import gradio as gr
from audio import predict_all  # This is your custom module for predictions
import re  # Regular expressions for text processing

# Additional CSS for styling the confidence bars and the result layout
additional_css = """
/* CSS for the confidence bars */
.confidence-section {
    display: flex;
    align-items: center;
    margin-top: 10px;
}

.confidence-label {
    margin-right: 10px;
    font-weight: bold;
}
.confidence-bar {
    height: 20px;
    width: 100%;
    background-color: #eee;
    border-radius: 10px;
    margin: 10px 0;
}

.confidence-fill {
    height: 100%;
    border-radius: 10px;
    background-color: #4caf50; /* Change color based on confidence level if desired */
    text-align: center;
    color: white;
    line-height: 20px;
}
/* Additional CSS for styling the rest of your results */
"""

# Function to generate custom HTML for the confidence bar
def custom_confidence_bar(confidence):
    color = "#4caf50" if confidence > 75 else "#FFC107" if confidence > 50 else "#F44336"
    return f"""
<div class="confidence-section">
    <span class="confidence-label">Model Confidence:</span>
    <div class="confidence-bar">
        <div class="confidence-fill" style="width: {confidence}%; background-color: {color};">
            {confidence}%
        </div>
    </div>
</div>
    """
    

# Function to extract score level from message
def extract_score_level(message):
    match = re.search(r'Score: (\d+)-(\d+)', message)
    score_level = f"{match.group(1)} of 10" if match else "N/A"
    return score_level
def message_markdown(label, message, task, score_level):
    md = f'''# {label}
    **Model Prediction:** {message}
    
    **{task} Score:** {score_level}
    '''
    return md
# Function to process the audio file and analyze it
def analyze_audio(audio_data):
    # Assuming predict_all returns a tuple of (message, confidence) for accuracy and fluency
    accuracy, fluency = predict_all(audio_data)

    # Unpack the results
    accuracy_message, accuracy_confidence = accuracy
    fluency_message, fluency_confidence = fluency

    # Extract the score level from the message
    accuracy_score = extract_score_level(accuracy_message)
    fluency_score = extract_score_level(fluency_message)

    # Remove the score level from the original message
    accuracy_message = accuracy_message.split(",")[1].strip() if "," in accuracy_message else accuracy_message
    fluency_message = fluency_message.split(",")[1].strip() if "," in fluency_message else fluency_message

    # Generate the confidence bar HTML
    accuracy_confidence_html = custom_confidence_bar(accuracy_confidence * 100)
    fluency_confidence_html = custom_confidence_bar(fluency_confidence * 100)
    
    accuracy_markdown = message_markdown('Accuracy of Pronunciation', accuracy_message, 'Pronunciation', accuracy_score)
    
    fluency_markdown = message_markdown('Speaker Fluency', fluency_message, 'Fluency', fluency_score)
    
    return accuracy_markdown, accuracy_confidence_html, fluency_markdown, fluency_confidence_html

# Define the Gradio interface
iface = gr.Interface(
    fn=analyze_audio,
    inputs=gr.Audio(label="Upload Audio"),
    outputs=[
        gr.Markdown(label="Accuracy Score Level"),
        gr.HTML(label="Accuracy Confidence"),
        gr.Markdown(label="Fluency Score Level"),
        gr.HTML(label="Fluency Confidence"),
    ],
    css=additional_css,
    title="Audio Analysis Tool",
    description="Upload an audio file to analyze its accuracy and fluency."
)

# Run the Gradio app
if __name__ == "__main__":
    iface.launch()