import streamlit as st
from utils import validate_sequence, predict
from model import models
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

def main():
    st.set_page_config(layout="wide")  # Keep the wide layout for overall flexibility
    st.title("AA Property Inference Demo", anchor=None)

    # Styling for the app to use monospace font
    st.markdown("""
        <style>
        .reportview-container {
            font-family: 'Courier New', monospace;
        }
        </style>
        """, unsafe_allow_html=True)

    # Input section in the sidebar
    sequence = st.sidebar.text_input("Enter your amino acid sequence:")
    uploaded_file = st.sidebar.file_uploader("Or upload a CSV file with amino acid sequences", type="csv")
    analyze_pressed = st.sidebar.button("Analyze Sequence")
    show_graphs = st.sidebar.checkbox("Show Prediction Graphs")

    sequences = [sequence] if sequence else []
    if uploaded_file:
        df = pd.read_csv(uploaded_file)
        sequences.extend(df['sequence'].tolist())

    results = []
    all_data = {}
    if analyze_pressed:
        for seq in sequences:
            if validate_sequence(seq):
                model_results = {}
                graph_data = {}
                for model_name, model in models.items():
                    prediction, confidence = predict(model, seq)
                    model_results[f"{model_name}_prediction"] = prediction
                    model_results[f"{model_name}_confidence"] = round(confidence, 3)
                    graph_data[model_name] = (prediction, confidence)
                results.append({"Sequence": seq, **model_results})
                all_data[seq] = graph_data
            else:
                st.sidebar.error(f"Invalid sequence: {seq}")

        if results:
            results_df = pd.DataFrame(results)
            st.write("### Results")
            st.dataframe(results_df.style.format(precision=3), width=None, height=None)
            
            if show_graphs and all_data:
                st.write("## Graphs")
                plot_prediction_graphs(all_data)

def plot_prediction_graphs(data):
    # Function to plot graphs for predictions
    for model_name in models.keys():
        plt.figure(figsize=(10, 4))
        predictions = {seq: values[model_name][1] for seq, values in data.items()}  # Using confidence for ordering
        # Sorting sequences based on confidence, descending
        sorted_sequences = sorted(predictions.items(), key=lambda x: x[1], reverse=True)
        sequences = [x[0] for x in sorted_sequences]
        conf_values = [x[1] for x in sorted_sequences]
        sns.barplot(x=sequences, y=conf_values, palette="viridis")
        plt.title(f'Confidence Scores for {model_name.capitalize()} Model')
        plt.xlabel('Sequences')
        plt.ylabel('Confidence')
        plt.xticks(rotation=45)  # Rotate x labels for better visibility
        st.pyplot(plt)  # Display each plot below the results table

if __name__ == "__main__":
    main()