Update app.py

app.py

@@ -10,59 +10,46 @@ warnings.filterwarnings("ignore")
 # ===================================================================
 # 1. LOAD THE TRAINED MODEL
 # ===================================================================
-# Make sure your trained model file is in the same directory
 MODEL_PATH = "glucose_predictor_model.pkl" 
 try:
     with open(MODEL_PATH, 'rb') as file:
         model = pickle.load(file)
     print("✅ Model loaded successfully!")
 except FileNotFoundError:
-    print(f"❌ CRITICAL ERROR: Model file '{MODEL_PATH}' not found. Please upload it to your Hugging Face Space.")
+    print(f"❌ CRITICAL ERROR: Model file '{MODEL_PATH}' not found.")
     model = None
 
 # ===================================================================
-# 2. SIMULATION AND PREDICTION LOGIC
+# 2. CORE PREDICTION LOGIC
 # ===================================================================
 
-def predict_glucose(current_glucose):
+def predict_glucose(current_glucose, hba1c, gender, hr_mean, hrv_rmssd, temp_mean):
     """
-    This is the core function. It takes only the current glucose, simulates
-    all other features, and returns the model's prediction.
+    This is the core prediction engine. It takes the primary inputs from the UI,
+    simulates the rest, engineers all features, and returns a prediction.
     """
     if model is None:
-        return "Model not loaded. Check file path.", "Error: Model file not found."
+        return "Model not loaded. Check file path."
 
-    # --- a) Simulate Demographics ---
-    # Randomly select a gender and a realistic HbA1c value
-    simulated_gender = np.random.choice(["Male", "Female"])
-    simulated_hba1c = np.round(np.random.uniform(5.5, 9.0), 1)
-
-    # --- b) Simulate a realistic 1-hour glucose history ---
-    history_size = 12 # 60 minutes of 5-min readings
-    trend = np.random.uniform(-2, 2) # Simulate a slight trend
+    # --- a) Simulate a realistic 1-hour glucose history ---
+    history_size = 12
+    trend = np.random.uniform(-2, 2)
     noise = np.random.normal(0, 3, history_size)
     glucose_history = np.linspace(current_glucose - trend * history_size, current_glucose, history_size) + noise
     glucose_history = np.round(glucose_history).astype(int)
 
-    # --- c) Simulate realistic wearable data ---
-    hr_mean = np.random.uniform(60, 90)
+    # --- b) Simulate the standard deviations and accelerometer data ---
     hr_std = np.random.uniform(2, 8)
-    temp_mean = np.random.uniform(32.0, 35.5)
     temp_std = np.random.uniform(0.05, 0.25)
-    acc_x_mean = np.random.uniform(-35, -5)
-    acc_y_mean = np.random.uniform(5, 35)
-    acc_z_mean = np.random.uniform(25, 55)
+    acc_x_mean, acc_y_mean, acc_z_mean = np.random.uniform(-35, -5), np.random.uniform(5, 35), np.random.uniform(25, 55)
     acc_std = np.random.uniform(1.0, 4.0)
-    hrv_rmssd = np.random.uniform(25, 60) # Typical RMSSD values in ms
 
-    # --- d) Assemble all features into a dictionary ---
+    # --- c) Assemble all features into a dictionary ---
     features = {
-        'glucose': current_glucose,
-        'hr_mean_5min': hr_mean, 'hr_std_5min': hr_std,
+        'glucose': current_glucose, 'hr_mean_5min': hr_mean, 'hr_std_5min': hr_std,
         'temp_mean_5min': temp_mean, 'temp_std_5min': temp_std,
-        'acc_x_mean_5min': acc_x_mean, 'acc_x_std_5min': acc_std,
-        'acc_y_mean_5min': acc_y_mean, 'acc_y_std_5min': acc_std,
-        'acc_z_mean_5min': acc_z_mean, 'acc_z_std_5min': acc_std,
+        'acc_x_mean_5min': acc_x_mean, 'acc_x_std_5min': acc_std, 'acc_y_mean_5min': acc_y_mean,
+        'acc_y_std_5min': acc_std, 'acc_z_mean_5min': acc_z_mean, 'acc_z_std_5min': acc_std,
         'glucose_lag_5min': glucose_history[-2], 'glucose_lag_15min': glucose_history[-4],
         'glucose_lag_30min': glucose_history[-7], 'glucose_lag_60min': glucose_history[0],
         'glucose_rolling_mean_15min': np.mean(glucose_history[-3:]),
@@ -72,83 +59,98 @@ def predict_glucose(current_glucose):
         'glucose_rolling_mean_60min': np.mean(glucose_history),
         'glucose_rolling_std_60min': np.std(glucose_history),
         'glucose_roc': current_glucose - glucose_history[-2],
-        'hba1c': simulated_hba1c,
-        'gender_is_female': 1 if simulated_gender == "Female" else 0,
+        'hba1c': hba1c, 'gender_is_female': 1 if gender == "Female" else 0,
         'hrv_rmssd_5min': hrv_rmssd
     }
 
-    # --- e) Create the final DataFrame for prediction ---
-    # The order of columns MUST match the order used during training.
-    # It's robust to use the model's expected feature names if available.
+    # --- d) Create the final DataFrame for prediction ---
     try:
         feature_order = model.get_booster().feature_names
-    except AttributeError: # Fallback for some model versions
-        feature_order = list(X.columns) # Assumes X is defined from training script
+    except AttributeError:
+        # Fallback if the feature names aren't in the model object
+        feature_order = ['glucose', 'hr_mean_5min', 'hr_std_5min', 'temp_mean_5min', 'temp_std_5min', 'acc_x_mean_5min', 'acc_x_std_5min', 'acc_y_mean_5min', 'acc_y_std_5min', 'acc_z_mean_5min', 'acc_z_std_5min', 'glucose_lag_5min', 'glucose_lag_15min', 'glucose_lag_30min', 'glucose_lag_60min', 'glucose_rolling_mean_15min', 'glucose_rolling_std_15min', 'glucose_rolling_mean_30min', 'glucose_rolling_std_30min', 'glucose_rolling_mean_60min', 'glucose_rolling_std_60min', 'glucose_roc', 'hba1c', 'gender_is_female', 'hrv_rmssd_5min']
 
     input_df = pd.DataFrame([features])[feature_order]
 
-    # --- f) Make Prediction ---
+    # --- e) Make Prediction ---
     prediction = model.predict(input_df)[0]
-    
-    # --- g) Format the output for display ---
-    prediction_text = f"{prediction:.0f} mg/dL"
-    status_text = (
-        f"Simulated Inputs:\n"
-        f"  - Gender: {simulated_gender}, HbA1c: {simulated_hba1c}%\n"
-        f"  - Heart Rate: {hr_mean:.1f} bpm, HRV (RMSSD): {hrv_rmssd:.1f} ms\n"
-        f"  - Recent Trend (ROC): {features['glucose_roc']} mg/dL/5min"
-    )
-    
-    return prediction_text, status_text
+    return f"{prediction:.0f} mg/dL"
+
+# ===================================================================
+# 3. RANDOM SCENARIO GENERATOR
+# ===================================================================
 
-def generate_sample_input():
-    """Generates a random glucose value to populate the input slider."""
-    return np.random.randint(90, 180)
+def generate_random_scenario_and_predict():
+    """Generates random values for all inputs and triggers a prediction."""
+    # Generate random values
+    random_glucose = np.random.randint(70, 250)
+    random_hba1c = np.round(np.random.uniform(5.5, 10.0), 1)
+    random_gender = np.random.choice(["Male", "Female"])
+    random_hr = np.random.randint(60, 100)
+    random_hrv = np.random.randint(20, 70)
+    random_temp = np.round(np.random.uniform(32.0, 36.0), 1)
+
+    # Get a prediction based on these random values
+    prediction = predict_glucose(random_glucose, random_hba1c, random_gender, random_hr, random_hrv, random_temp)
+    
+    # Return all the generated values to update the UI
+    return random_glucose, random_hba1c, random_gender, random_hr, random_hrv, random_temp, prediction
 
 # ===================================================================
-# 3. CREATE THE GRADIO INTERFACE
+# 4. CREATE THE GRADIO INTERFACE
 # ===================================================================
 
-with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue", secondary_hue="sky")) as demo:
+with gr.Blocks(theme=gr.themes.Monochrome()) as demo:
     gr.Markdown(
         """
         # 🩸 Glucose Guardian: AI-Powered 30-Minute Prediction
-        This is a simulation of the AI model we trained. Enter a **current glucose** value to see the model's 30-minute forecast. 
-        The model automatically simulates other sensor and demographic data to make its prediction.
+        A simulation of an AI model that predicts future glucose levels. You can either adjust the sliders yourself
+        or click the "Generate Random Vitals" button to create a new scenario.
         """
     )
-    
+
     with gr.Row():
+        # --- INPUTS COLUMN ---
         with gr.Column(scale=1):
-            gr.Markdown("###  CGM Input")
-            glucose_input = gr.Slider(
-                label="Current Glucose Level (mg/dL)", 
-                minimum=40, maximum=400, step=1, value=125
-            )
+            with gr.Group():
+                gr.Markdown("###  CGM Input")
+                glucose_input = gr.Slider(label="Current Glucose (mg/dL)", minimum=40, maximum=400, step=1, value=125)
+
+            with gr.Group():
+                gr.Markdown("### 👤 Demographics")
+                gender_input = gr.Radio(label="Gender", choices=["Male", "Female"], value="Female")
+                hba1c_input = gr.Slider(label="HbA1c (%)", minimum=5.0, maximum=12.0, step=0.1, value=6.5)
+
+            with gr.Group():
+                gr.Markdown("### ⌚ Wearable Vitals (Simulated)")
+                hr_input = gr.Slider(label="❤️ Heart Rate (bpm)", minimum=50, maximum=160, step=1, value=75)
+                hrv_input = gr.Slider(label="🧬 HRV - RMSSD (ms)", minimum=10, maximum=120, step=1, value=45)
+                temp_input = gr.Slider(label="🌡️ Skin Temperature (°C)", minimum=30.0, maximum=38.0, step=0.1, value=34.5)
+        
+        # --- OUTPUTS AND BUTTONS COLUMN ---
+        with gr.Column(scale=2):
+            gr.Markdown("### 🔮 AI Forecast")
+            prediction_output = gr.Textbox(label="Predicted Glucose in 30 Minutes", interactive=False, elem_id="prediction_textbox")
             
-            with gr.Row():
-                sample_btn = gr.Button("Generate Sample Input")
-                predict_btn = gr.Button("Predict Future Glucose", variant="primary")
+            predict_btn = gr.Button("🔮 Predict Future Glucose (from sliders)", variant="primary")
+            random_btn = gr.Button("🎲 Generate Random Vitals & Predict")
             
-        with gr.Column(scale=2):
-            gr.Markdown("### 🔮 AI Forecast & Simulated Data")
-            prediction_output = gr.Textbox(label="Predicted Glucose in 30 Minutes", interactive=False)
-            status_output = gr.Textbox(label="Automatically Simulated Data (for this prediction)", interactive=False, lines=4)
+            gr.Markdown("---")
+            gr.Markdown("*Medical Disclaimer: This is a research project and is **NOT** an approved medical device. Predictions are for educational and demonstrational purposes only.*")
 
-    # --- Link buttons to functions ---
-    sample_btn.click(
-        fn=generate_sample_input,
-        inputs=[],
-        outputs=[glucose_input]
-    )
+
+    # --- Link UI components to functions ---
     predict_btn.click(
         fn=predict_glucose,
-        inputs=[glucose_input],
-        outputs=[prediction_output, status_output]
+        inputs=[glucose_input, hba1c_input, gender_input, hr_input, hrv_input, temp_input],
+        outputs=[prediction_output]
+    )
+    
+    random_btn.click(
+        fn=generate_random_scenario_and_predict,
+        inputs=[],
+        outputs=[glucose_input, hba1c_input, gender_input, hr_input, hrv_input, temp_input, prediction_output]
     )
-
-    gr.Markdown("---")
-    gr.Markdown("*Medical Disclaimer: This is a research project and is **NOT** an approved medical device. Predictions are for educational and demonstrational purposes only and should not be used for medical decisions.*")
 
 # Launch the app
 if __name__ == "__main__":