app.py

"""
Streamlit Web Application for Agricultural Production Forecasting.
"""
import streamlit as st
import pandas as pd
import numpy as np
import os
import sys
import plotly.graph_objects as go
from PIL import Image

# Add src to path
sys.path.append(os.path.dirname(os.path.abspath(__file__)))

from src.data_loader import DataLoader
from src.feature_engineering import FeatureEngineer
from src.predictor import Predictor
from src.policy_advisor import PolicyAdvisor
from src.visualization import Visualizer

# Page Config
st.set_page_config(
    page_title="Kaduna Agricultural Forecasting",
    page_icon="🌾",
    layout="wide"
)

# Title and Header
st.title("🌾 Kaduna State Agricultural Forecasting System")
st.markdown("""
This system predicts future crop production levels to support food security planning and policy decisions.
""")

# Sidebar
st.sidebar.header("Configuration")

# Load Data
@st.cache_data
def load_dataset():
    loader = DataLoader("Kaduna_State_Agricultural_Production_Data.csv")
    loader.clean_data()
    return loader

try:
    loader = load_dataset()
    available_crops = loader.get_available_crops()
except Exception as e:
    st.error(f"Error loading data: {e}")
    st.stop()

# Crop Selection
selected_crop = st.sidebar.selectbox("Select Crop", available_crops, index=available_crops.index("Maize") if "Maize" in available_crops else 0)

# Model Selection
model_options = ['ARIMA', 'SARIMAX', 'XGBoost', 'LSTM', 'GRU']
selected_model = st.sidebar.selectbox("Select Model", model_options, index=0)

# Forecast Horizon
forecast_years = st.sidebar.slider("Forecast Horizon (Years)", 1, 5, 3)

# Run Button
if st.sidebar.button("Generate Forecast"):
    with st.spinner(f"Generating forecast for {selected_crop} using {selected_model}..."):
        try:
            # Get Data
            crop_df = loader.get_crop_data(selected_crop)
            
            # Initialize Components
            predictor = Predictor()
            advisor = PolicyAdvisor()
            viz = Visualizer()
            
            # Generate Forecast
            # Note: This assumes models are already trained and saved in outputs/models/
            # If not, we might need to trigger training or handle the error
            try:
                forecast_df = predictor.generate_forecast(
                    selected_crop, 
                    selected_model.lower(), 
                    years_ahead=forecast_years,
                    historical_data=crop_df
                )
            except FileNotFoundError:
                st.warning(f"Model {selected_model} for {selected_crop} not found. Please train the model first using the CLI or check available models.")
                st.info(f"Try running: `python main.py --crop \"{selected_crop}\" --model {selected_model.lower()} --train` in your terminal.")
                st.stop()
                
            # Layout
            col1, col2 = st.columns([2, 1])
            
            with col1:
                st.subheader("Production Forecast")
                
                # Plot
                fig = viz.plot_forecast_interactive(
                    crop_df, forecast_df, selected_crop, selected_model
                )
                st.plotly_chart(fig, use_container_width=True)
                
                # Data Table
                st.subheader("Forecast Data")
                st.dataframe(forecast_df)
                
            with col2:
                st.subheader("Policy Analysis")
                
                # Policy Analysis
                analysis = advisor.analyze_forecast(crop_df, forecast_df, selected_crop)
                recs = advisor.generate_recommendations(analysis)
                
                # Display Recommendations
                for rec in recs:
                    if "ALERT" in rec or "Shortage" in rec:
                        st.error(rec)
                    elif "NOTICE" in rec or "Oversupply" in rec:
                        st.warning(rec)
                    elif "Action" in rec:
                        st.info(rec)
                    else:
                        st.write(rec)
                        
                # Download Policy Brief
                brief_text = "\n".join(recs)
                st.download_button(
                    label="Download Policy Brief",
                    data=brief_text,
                    file_name=f"{selected_crop}_policy_brief.txt",
                    mime="text/plain"
                )

        except Exception as e:
            st.error(f"An error occurred: {e}")
            import traceback
            st.code(traceback.format_exc())

# Historical Data Tab
with st.expander("View Historical Data"):
    crop_df = loader.get_crop_data(selected_crop)
    st.dataframe(crop_df)
    
    # Simple stats
    st.write("Summary Statistics:")
    st.write(crop_df.describe())

# Footer
st.markdown("---")
st.markdown("Built with ❤️ for Kaduna State Ministry of Agriculture")