Update app.py

app.py

@@ -1,47 +1,45 @@
 import streamlit as st
 import pandas as pd
-import pickle
 from prophet import Prophet
 from datetime import datetime, timedelta
 import numpy as np
-import plotly.graph_objects as go
-
-# Load the trained models (optional, for initialization or fallback)
-@st.cache_resource
-def load_model():
-    try:
-        with open('model.pkl', 'rb') as f:
-            models = pickle.load(f)
-        return models
-    except FileNotFoundError:
-        return None
 
 # Prepare data for Prophet
 def prepare_prophet_data(usage_series):
-    # Create a date range for the last 60 days
     end_date = datetime.now()
     start_date = end_date - timedelta(days=len(usage_series) - 1)
     dates = [start_date + timedelta(days=i) for i in range(len(usage_series))]
-    
-    # Create Prophet-compatible DataFrame
     prophet_df = pd.DataFrame({
         'ds': dates,
         'y': usage_series
     })
+    prophet_df['cap'] = 60  # Max observed usage
+    prophet_df['floor'] = 0
     return prophet_df
 
 # Train or update Prophet model with user-provided usage series
 def train_model_with_usage(usage_series):
+    print("Training with changepoint_prior_scale=0.002, usage:", usage_series)  # Debug
     prophet_df = prepare_prophet_data(usage_series)
-    model = Prophet(yearly_seasonality=False, weekly_seasonality=True, daily_seasonality=True)
+    model = Prophet(
+        yearly_seasonality=False,
+        weekly_seasonality=True,
+        daily_seasonality=True,
+        changepoint_prior_scale=0.002,
+        growth='logistic'
+    )
     model.fit(prophet_df)
     return model
 
 # Function to make forecasts
 def make_forecast(model, periods):
     future = model.make_future_dataframe(periods=periods)
+    future['cap'] = 60
+    future['floor'] = 0
     forecast = model.predict(future)
-    return round(forecast['yhat'].tail(periods).sum())  # Round to nearest integer
+    daily_forecasts = forecast['yhat'].tail(periods).tolist()
+    print(f"Daily forecasts for {periods} days:", [round(y) for y in daily_forecasts])  # Debug
+    return round(sum(max(0, y) for y in daily_forecasts))  # Clip negative values
 
 # Function to validate input
 def validate_usage_series(usage_str):
@@ -58,110 +56,59 @@ def validate_usage_series(usage_str):
 # Main Streamlit app
 def main():
     st.title("SmartLab Consumables Forecast")
+    if st.button("Clear Cache"):
+        st.cache_resource.clear()
+        st.write("Cache cleared!")
     
-    # Load pre-trained models (optional, for reference)
-    models = load_model()
-    
-    # Input form
     st.header("Input Parameters")
     consumable_type = st.selectbox("Consumable Type", ['Filters', 'Reagents', 'Vials'])
     usage_series = st.text_input("Last 60 Days Usage (comma-separated)", "")
     current_stock = st.number_input("Current Stock", min_value=0, value=0)
     
     if st.button("Generate Forecast"):
-        # Validate inputs
         usage_list, error = validate_usage_series(usage_series)
         if error:
             st.error(error)
             return
+        st.write("Debug: Input usage series:", usage_list)  # Debug
         
-        # Train a new model with the user-provided usage series
         try:
             model = train_model_with_usage(usage_list)
         except Exception as e:
             st.error(f"Error training model: {str(e)}")
             return
         
-        # Forecast for 7, 14, and 30 days
         forecast_7 = make_forecast(model, 7)
         forecast_14 = make_forecast(model, 14)
         forecast_30 = make_forecast(model, 30)
         
-        # Display forecasts
         st.header("Forecast Results")
         st.write(f"**7-Day Forecast**: {forecast_7} units")
         st.write(f"**14-Day Forecast**: {forecast_14} units")
         st.write(f"**30-Day Forecast**: {forecast_30} units")
         
-        # Threshold alerting
         st.header("Threshold Alerts")
         if current_stock < forecast_7:
             st.warning(f"Alert: Current stock ({current_stock}) is below 7-day forecast ({forecast_7}). 🚩")
+        else:
+            st.write("No alert for 7-day forecast.")
         if current_stock < forecast_14:
             st.warning(f"Alert: Current stock ({current_stock}) is below 14-day forecast ({forecast_14}). 🚩")
+        else:
+            st.write("No alert for 14-day forecast.")
         if current_stock < forecast_30:
             st.warning(f"Alert: Current stock ({current_stock}) is below 30-day forecast ({forecast_30}). 🚩")
+        else:
+            st.write("No alert for 30-day forecast.")
         
-        # Order suggestions
         st.header("Order Suggestions")
-        order_7 = max(0, round(forecast_7 - current_stock))  # Round to nearest integer
-        order_14 = max(0, round(forecast_14 - current_stock))  # Round to nearest integer
-        order_30 = max(0, round(forecast_30 - current_stock))  # Round to nearest integer
+        order_7 = max(0, round(forecast_7 - current_stock))
+        order_14 = max(0, round(forecast_14 - current_stock))
+        order_30 = max(0, round(forecast_30 - current_stock))
         
         st.write(f"**For 7 Days**: Order {order_7} additional units.")
         st.write(f"**For 14 Days**: Order {order_14} additional units.")
         st.write(f"**For 30 Days**: Order {order_30} additional units.")
-        
-        # Graphical representation for forecast
-        st.header("Forecast Visualization")
-        forecast_data = pd.DataFrame({
-            'Period': ['7 Days', '14 Days', '30 Days'],
-            'Units': [forecast_7, forecast_14, forecast_30]
-        })
-        fig_forecast = go.Figure()
-        fig_forecast.add_trace(go.Scatter(
-            x=forecast_data['Period'],
-            y=forecast_data['Units'],
-            mode='lines+markers',
-            name='Forecasted Units',
-            line=dict(color='blue'),
-            marker=dict(size=10)
-        ))
-        fig_forecast.update_layout(
-            title='Consumable Usage Forecast',
-            xaxis_title='Time Period',
-            yaxis_title='Units',
-            template='plotly_white'
-        )
-        st.plotly_chart(fig_forecast)
-        
-        # Graphical representation for threshold alerts
-        st.header("Threshold Alerts Visualization")
-        alert_data = pd.DataFrame({
-            'Category': ['Current Stock', '7-Day Forecast', '14-Day Forecast', '30-Day Forecast'],
-            'Units': [current_stock, forecast_7, forecast_14, forecast_30],
-            'Alert': [
-                False,
-                current_stock < forecast_7,
-                current_stock < forecast_14,
-                current_stock < forecast_30
-            ]
-        })
-        fig_alerts = go.Figure()
-        fig_alerts.add_trace(go.Bar(
-            x=alert_data['Category'],
-            y=alert_data['Units'],
-            marker_color=['green'] + ['red' if alert else 'blue' for alert in alert_data['Alert'][1:]],
-            text=[f"🚩" if alert else "" for alert in alert_data['Alert']],
-            textposition='auto'
-        ))
-        fig_alerts.update_layout(
-            title='Stock vs Forecast with Alerts (🚩 indicates low stock)',
-            xaxis_title='Category',
-            yaxis_title='Units',
-            template='plotly_white'
-        )
-        st.plotly_chart(fig_alerts)
 
 if __name__ == "__main__":
-    main()  
+    main()