student is answering the question: '{question_choice}'. 
            {picture_description_inclusion}
            {system_message}
            """
        },
        {"role": "user", "content": message}
    ]

    response = openai.ChatCompletion.create(
        model='gpt-3.5-turbo',
        messages=conversation,
        temperature=0.6,
        max_tokens=550,  # Limiting the response to 550 tokens
        stream=True
    )

    partial_message = ""
    for chunk in response:
        if len(chunk['choices'][0]['delta']) != 0:
            partial_message = partial_message + chunk['choices'][0]['delta']['content']
            yield partial_message

# Gradio Interface
iface = gr.Interface(
    fn=predict,
    inputs=[
        gr.Radio(RadinMas.questions, label="Choose a question", default=RadinMas.questions[0]),  # Dropdown for question choice
        gr.Radio(["Brief Feedback", "Moderate Feedback", "Comprehensive Feedback"], label="Choose a feedback level", default="Brief Feedback"),  # Ensure this line is present
        gr.inputs.Audio(source="microphone", type="filepath")  # Audio input
    ],
    outputs=gr.inputs.Textbox(),  # Using inputs.Textbox as an output to make it editable
    description=img_html + '''
        <div style="text-align: center; font-size: medium;">
            <a href="https://forms.moe.edu.sg/forms/J0lmkJ" target="_blank">
                &#x1F4DD; Click here to provide feedback on the initial prototype of the Oral Coach &#x1F4DD;
            </a>
        </div>
    ''',  # Corrected string concatenation
    css="custom.css"  # Link to the custom CSS file
)

iface.queue(max_size=99, concurrency_count=40).launch(debug=True)