app.py

import streamlit as st
import pandas as pd
import re
from config import MODEL_PATH, ENCODER_DIR, OPENAI_API_KEY, OPENAI_BASE_URL
from utils import load_model, load_label_encoders
from prediction import predict_susceptibility
from ai_assistant import initialize_openai_client, get_ai_response

# Load assets
model = load_model(MODEL_PATH)
encoders = load_label_encoders(ENCODER_DIR)

# Initialize OpenAI client
client = initialize_openai_client(OPENAI_API_KEY, OPENAI_BASE_URL)

# Streamlit UI
st.set_page_config(page_title="Microbial Susceptibility Analyzer", layout="wide")

st.sidebar.title("Navigation")
page = st.sidebar.radio("Go to", ["Home", "Susceptibility Analysis", "Data Upload", "About"])

# Home Page
if page == "Home":
    st.title("Microbial Susceptibility Analyzer")
    st.image("bacteria.jpeg", use_container_width=True)
    st.markdown("""
    **Welcome to the Microbial Susceptibility Analyzer!**  
    This app helps analyze **antibiotic resistance** using **machine learning and rule-based decisions**.
    - Predict microbial susceptibility.
    - Ask an AI assistant for expert advice.
    - Upload datasets for batch predictions.
    """)

# Susceptibility Analysis Page
elif page == "Susceptibility Analysis":
    st.title("Susceptibility Prediction")
    
    # Initialize session state for messages if not exists
    if 'messages' not in st.session_state:
        st.session_state.messages = []
    
    # Create two columns for layout
    col1, col2 = st.columns([1, 1])
    
    with col1:
        with st.form("prediction_form"):
            organism = st.selectbox('Organism', options=encoders['organism'].keys())
            antibiotic = st.selectbox('Antibiotic', options=encoders['antibiotic'].keys())
            was_positive = st.selectbox('Was Positive', options=[1, 0])
            submit_button = st.form_submit_button("Predict")
        
        if submit_button:
            # Store inputs in session state
            st.session_state['current_organism'] = organism
            st.session_state['current_antibiotic'] = antibiotic
            st.session_state['current_was_positive'] = was_positive
            
            result = predict_susceptibility({
                'was_positive': was_positive,
                'organism': organism,
                'antibiotic': antibiotic
            }, model, encoders)
            
            st.subheader("Prediction Results")
            if "Error" in result:
                st.error(result["Error"])
            else:
                st.write(f"**Final Decision:** {result['Final Output']}")
                st.write(f"**Rule-Based Guidance:** {result['Rule Guidance']}")
                st.write(f"**Model Prediction:** {result['Model Prediction']}")
                st.write(f"**Decision Explanation:** {result['Decision Reason']}")
                
                # Clear previous messages when new prediction is made
                st.session_state.messages = []
    
    with col2:
        st.subheader("DeepSeek AI Assistant")
        
        # Only show assistant if a prediction has been made
        if 'current_organism' in st.session_state:
            st.markdown(f"Ask about **{st.session_state.get('current_organism')}** and **{st.session_state.get('current_antibiotic')}**:")
            
            # Example prompts as buttons
            example_prompts = [
                "Explain why this combination might show resistance",
                "Suggest alternative antibiotics for this organism",
                "What resistance mechanisms are common here?",
                "How should this result influence treatment decisions?"
            ]
            
            # Create a unique key for each button
            for i, prompt in enumerate(example_prompts):
                if st.button(prompt, key=f"prompt_{i}"):
                    # Create context-enhanced prompt
                    enhanced_prompt = f"For organism {st.session_state.get('current_organism')} " \
                                     f"and antibiotic {st.session_state.get('current_antibiotic')}: {prompt}"
                    response = get_ai_response(client, enhanced_prompt)
                    
                    # Display the user prompt and AI response
                    st.chat_message("user").markdown(prompt)
                    st.chat_message("assistant").markdown(response)
            
            # Chat input
            user_prompt = st.chat_input("Ask about this result...")
            if user_prompt:
                # Create context-enhanced prompt
                enhanced_prompt = f"For organism {st.session_state.get('current_organism')} " \
                                 f"and antibiotic {st.session_state.get('current_antibiotic')}: {user_prompt}"
                response = get_ai_response(client, enhanced_prompt)
                
                # Display the user prompt and AI response
                st.chat_message("user").markdown(user_prompt)
                st.chat_message("assistant").markdown(response)
        else:
            st.info("Make a prediction first to get specific AI assistance")

# Data Upload Page
elif page == "Data Upload":
    st.title("Batch Prediction: Upload CSV")
    uploaded_file = st.file_uploader("Upload your CSV file", type=["csv"])
    
    if uploaded_file:
        df = pd.read_csv(uploaded_file)
        st.write("Uploaded Data Preview:", df.head())
        
        if st.button("Predict for Dataset"):
            df["Prediction"] = df.apply(lambda row: predict_susceptibility(row.to_dict(), model, encoders)["Final Output"], axis=1)
            st.write("Prediction Results", df)

# About Page
elif page == "About":
    st.title("About this App")
    st.markdown("""
    - Developed by **Okunromade Joseph Oluwaseun**  
    - Uses **Machine Learning & Rule-based AI**  
    - Integrated with **DeepSeek AI** for advanced queries  
    - Matric No: 22/SCI01/172
    """)