app.py

import os
import numpy as np
import pandas as pd
import joblib
import pickle
import streamlit as st
import seaborn as sns
from streamlit_option_menu import option_menu
import time
import matplotlib.pyplot as plt
import json
import google.generativeai as genai
from dotenv import load_dotenv
from transformers import pipeline

# Set page config with icon
st.set_page_config(page_title="Disease Prediction", page_icon="🩺", layout="wide")

diabetes_model = pickle.load(open('diabetes/diabetes_model.sav', 'rb'))
# asthama_model = pickle.load(open('asthama/model.pkl', 'rb'))

import joblib
asthama_model = joblib.load("asthama/model.pkl")

cardio_model = pickle.load(open('cardio_vascular/xgboost_cardiovascular_model.pkl', 'rb'))

# stroke_model = pickle.load(open('stroke/stroke_model.sav', 'rb'))

stroke_model = joblib.load("stroke/finalized_model.pkl")

prep_asthama = pickle.load(open('asthama/preprocessor.pkl', 'rb'))

# sleep_model = pickle.load(open('sleep_health/best_model.pkl', 'rb'))
# scaler = pickle.load(open('sleep_health/scaler.pkl', 'rb'))
# label_encoder = pickle.load(open('sleep_health/label_encoders.pkl', 'rb'))

# At the beginning of your app, when loading models:
try:
    sleep_model = pickle.load(open('sleep_health/svc_model.pkl', 'rb'))
    scaler = pickle.load(open('sleep_health/scaler.pkl', 'rb'))
    label_encoder = pickle.load(open('sleep_health/label_encoders.pkl', 'rb'))
    
    # Store the expected feature names if available
    # This depends on how your model was saved/trained
    # if hasattr(sleep_model, 'feature_names_in_'):
    #     expected_features = sleep_model.feature_names_in_
    # else:
    #     # Try to load feature names from a separate file
    #     try:
    #         with open('sleep_health/feature_names.pkl', 'rb') as f:
    #             expected_features = pickle.load(f)
    #     except:
    #         st.warning("Warning: Feature names not found. Predictions may be inaccurate.")
    #         expected_features = None
            
except FileNotFoundError:
    st.error("Error: Model files not found. Please upload the model files.")
    st.stop()

with st.sidebar:
    st.title("🩺 Disease Prediction")
    
    selected = option_menu(
        menu_title="Navigation",
        options=['Home',  'Diabetes Prediction','Hypertension Prediction', 'Cardiovascular Disease Prediction', 'Stroke Prediction','Asthma Prediction', 'Sleep Health Analysis','Mental-Analysis','Medical Consultant', 'Data Visualization'],
        icons=['house', 'activity', 'lungs', 'heart-pulse', 'brain', 'bar-chart', 'chat'],
        menu_icon="cast",
        default_index=0,
        styles={
            "container": {"padding": "5px", "background-color": "#111111"},  # Darker background
            "icon": {"color": "#FF0000", "font-size": "20px"},  # Red icons
            "nav-link": {"font-size": "16px", "text-align": "left", "margin": "0px", "color": "#FFFFFF"},  # White text
            "nav-link-selected": {"background-color": "#FF0000", "color": "#FFFFFF"},
        }
    )

# 'Mental-Analysis',

# 'Checkbox-to-disease-predictor', 
# 'Text-based Disease Prediction', 
# Utility function to safely convert input to float
def safe_float(value, default=0.0):
    try:
        return float(value)
    except ValueError:
        return default  # Assigns default value if conversion fails


# 🚀 Home Page
if selected == 'Home':
    st.title("🩺 Early Prediction of Health & Lifestyle Diseases")

    st.markdown("""
    ## Welcome to the **Early Prediction of Health & Lifestyle Diseases**!  
    This tool provides **early prediction and analysis** for various health conditions using **Machine Learning & NLP**.
    
    ### 🏥 Available Features:
    - **🩸 Disease Predictors**:
      - Diabetes Prediction   
      - Hypertension Prediction  
      - Cardiovascular Disease Prediction 
      - Stroke Prediction
      - Asthma Prediction 
      - Sleep Health Analysis
    # - **✅ Checkbox-based Lifestyle Disease Predictor** using **BiomedNLP-PubMedBERT**  
    - **🤖 Medical Consultant** (Ask health-related questions)  
    # - **🧠 Mental Health Assessment**  
    - **📊 Data Visualizer** (Analyze trends in health conditions)  
                
    👉 Select an option from the sidebar to proceed!  
    """)

    with st.expander("🚀 Quick Start Guide"):
        st.write("""
        1. Select a **health prediction model** from the sidebar.
        2. Enter your details in the input fields.
        3. Click **Predict** to get your result.
        4. View personalized **health insights & recommendations**.
        """)


    # Disclaimer Section
    st.markdown("---")
    # st.markdown("""
    # **⚠️ Disclaimer:** This application has been developed using **real-world healthcare datasets** sourced from Kaggle:  

    # - [Stroke Prediction Dataset](http://kaggle.com/code/chanchal24/stroke-prediction-using-python/input?select=healthcare-dataset-stroke-data.csv)  
    # - [Asthma Analysis & Prediction](https://www.kaggle.com/code/bryamblasrimac/asthma-eda-prediction-f2score-85/input)  
    # - [Diabetes Dataset](https://www.kaggle.com/datasets/mathchi/diabetes-data-set)  
    # - [Cardiovascular Disease Dataset](https://www.kaggle.com/datasets/sulianova/cardiovascular-disease-dataset)  
    # - [Sentiment Analysis for Mental Health](https://www.kaggle.com/datasets/suchintikasarkar/sentiment-analysis-for-mental-health)  
    # - [Sleep Health Analysis](https://www.kaggle.com/datasets/uom190346a/sleep-health-and-lifestyle-dataset)
    # \


    st.markdown("""
    The predictions are generated using **machine learning models** trained on these datasets, incorporating **evaluation metrics and graphical insights** to enhance interpretability.  

    However, this tool has **not undergone clinical validation** and should be used **for informational and educational purposes only**. It is not intended to serve as a substitute for professional medical diagnosis or treatment. Always consult a qualified healthcare provider for medical advice.
    """)


if selected == 'Diabetes Prediction':
    st.title('🩸 Diabetes Prediction using ML (SVC)')
    st.image("https://cdn-icons-png.flaticon.com/512/2919/2919950.png", width=100)

    st.markdown("""
    This model predicts the likelihood of **Diabetes** based on various health parameters.  
    Please enter the required medical details below and click **"Diabetes Test Result"** to get the prediction.
    """)



    # Create columns for better input organization
    col1, col2 = st.columns(2)

    with col1:
        gender = st.radio("Gender", ["Male", "Female"], horizontal=True)
        
        # For females, show Pregnancies input; for males, set to 0
        if gender == "Female":
            Pregnancies = safe_float(st.text_input("Number of Pregnancies", "0"))
        else:
            Pregnancies = 0.0
            st.info("Pregnancies set to 0 for males")
            
        Glucose = safe_float(st.text_input("Glucose Level", "100"))
        BloodPressure = safe_float(st.text_input("Blood Pressure", "80"))
        SkinThickness = safe_float(st.text_input("Skin Thickness", "20"))

    with col2:
        Insulin = safe_float(st.text_input("Insulin Level", "79"))
        BMI = safe_float(st.text_input("BMI (Body Mass Index)", "25.0"))
        DiabetesPedigreeFunction = safe_float(st.text_input("Diabetes Pedigree Function", "0.5"))
        Age = st.number_input("Enter Age", min_value=10, max_value=100, value=30, step=1)

    with col1:
        if st.button('Diabetes Test Result'):
            try:
                input_data = np.array([[Pregnancies, Glucose, BloodPressure, SkinThickness, Insulin, BMI, DiabetesPedigreeFunction, Age]])
                
                with st.spinner("⏳ Predicting... Please wait..."):
                    time.sleep(2)  # Simulating delay (remove in actual use)
                    diab_prediction = diabetes_model.predict(input_data)
                
                result = "🛑 The person is diabetic" if diab_prediction[0] == 1 else "✅ The person is not diabetic"
                if diab_prediction[0] == 0:
                    # st.balloons()  # Or use st.confetti() if you install the library
                    st.success(result)
                else:
                    st.error(result)

            except Exception as e:
                st.error(f"❌ Error: {e}")


if selected == 'Asthma Prediction':
    st.title('🌬️ Asthma Prediction using ML')
    st.image("https://cdn-icons-png.flaticon.com/512/3462/3462191.png", width=100)

    st.markdown("""
    This model predicts the likelihood of **Asthma** based on health factors.  
    Enter your details and click **"Asthma Test Result"** to get the prediction.
    """)

    col1, col2 = st.columns(2)

    with col1:
        Gender_Male = st.radio("Gender", ["Female", "Male"])
        Gender_Male = 1 if Gender_Male == "Male" else 0

        Smoking_Status = st.radio("Smoking Status", ["Non-Smoker", "Ex-Smoker"])
        Smoking_Status_Ex_Smoker = 1 if Smoking_Status == "Ex-Smoker" else 0
        Smoking_Status_Non_Smoker = 1 if Smoking_Status == "Non-Smoker" else 0

    with col2:
        # Use actual age as input instead of normalized value
        actual_age = st.slider("Age", min_value=18, max_value=85, value=40, help="Your actual age in years")
        
        # Convert actual age to normalized value (0.0 to 0.914894)
        # Assuming normalization was done with min_age=18 and max_age=90
        min_age = 18
        max_age = 90
        Age = (actual_age - min_age) / (max_age - min_age)
        
        # Show the normalized value for reference (can be hidden in final version)
        st.info(f"Normalized age value (used by model): {Age:.6f}")
        
        Peak_Flow = st.slider("Peak Flow (L/sec)", min_value=0.1, max_value=1.0, value=0.5)

    with col1:
        if st.button('Asthma Test Result'):
            try:
                # Prepare raw input
                raw_input = np.array([[Gender_Male, Smoking_Status_Ex_Smoker, Smoking_Status_Non_Smoker, Age, Peak_Flow]])

                # Check if preprocessing is needed
                if prep_asthama is not None and hasattr(prep_asthama, "transform"):
                    processed_input = prep_asthama.transform(raw_input)  # Use transform if prep_asthama exists
                else:
                    processed_input = raw_input  # If no scaler, use raw input

                with st.spinner("⏳ Predicting... Please wait..."):
                    time.sleep(2)  # Simulating delay (remove in actual use)
                    asthma_prediction = asthama_model.predict(processed_input)

                result = "🛑 High risk of asthma" if asthma_prediction[0] == 1 else "✅ Low risk of asthma"
                if asthma_prediction[0] == 0:
                    # st.balloons()
                    st.success(result)
                else:
                    st.error(result)
                
                # Add risk factor analysis
                st.subheader("Risk Factor Analysis")
                risk_factors = []
                
                if actual_age > 60:
                    risk_factors.append("⚠️ Age is a risk factor for asthma")
                if Smoking_Status == "Ex-Smoker":
                    risk_factors.append("⚠️ Smoking history increases asthma risk")
                if Peak_Flow < 0.5:
                    risk_factors.append("⚠️ Low peak flow readings may indicate restricted airways")
                
                if risk_factors:
                    for factor in risk_factors:
                        st.markdown(factor)
                else:
                    st.markdown("✅ No significant risk factors identified")

            except Exception as e:
                st.error(f"❌ Error: {e}")
                st.info("If you have access to the preprocessing pipeline, you can check the exact age normalization formula used during model training.")


if selected == 'Cardiovascular Disease Prediction':
    st.title('❤️ Cardiovascular Disease Prediction')
    st.image("https://cdn-icons-png.flaticon.com/512/2919/2919950.png", width=100)

    st.markdown("""
    This model predicts the likelihood of **Cardiovascular Disease** based on various health parameters.  
    Please enter the required medical details below and click **"Cardio Test Result"** to get the prediction.
    """)

    # Input Fields
    col1, col2 = st.columns(2)

    with col1:
        age = st.number_input("Age", min_value=29, max_value=64, value=40, step=1)
        ap_hi = st.slider("Systolic Blood Pressure (ap_hi)", min_value=90, max_value=180, value=120)
        ap_lo = st.slider("Diastolic Blood Pressure (ap_lo)", min_value=60, max_value=120, value=80)
        weight = st.number_input("Weight (kg)", min_value=40.0, max_value=180.0, value=70.0, step=0.1)

    with col2:
        cholesterol = st.radio("Cholesterol Level", ["Normal", "Above Normal", "Well Above Normal"])
        cholesterol = {"Normal": 1, "Above Normal": 2, "Well Above Normal": 3}[cholesterol]

        gluc = st.radio("Glucose Level", ["Normal", "Above Normal", "Well Above Normal"])
        gluc = {"Normal": 1, "Above Normal": 2, "Well Above Normal": 3}[gluc]

        smoke = st.radio("Smoking Status", ["No", "Yes"])
        smoke = 1 if smoke == "Yes" else 0

        alco = st.radio("Alcohol Consumption", ["No", "Yes"])
        alco = 1 if alco == "Yes" else 0

        active = st.radio("Physically Active", ["No", "Yes"])
        active = 1 if active == "Yes" else 0

    # Prediction Button
    if st.button('Cardio Test Result'):
        try:
            # Preparing Input Data
            input_data = np.array([[age, ap_hi, ap_lo, cholesterol, gluc, smoke, alco, active, weight]])

            with st.spinner("⏳ Predicting... Please wait..."):
                time.sleep(2)  # Simulating Model Inference
                cardio_prediction = cardio_model.predict(input_data)

            # Display Result
            result = "🛑 High risk of cardiovascular disease" if cardio_prediction[0] == 1 else "✅ Low risk of cardiovascular disease"
            if cardio_prediction[0] == 0:
                # st.balloons()
                st.success(result)

        except Exception as e:
            st.error(f"❌ Error: {e}")





if selected == 'Stroke Prediction':
    st.title('🧠 Stroke Prediction using ML')
    st.image("https://cdn-icons-png.flaticon.com/512/3209/3209265.png", width=100)

    st.markdown("""
    This model predicts the likelihood of **Stroke** based on various health factors.  
    Enter your details and click **"Stroke Test Result"** to get the prediction.
    """)

    col1, col2 = st.columns(2)

    with col1:
        Age = st.number_input("Age", min_value=0, max_value=82, value=50, step=1)
        Hypertension = st.radio("Hypertension", ["No", "Yes"])
        Hypertension = 1 if Hypertension == "Yes" else 0

        Heart_Disease = st.radio("Heart Disease", ["No", "Yes"])
        Heart_Disease = 1 if Heart_Disease == "Yes" else 0

    with col2:
        Ever_Married = st.radio("Ever Married", ["No", "Yes"])
        Ever_Married = 1 if Ever_Married == "Yes" else 0

        Avg_Glucose_Level = st.slider("Average Glucose Level", min_value=55.23, max_value=267.61, value=120.0)
        BMI = st.slider("BMI", min_value=13.5, max_value=97.6, value=25.0)

        Smoking_Status = st.selectbox("Smoking Status", ["Never Smoked", "Former Smoker", "Smokes", "Unknown"])
        Smoking_Status = {"Never Smoked": 0, "Former Smoker": 1, "Smokes": 2, "Unknown": 3}[Smoking_Status]

    with col1:
        if st.button('Stroke Test Result'):
            try:
                input_data = np.array([[Age, Hypertension, Heart_Disease, Ever_Married, Avg_Glucose_Level, BMI, Smoking_Status]])

                with st.spinner("⏳ Predicting... Please wait..."):
                    time.sleep(2)
                    stroke_prediction = stroke_model.predict(input_data)

                result = "🛑 High risk of stroke" if stroke_prediction[0] == 1 else "✅ Low risk of stroke"
                if stroke_prediction[0] == 0:
                    st.balloons()
                st.success(result)

            except Exception as e:
                st.error(f"❌ Error: {e}")




if selected == 'Data Visualization':
    # st.set_page_config(page_title="Data Visualizer",
    #                 page_icon="📊", layout="centered")
    st.title(" 📊 Data Visualization")

    working_dir = os.path.dirname(os.path.abspath(__file__))

    folder_path = f"{working_dir}/data_csv"

    files_list = [f for f in os.listdir(folder_path) if f.endswith('.csv')]

    selected_file = st.selectbox("Select a file", files_list, index=None)

    if selected_file:

        file_path = os.path.join(folder_path, selected_file)

        df = pd.read_csv(file_path)

        columns = df.columns.tolist()

        col1, col2 = st.columns(2)

        with col1:
            st.write("")
            st.write(df.head())

        with col2:
            x_axis = st.selectbox("Select X-axis", options=columns + ["None"])
            y_axis = st.selectbox("Select Y-axis", options=columns + ["None"])

            plot_list = ["Line Plot", "Bar Plot", "Scatter Plot", "Histogram", "Box Plot", "Distribution Plot", "Count Plot", "Pair Plot"]

            selected_plot = st.selectbox("Select a plot", options=plot_list, index=None)

            # st.write(x_axis)
            # st.write(y_axis)
            # st.write(selected_plot)

        if st.button("Generate Plot"):

            fig, ax = plt.subplots(figsize=(6,4))

            if selected_plot == "Line Plot":
                sns.lineplot(x=x_axis, y=y_axis, data=df, ax=ax)

            elif selected_plot == "Bar Plot":
                sns.barplot(x=x_axis, y=y_axis, data=df, ax=ax)
            
            elif selected_plot == "Scatter Plot":
                sns.scatterplot(x=x_axis, y=y_axis, data=df, ax=ax)
            
            elif selected_plot == "Histogram":
                sns.histplot(df[x_axis], ax=ax)
            
            elif selected_plot == "Box Plot":
                sns.boxplot(x=x_axis, y=y_axis, data=df, ax=ax)

            elif selected_plot == "Distribution Plot":
                sns.kdeplot(df[x_axis], ax=ax)
            
            elif selected_plot == "Count Plot":
                sns.countplot(x=x_axis, data=df, ax=ax)
            
            elif selected_plot == "Pair Plot":
                sns.pairplot(df, ax=ax)

            ax.tick_params(axis="x", labelsize=10)
            ax.tick_params(axis="y", labelsize=10)

            plt.title(f"{selected_plot} of {x_axis} vs {y_axis}", fontsize=12)
            plt.xlabel(x_axis, fontsize=10)
            plt.ylabel(y_axis, fontsize=10)
            
            st.pyplot(fig)


# if selected == 'Medical Consultant':
#     st.title("🩺 Medical Consultant Chatbot")
#     st.markdown("### Discuss Your Health Concerns with Our AI-powered Chatbot")
#     st.write("Ask about **Diabetes, Asthma, Stroke, Cardiovascular Disease, or Mental Health.**")

#     genai.configure(api_key="AIzaSyAwyi9c5OdvLoWrv5lFi1jZDEYwuprQAKE")

#     # Custom Styling
#     st.markdown("""
#         <style>
#             .prompt-box { 
#                 background-color: #000000; 
#                 padding: 12px; 
#                 border-radius: 8px; 
#                 font-size: 14px; 
#                 font-family: sans-serif;
#                 margin-bottom: 10px;
#                 border: 1px solid #dee2e6;
#                 text-align: center;
#             }
#         </style>
#     """, unsafe_allow_html=True)

#     st.markdown("#### 💡 Common Health Queries")

#     prompt_options = [
#         ("Diabetes – Diet", "What foods should I eat if I have diabetes?"),
#         ("Diabetes – Exercise", "What type of workouts help control blood sugar levels?"),
#         ("Asthma – Triggers", "What are common asthma triggers?"),
#         ("Asthma – Treatment", "What are the best medications for asthma?"),
#         ("Stroke – Symptoms", "What are the early warning signs of a stroke?"),
#         ("Stroke – Prevention", "How can I reduce my risk of stroke?"),
#         ("Cardiovascular – Heart Health", "How can I reduce my risk of heart disease?"),
#         ("Cardiovascular – Blood Pressure", "What lifestyle changes can lower high blood pressure?"),
#         ("Mental Health – Stress Management", "How can I manage stress effectively?"),
#         ("Mental Health – Sleep Disorders", "What are the causes and treatments for sleep disorders?")
#     ]

#     # Display prompts in two columns (2 prompts per row)
#     cols = st.columns(2)
#     for i in range(0, len(prompt_options), 2):
#         with cols[0]: 
#             if i < len(prompt_options):
#                 label, prompt = prompt_options[i]
#                 st.markdown(f"""<div class="prompt-box"><strong>{label}</strong><br>{prompt}</div>""", unsafe_allow_html=True)

#         with cols[1]: 
#             if i+1 < len(prompt_options):
#                 label, prompt = prompt_options[i+1]
#                 st.markdown(f"""<div class="prompt-box"><strong>{label}</strong><br>{prompt}</div>""", unsafe_allow_html=True)

#     # Initialize chat history if not present
#     if "chat_history" not in st.session_state:
#         st.session_state.chat_history = []

#     # Display previous chat history
#     for message in st.session_state.chat_history:
#         with st.chat_message(message["role"]):
#             st.markdown(message["content"])

#     # User input field
#     user_prompt = st.chat_input("Ask about Diabetes, Asthma, Stroke, Cardiovascular Disease, or Mental Health...")

#     # List of allowed topics
#     allowed_keywords = ["diabetes", "asthma", "stroke", "cardiovascular", "heart", "blood pressure", 
#                         "mental health", "depression", "stress", "cholesterol", "sleep disorders"]

#     if user_prompt:
#         # Display user message
#         st.chat_message("user").markdown(user_prompt)
#         st.session_state.chat_history.append({"role": "user", "content": user_prompt})

#         # Restriction: Only process if related to health topics
#         if any(keyword in user_prompt.lower() for keyword in allowed_keywords):
#             model = genai.GenerativeModel("gemini-2.0-flash")
#             response = model.generate_content(user_prompt)

#             if response and hasattr(response, "text"):
#                 assistant_response = response.text
#             else:
#                 assistant_response = "I'm sorry, I couldn't generate a response."

#             st.session_state.chat_history.append({"role": "assistant", "content": assistant_response})

#             # Display assistant's response
#             with st.chat_message("assistant"):
#                 st.markdown(assistant_response)
#         else:
#             # Restriction message
#             restriction_msg = "**⚠️ This chatbot only responds to health-related topics.**\nPlease ask about Diabetes, Asthma, Stroke, Cardiovascular Disease, or Mental Health."
#             st.session_state.chat_history.append({"role": "assistant", "content": restriction_msg})
            
#             with st.chat_message("assistant"):
#                 st.markdown(restriction_msg)


# if selected == 'Checkbox-to-disease-predictor':
# # Load transformer model
#     classifier = pipeline("zero-shot-classification", model="microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract")
#     # Use a pipeline as a high-level helper

#     # from transformers import pipeline

#     # pipe = pipeline("fill-mask", model="microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext")


#     # Define symptoms for each disease
#     diseases = {
#         "Diabetes": ["Frequent urination", "Increased thirst", "Unexplained weight loss", "Fatigue", "Blurred vision"],
#         "Hypertension": ["Headache", "Dizziness", "Chest pain", "Shortness of breath", "Nosebleeds"],
#         "Obesity": ["Excess body fat", "Breathlessness", "Joint pain", "Increased sweating", "Low energy levels"],
#         "Cardiovascular Disease": ["Chest pain", "Shortness of breath", "Dizziness", "Irregular heartbeat", "Fatigue"],
#         "COPD": ["Chronic cough", "Shortness of breath", "Wheezing", "Chest tightness", "Frequent respiratory infections"],
#         "Liver Disease": ["Jaundice", "Abdominal pain", "Swelling in legs", "Chronic fatigue", "Nausea"],
#         "Kidney Disease": ["Swelling in legs", "Fatigue", "Loss of appetite", "Changes in urination", "Muscle cramps"],
#         "Metabolic Syndrome": ["High blood sugar", "High blood pressure", "Increased waist size", "High cholesterol", "Fatigue"],
#         "Osteoarthritis": ["Joint pain", "Stiffness", "Swelling", "Reduced flexibility", "Bone spurs"],
#         "Gastroesophageal Reflux Disease": ["Heartburn", "Acid reflux", "Difficulty swallowing", "Chronic cough", "Sore throat"],
#         "Depression": ["Persistent sadness", "Loss of interest", "Sleep disturbances", "Fatigue", "Difficulty concentrating"],
#         "Sleep Apnea": ["Loud snoring", "Pauses in breathing", "Daytime drowsiness", "Morning headaches", "Irritability"],
#     }

#     # Streamlit UI
#     st.title("🩺 Hybrid Symptom Checker")
#     st.write("Select your symptoms and get AI-powered predictions!")

#     selected_symptoms = []

#     # Create symptom selection with markdown separation and three columns
#     disease_keys = list(diseases.keys())

#     for i in range(0, len(disease_keys), 3):
#         cols = st.columns(3)
#         for j in range(3):
#             if i + j < len(disease_keys):
#                 disease = disease_keys[i + j]
#                 with cols[j]:
#                     st.markdown(f"### {disease}")
#                     for symptom in diseases[disease]:
#                         if st.checkbox(symptom, key=f"{disease}_{symptom}"):
#                             selected_symptoms.append(symptom)

#     if st.button("🔍 Predict Disease"):
#         if selected_symptoms:
#             user_input = ", ".join(selected_symptoms)  # Convert symptoms to text

#             # 1️⃣ Custom Symptom Matching Approach
#             disease_scores = {disease: 0 for disease in diseases.keys()}
#             for disease, symptoms in diseases.items():
#                 matches = sum(symptom in selected_symptoms for symptom in symptoms)
#                 disease_scores[disease] = matches / len(symptoms)  # Normalize by symptom count

#             # Normalize to percentage
#             symptom_match_scores = {d: round(score * 100, 2) for d, score in disease_scores.items()}

#             # 2️⃣ AI Model Prediction
#             ai_results = classifier(user_input, list(diseases.keys()))
#             ai_scores = {ai_results["labels"][i]: round(ai_results["scores"][i] * 100, 2) for i in range(len(ai_results["labels"]))}

#             # 3️⃣ Hybrid Score Calculation (Average of Both Scores)
#             final_scores = {}
#             for disease in diseases.keys():
#                 symptom_score = symptom_match_scores.get(disease, 0)
#                 ai_score = ai_scores.get(disease, 0)
#                 final_scores[disease] = round((symptom_score + ai_score) / 2, 2)  # Averaging

#             # Sort by final score
#             sorted_final_scores = sorted(final_scores.items(), key=lambda x: x[1], reverse=True)

#             # Display results
#             st.write("### 🔬 Possible Conditions (Hybrid Model Prediction):")
#             for disease, score in sorted_final_scores:
#                 if score > 0:
#                     st.write(f"🩺 {disease}: {score}% match")
#         else:
#             st.write("⚠️ Please select at least one symptom.")


import torch
from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer

import os
from huggingface_hub import login

login(token=os.environ.get("HF_TOKEN"))

if selected == 'Mental-Analysis':
    # Load the Hugging Face model
    model_name = "mental/mental-roberta-base"
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForSequenceClassification.from_pretrained(model_name)

    # Sidebar with title and markdown
    st.sidebar.title("🧠 Mental Health Analysis")
    st.sidebar.markdown("""
    Analyze mental health symptoms using a **pre-trained AI model**.  
    This tool predicts **Depression and Anxiety** based on text input.
    """)

    # Main content
    st.title("🔬 Mental Health Text Analysis")
    st.markdown("Enter a description of your mental state, and the AI will predict possible conditions.")

    # User input
    user_input = st.text_area("Describe your symptoms (e.g., 'I feel hopeless and anxious all the time.'):")
    
    if st.button("Analyze"):
        if user_input:
            # Tokenize input
            inputs = tokenizer(user_input, return_tensors="pt", truncation=True, padding=True)

            # Get raw logits from the model
            with torch.no_grad():
                outputs = model(**inputs)
            logits = outputs.logits

            # Apply sigmoid activation to get independent probabilities
            probs = torch.sigmoid(logits).squeeze().tolist()

            # Map to labels
            label_mapping = {
                0: "Depression",
                1: "Anxiety"
            }
            predictions = {label_mapping[i]: round(probs[i] * 100, 2) for i in range(len(probs))}

            # Display predictions
            st.write("### Predictions:")
            for label, score in predictions.items():
                st.write(f"🩺 **{label}**: {score}% confidence")

            # Sort for better visualization
            sorted_labels = sorted(predictions.keys(), key=lambda x: predictions[x], reverse=True)
            sorted_scores = [predictions[label] for label in sorted_labels]

            # Plot using Seaborn
            fig, ax = plt.subplots(figsize=(4, 2.5))  # Compact size
            sns.barplot(x=sorted_scores, y=sorted_labels, palette="coolwarm", ax=ax)

            # Labels & title
            ax.set_xlabel("Risk Probability (%)")
            ax.set_title("Mental Health Risk Assessment")
            ax.set_xlim(0, 100)
            
            # Add percentages inside bars
            for i, (score, label) in enumerate(zip(sorted_scores, sorted_labels)):
                ax.text(score - 5, i, f"{score}%", va='center', ha='right', color='white', fontsize=10, fontweight='bold')

            # Display the chart in a single column
            st.pyplot(fig)


if selected == 'Sleep Health Analysis':
    st.title("🌙 Sleep Health Analysis")
    st.image("https://cdn-icons-png.flaticon.com/512/1205/1205526.png", width=100)
    
    st.markdown("""
    This model predicts the likelihood of **Sleep Disorders** based on various health factors.  
    Enter your details and click **"Sleep Health Test Result"** to get the prediction.
    """)

    # Load models
    try:
        sleep_model = pickle.load(open('sleep_health/svc_model.pkl', 'rb'))
        scaler = pickle.load(open('sleep_health/scaler.pkl', 'rb'))
        label_encoder = pickle.load(open('sleep_health/label_encoders.pkl', 'rb'))
    except FileNotFoundError:
        st.error("Error: Model files not found. Please upload the model files.")
        st.stop()

    # Input fields for user data
    col1, col2 = st.columns(2)

    with col1:
        gender = st.selectbox('Gender', ['Male', 'Female'], key='gender_sleep')
        age = st.slider("Age", min_value=27, max_value=59, value=35, 
                      help="Age range in dataset: 27-59 years", key='age_sleep')
        occupation = st.selectbox("Occupation", 
            ['Software Engineer', 'Doctor', 'Sales Representative', 'Teacher', 'Business',
             'Scientist', 'Accountant', 'Engineer'], key='occupation_sleep')
        sleep_duration = st.slider("Sleep Duration (hours)", 
            min_value=5.8, max_value=8.5, value=6.5, step=0.1,
            help="Dataset range: 5.8-8.5 hours", key='sleep_duration')
        quality_of_sleep = st.slider('Quality of Sleep', 
            min_value=4, max_value=9, value=6,
            help="Higher is better (4-9 scale)", key='quality_sleep')
        physical_activity_level = st.slider('Physical Activity Level (minutes/day)', 
            min_value=30, max_value=90, value=45,
            help="Physical activity in minutes per day", key='activity_sleep')

    with col2:
        stress_level = st.slider('Stress Level', 
            min_value=3, max_value=8, value=6, 
            help="Higher values indicate higher stress (3-8 scale)", key='stress_sleep')
        bmi_category = st.selectbox("BMI Category", 
            ["Normal", "Overweight", "Obese"], key='bmi_sleep')
        
        # For blood pressure, let's use two separate inputs for systolic/diastolic
        col2a, col2b = st.columns(2)
        with col2a:
            systolic = st.slider("Blood Pressure (Systolic)", 
                min_value=110, max_value=140, value=125, key='bp_sys_sleep')
        with col2b:
            diastolic = st.slider("Diastolic", 
                min_value=70, max_value=95, value=80, key='bp_dia_sleep')
        
        blood_pressure = f"{systolic}/{diastolic}"
        
        heart_rate = st.slider("Heart Rate (bpm)", 
            min_value=65, max_value=86, value=75, 
            help="Normal range: 60-100 bpm", key='hr_sleep')
        daily_steps = st.slider("Daily Steps", 
            min_value=3000, max_value=10000, value=6000, step=500,
            help="Recommended: 7,000-10,000 steps/day", key='steps_sleep')

    # Create a button to trigger prediction
    if st.button('Sleep Health Test Result', key='sleep_test_button'):
        try:
            # Prepare input data
            input_data = {
                'Gender': gender,
                'Age': age,
                'Occupation': occupation,
                'Sleep Duration': sleep_duration,
                'Quality of Sleep': quality_of_sleep,
                'Physical Activity Level': physical_activity_level,
                'Stress Level': stress_level,
                'BMI Category': bmi_category,
                'Blood Pressure': blood_pressure,
                'Heart Rate': heart_rate,
                'Daily Steps': daily_steps
            }
            
            # Process and predict
            df = pd.DataFrame([input_data])
            
            # Apply label encoding to categorical features
            for col, encoder in label_encoder.items():
                if col in df.columns:
                    df[col] = encoder.transform([df[col].iloc[0]])[0]
            
            # Feature Engineering and Preprocessing
            columns_to_drop = ["Physical Activity Level", "Person ID"]
            for col in columns_to_drop:
                if col in df.columns:
                    df = df.drop(columns=[col])
            
            # Create dummy variables with the correct column names expected by the model
            df = pd.get_dummies(df)
            
            # Get the expected feature names from the model
            # You might need to store these feature names during training
            expected_features = [
                'Age', 'Sleep Duration', 'Quality of Sleep', 'Stress Level',
                'Heart Rate', 'Daily Steps', 'Gender_Female', 'Gender_Male',
                'Occupation_Accountant', 'Occupation_Business', 'Occupation_Doctor',
                'Occupation_Engineer', 'Occupation_Sales Representative', 
                'Occupation_Scientist', 'Occupation_Software Engineer', 'Occupation_Teacher',
                'BMI Category_Normal', 'BMI Category_Obese', 'BMI Category_Overweight',
                'Blood Pressure_110/70', 'Blood Pressure_120/80', 'Blood Pressure_125/80',
                'Blood Pressure_130/85', 'Blood Pressure_140/90'
            ]
            
            # Create a DataFrame with expected features filled with zeros
            prediction_df = pd.DataFrame(0, index=[0], columns=expected_features)
            
            # Fill in the values from our current DataFrame
            for col in df.columns:
                if col in prediction_df.columns:
                    prediction_df[col] = df[col].values
                    
            # Scale with proper error handling
            with st.spinner("⏳ Predicting... Please wait..."):
                time.sleep(2)
                # Use the properly formatted DataFrame
                prediction = sleep_model.predict(prediction_df)
            
            # Display result
            result = "🛑 High risk of sleep disorder" if prediction[0] == 1 else "✅ Low risk of sleep disorder"
            if prediction[0] == 0:
                st.balloons()
            st.success(result)
            
            # Show risk factors based on input
            st.subheader("Risk Factor Analysis")
            risk_factors = []
            
            if sleep_duration < 6.0:
                risk_factors.append("⚠️ Low sleep duration (less than 6 hours)")
            if quality_of_sleep < 6:
                risk_factors.append("⚠️ Poor sleep quality")
            if stress_level > 6:
                risk_factors.append("⚠️ High stress levels")
            if bmi_category in ["Overweight", "Obese"]:
                risk_factors.append(f"⚠️ {bmi_category} BMI category")
            if int(systolic) > 130 or int(diastolic) > 85:
                risk_factors.append("⚠️ Elevated blood pressure")
            if heart_rate > 80:
                risk_factors.append("⚠️ Elevated heart rate")
            if daily_steps < 5000:
                risk_factors.append("⚠️ Low daily activity (steps)")
                
            if risk_factors:
                st.markdown("##### Potential Risk Factors:")
                for factor in risk_factors:
                    st.markdown(factor)
            else:
                st.markdown("✅ No significant risk factors identified.")

        except Exception as e:
            st.error(f"❌ Error: {e}")




if selected=='Hypertension Prediction':
    st.title("Hypertension Risk Prediction App")
    st.markdown("This application uses an Extra Trees Classifier model to predict hypertension risk based on patient health data.")

    # Load the model and scaler
    try:
        hypertension_model = pickle.load(open('hypertension/extratrees_model.pkl', 'rb'))
        hypertension_scaler = pickle.load(open('hypertension/scaler.pkl', 'rb'))
        st.success("Model and scaler loaded successfully!")
    except Exception as e:
        st.error(f"Error loading model or scaler: {e}")
        st.info("Please check that model and scaler files are in the correct location.")
        st.warning("Expected path: 'hypertension/extratrees_model.pkl' and 'hypertension/scaler.pkl'")

    # Define input section
    st.subheader("Patient Information")

    # Create two columns for input layout
    col1, col2 = st.columns(2)

    with col1:
        male = st.radio("Gender", options=[0, 1], format_func=lambda x: "Female" if x == 0 else "Male")
        age = st.slider("Age", min_value=32, max_value=70, value=49, help="Patient's age (32-70 years)")
        cigs_per_day = st.slider("Cigarettes Per Day", min_value=0.0, max_value=70.0, value=0.0, step=1.0)
        bp_meds = st.radio("On Blood Pressure Medication", options=[0.0, 1.0], format_func=lambda x: "No" if x == 0.0 else "Yes")
        tot_chol = st.slider("Total Cholesterol", min_value=107.0, max_value=500.0, value=234.0, step=1.0, help="mg/dL")

    with col2:
        sys_bp = st.slider("Systolic Blood Pressure", min_value=83.5, max_value=295.0, value=128.0, step=0.5, help="mmHg")
        dia_bp = st.slider("Diastolic Blood Pressure", min_value=48.0, max_value=142.5, value=82.0, step=0.5, help="mmHg")
        bmi = st.slider("BMI", min_value=15.54, max_value=56.80, value=25.40, step=0.01)
        heart_rate = st.slider("Heart Rate", min_value=44.0, max_value=143.0, value=75.0, step=1.0, help="beats per minute")
        glucose = st.slider("Glucose", min_value=40.0, max_value=394.0, value=78.0, step=1.0, help="mg/dL")

    # Prediction button
    predict_button = st.button("Predict Hypertension Risk")

    if predict_button:
        # Create input dataframe
        input_data = pd.DataFrame({
            'male': [male],
            'age': [age],
            'cigsPerDay': [cigs_per_day],
            'BPMeds': [bp_meds],
            'totChol': [tot_chol],
            'sysBP': [sys_bp],
            'diaBP': [dia_bp],
            'BMI': [bmi],
            'heartRate': [heart_rate],
            'glucose': [glucose]
        })
        
        # Display input data
        st.subheader("Input Data:")
        st.dataframe(input_data)
        
        # Identify numerical columns to scale
        num_cols = ['age', 'cigsPerDay', 'totChol', 'sysBP', 'diaBP', 'BMI', 'heartRate', 'glucose']
        
        try:
            # Scale the numerical features
            input_data[num_cols] = hypertension_scaler.transform(input_data[num_cols])
            
            # Make prediction
            prediction = hypertension_model.predict(input_data)[0]
            prediction_prob = hypertension_model.predict_proba(input_data)[0]
            
            # Display prediction results
            st.subheader("Prediction Result:")
            
            # Create columns for results
            res_col1, res_col2 = st.columns(2)
            
            with res_col1:
                if prediction == 0:
                    st.success("✅ Low Risk of Hypertension")
                else:
                    st.error("🚨 High Risk of Hypertension")
            
            with res_col2:
                # Visualization
                st.write(f"Probability of Low Risk: {prediction_prob[0]:.2f}")
                st.write(f"Probability of High Risk: {prediction_prob[1]:.2f}")
                
                # Add progress bar
                st.progress(float(prediction_prob[1]))
        
        except Exception as e:
            st.error(f"Error during prediction: {e}")
        # st.info("Please check that all inputs are valid and within the expected ranges.")


if selected == 'Medical Consultant':
    st.title("🩺 Medical Consultant Chatbot")
    st.markdown("### Discuss Your Health Concerns with Our AI-powered Chatbot")
    st.write("Our AI can help with **medical questions, symptom analysis, and health recommendations**.")
    
    # Initialize API
    genai.configure(api_key="AIzaSyAcXexC7cNXrRTCYj6Dg7ZFYVQZH8a5PMw")  # Replace with your actual API key

    # Custom Styling for suggestions
    st.markdown("""
        <style>
            .prompt-box { 
                background-color: #222222; 
                padding: 12px; 
                border-radius: 8px; 
                font-size: 14px; 
                font-family: sans-serif;
                margin-bottom: 10px;
                border: 1px solid #444444;
                text-align: center;
                cursor: pointer;
            }
            .prompt-box:hover {
                background-color: #333333;
            }
        </style>
    """, unsafe_allow_html=True)
    
    # Common medical questions as suggestions
    st.markdown("#### 💡 Common Health Questions")
    
    prompt_options = [
        ("Diabetes", "What are early warning signs of diabetes?"),
        ("Hypertension", "How can I manage my blood pressure naturally?"),
        ("Heart Health", "What lifestyle changes help reduce cardiovascular risk?"),
        ("Asthma", "What triggers asthma attacks and how can I prevent them?"),
        ("Stroke", "What are the warning signs of a stroke?"),
        ("Sleep Health", "How does poor sleep affect my overall health?"),
        ("Mental Health", "What are common symptoms of anxiety?"),
        ("Preventive Care", "What preventive screenings should I get at my age?"),
        ("Exercise", "How much exercise do I need for good health?"),
        ("Nutrition", "What diet changes can improve my heart health?")
    ]
    
    # Display prompts in two columns
    cols = st.columns(2)
    for i in range(0, len(prompt_options), 2):
        with cols[0]: 
            if i < len(prompt_options):
                label, prompt = prompt_options[i]
                st.markdown(f"""<div class="prompt-box" onclick="document.querySelector('#medical-chat-input').value='{prompt}';"><strong>{label}</strong><br>{prompt}</div>""", unsafe_allow_html=True)
        
        with cols[1]: 
            if i+1 < len(prompt_options):
                label, prompt = prompt_options[i+1]
                st.markdown(f"""<div class="prompt-box" onclick="document.querySelector('#medical-chat-input').value='{prompt}';"><strong>{label}</strong><br>{prompt}</div>""", unsafe_allow_html=True)
    
    # Initialize chat history if not present
    if "medical_chat_history" not in st.session_state:
        st.session_state.medical_chat_history = []
        # Add welcome message
        welcome_msg = {
            "role": "assistant", 
            "content": """👋 Welcome to your Medical Consultant! I can help answer questions about:

- Health concerns and symptoms
- Disease prevention and management
- Lifestyle recommendations
- Understanding medical conditions
            
How can I assist with your health questions today?"""
        }
        st.session_state.medical_chat_history.append(welcome_msg)
    
    # Chat container
    chat_container = st.container()
    with chat_container:
        # Display previous chat history
        for message in st.session_state.medical_chat_history:
            with st.chat_message(message["role"]):
                st.markdown(message["content"])
    
    # User input field
    user_prompt = st.chat_input("Ask about health concerns, symptoms, or lifestyle questions...", key="medical-chat-input")
    
    # Define medical topics for feature recommendations
    medical_topics = {
        "diabetes": "Diabetes Prediction",
        "blood sugar": "Diabetes Prediction",
        "hypertension": "Hypertension Prediction",
        "blood pressure": "Hypertension Prediction", 
        "heart": "Cardiovascular Disease Prediction",
        "cardiovascular": "Cardiovascular Disease Prediction",
        "asthma": "Asthma Prediction",
        "breathing": "Asthma Prediction",
        "stroke": "Stroke Prediction",
        "sleep": "Sleep Health Analysis",
        "insomnia": "Sleep Health Analysis",
        "mental health": "Mental-Analysis",
        "depression": "Mental-Analysis",
        "anxiety": "Mental-Analysis",
        "stress": "Mental-Analysis"
    }
    
    if user_prompt:
        # Add user message to chat
        st.session_state.medical_chat_history.append({"role": "user", "content": user_prompt})
        
        # Display user message in chat
        with st.chat_message("user"):
            st.markdown(user_prompt)
        
        try:
            # Create system instruction
            system_instruction = """You are a medical consultant chatbot designed to provide helpful health information.
            
RULES:
- Provide accurate, concise medical information based on current scientific understanding
- Answer questions about symptoms, diseases, prevention, and health management
- Keep responses informative but brief (under 150 words)
- When uncertain, acknowledge limitations and recommend consulting a healthcare professional
- Avoid making definitive diagnoses or treatment recommendations
- Never claim to be an AI or language model - respond directly as a medical consultant
- Always clarify that your advice is informational and not a substitute for professional medical care
- When describing medical conditions, focus on factual information about symptoms, risk factors, and prevention
- Maintain a professional, empathetic tone
- If the user mentions specific symptoms, acknowledge them and provide information about possible causes
- Respond in a doctor-like manner when assessing symptoms or risk factors
- Use your knowledge to identify if the user's query relates to any specific medical conditions
- Do not suggest our prediction tools in every response - only when truly relevant

The user is interacting with a health prediction platform that offers the following tools:
- Diabetes Prediction
- Hypertension Prediction 
- Cardiovascular Disease Prediction
- Asthma Prediction
- Stroke Prediction
- Sleep Health Analysis
- Mental Health Analysis

TASK: First, determine if the user's query contains symptoms or mentions specific health conditions. If so, provide a doctor-like assessment. Only if appropriate, subtly suggest one of our health prediction tools at the end of your response.
"""

            # Generate a response using Gemini
            model = genai.GenerativeModel("gemini-2.0-flash")
            
            # Prepare chat context
            chat_context = []
            for msg in st.session_state.medical_chat_history[-5:]:  # Last 5 messages for context
                if msg["role"] == "user":
                    chat_context.append(f"User: {msg['content']}")
                else:
                    chat_context.append(f"Medical Consultant: {msg['content']}")
            
            # Add current query with additional analysis request
            full_prompt = f"""{system_instruction}

CONVERSATION HISTORY:
{chr(10).join(chat_context)}

USER QUERY: {user_prompt}

ANALYSIS INSTRUCTIONS:
1. First, determine if this query relates to any specific health conditions or symptoms
2. Provide a helpful medical response addressing the user's concerns
3. If appropriate, subtly suggest one relevant prediction tool at the end of your response (only if truly related)
4. Remember to be professional and avoid making definitive diagnoses
"""
            
            # Generate response
            response = model.generate_content(full_prompt)
            
            if response and hasattr(response, "text"):
                assistant_response = response.text
            else:
                assistant_response = "I'm sorry, I couldn't generate a response. Please try asking a different health-related question."
            
            # Save and display response
            st.session_state.medical_chat_history.append({"role": "assistant", "content": assistant_response})
            
            with st.chat_message("assistant"):
                st.markdown(assistant_response)
            
        except Exception as e:
            error_msg = f"I apologize, but I'm having trouble processing your request right now. Please try again with a different question."
            st.session_state.medical_chat_history.append({"role": "assistant", "content": error_msg})
            
            with st.chat_message("assistant"):
                st.markdown(error_msg)
        
        # Force refresh to update the chat
        st.rerun()