Load model metadata and images from Hopsworks, update model info display

app.py

@@ -12,6 +12,7 @@ This app:
 
 import os
 import sys
+import re
 import logging
 import gradio as gr
 from typing import Tuple
@@ -27,6 +28,7 @@ logger = logging.getLogger(__name__)
 
 # Global inference pipeline
 pipeline = None
+model_info_cache = None
 
 
 def initialize_app():
@@ -48,47 +50,186 @@ def initialize_app():
         pipeline.load_model_from_hopsworks()
         logger.info("Inference pipeline initialized successfully!")
 
+        # Load model metadata (metrics and images)
+        logger.info("Loading model metadata and images...")
+        load_model_metadata()
+
         return True
     except Exception as e:
         logger.error(f"Failed to initialize app: {e}")
         return False
 
 
+def load_model_metadata():
+    """Load model metadata and images from Hopsworks."""
+    global model_info_cache
+
+    try:
+        from phising_detection.utils.hopsworks_utils import connect_to_hopsworks
+
+        project = connect_to_hopsworks()
+        mr = project.get_model_registry()
+
+        # Get the same model that pipeline loaded
+        model_version = getattr(pipeline, 'model_version', None) if pipeline else None
+        model_name = pipeline.model_name if pipeline else "phishing_detector"
+
+        if model_version:
+            model_registry = mr.get_model(model_name, version=model_version)
+        else:
+            model_registry = mr.get_model(model_name)
+
+        # Download model artifacts to get images and metrics
+        model_dir = model_registry.download()
+
+        # Parse hyperparameters.txt for metrics
+        metrics = {}
+        hyperparams_path = os.path.join(model_dir, "hyperparameters.txt")
+        if os.path.exists(hyperparams_path):
+            with open(hyperparams_path, 'r') as f:
+                content = f.read()
+
+                # Extract model name (first line)
+                lines = content.split('\n')
+                if lines:
+                    first_line = lines[0].strip()
+                    if 'Phishing Detection Model' in first_line:
+                        model_type = first_line.split('-')[-1].strip() if '-' in first_line else 'Unknown'
+                        metrics['model_type'] = model_type
+
+                # Extract training info
+                if 'CV Folds:' in content:
+                    cv_match = re.search(r'CV Folds:\s*(\d+)', content)
+                    if cv_match:
+                        metrics['cv_folds'] = int(cv_match.group(1))
+
+                    iter_match = re.search(r'RandomizedSearchCV iterations:\s*(\d+)', content)
+                    if iter_match:
+                        metrics['search_iterations'] = int(iter_match.group(1))
+
+                    cv_score_match = re.search(r'Best CV Score:\s*([\d.]+)', content)
+                    if cv_score_match:
+                        metrics['best_cv_score'] = float(cv_score_match.group(1))
+
+                # Extract Test Performance metrics
+                if 'Test Performance:' in content:
+                    test_section = content.split('Test Performance:')[1].split('Features:')[0]
+
+                    acc_match = re.search(r'Accuracy:\s*([\d.]+)', test_section)
+                    if acc_match:
+                        metrics['test_accuracy'] = float(acc_match.group(1))
+
+                    prec_match = re.search(r'Precision:\s*([\d.]+)', test_section)
+                    if prec_match:
+                        metrics['test_precision'] = float(prec_match.group(1))
+
+                    rec_match = re.search(r'Recall:\s*([\d.]+)', test_section)
+                    if rec_match:
+                        metrics['test_recall'] = float(rec_match.group(1))
+
+                    f1_match = re.search(r'F1 Score:\s*([\d.]+)', test_section)
+                    if f1_match:
+                        metrics['test_f1_score'] = float(f1_match.group(1))
+
+                    roc_match = re.search(r'ROC-AUC:\s*([\d.]+)', test_section)
+                    if roc_match:
+                        metrics['test_roc_auc'] = float(roc_match.group(1))
+
+                # Extract number of features
+                if 'Features:' in content:
+                    features_section = content.split('Features:')[1].strip()
+                    # Count features in the list
+                    feature_list = re.findall(r"'([^']+)'", features_section)
+                    metrics['n_features'] = len(feature_list)
+                    metrics['feature_names'] = ', '.join(feature_list)
+
+        # Get image paths
+        confusion_path = os.path.join(model_dir, "evaluation_image_1.png")
+        importance_path = os.path.join(model_dir, "evaluation_image_2.png")
+
+        model_info_cache = {
+            'version': model_registry.version,
+            'metrics': metrics,
+            'confusion_matrix_path': confusion_path if os.path.exists(confusion_path) else None,
+            'feature_importance_path': importance_path if os.path.exists(importance_path) else None
+        }
+
+        logger.info(f"Loaded model metadata: version {model_registry.version}, {len(metrics)} metrics")
+
+    except Exception as e:
+        logger.error(f"Error loading model metadata: {e}")
+        model_info_cache = None
+
+
 def get_model_info() -> str:
     """Get model information as HTML."""
-    if not pipeline:
-        return "<p>Model information not available</p>"
+    if not model_info_cache:
+        return "<p>Model information not available. Try refreshing the page.</p>"
 
-    # TODO: Check if pipeline has model_metrics attribute
-    metrics = getattr(pipeline, 'model_metrics', None)
-    if not metrics:
-        return "<p>Model metrics not available</p>"
+    metrics = model_info_cache.get('metrics', {})
+    version = model_info_cache.get('version', 'unknown')
+    model_name = pipeline.model_name if pipeline else "phishing_detector"
 
-    version = getattr(pipeline, 'model_version', 'latest')
+    # Check if we have any metrics
+    if not metrics:
+        return "<p>No metrics found in model artifacts.</p>"
+
+    # Helper function to safely format metric values
+    def format_metric(key, default='N/A'):
+        value = metrics.get(key, default)
+        if value == default:
+            return default
+        if isinstance(value, float):
+            return f"{value:.4f}"
+        if isinstance(value, int):
+            return str(value)
+        return str(value)
+
+    # Get model type from metrics or use default
+    model_type = metrics.get('model_type', 'Unknown')
 
     html = f"""
     <div style="padding: 20px; background-color: #f5f5f5; border-radius: 10px; margin: 10px;">
-        <h3>Model Information</h3>
-        <p><strong>Model Name:</strong> {pipeline.model_name}</p>
-        <p><strong>Model Version:</strong> {version}</p>
+        <h3>🤖 Model Information</h3>
+        <p><strong>Model Type:</strong> {model_type}</p>
+        <p><strong>Model Name:</strong> {model_name}</p>
+        <p><strong>Version:</strong> {version}</p>
         <hr>
-        <h4>Performance Metrics</h4>
-        <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 10px;">
-            <div><strong>Test Accuracy:</strong> {metrics.get('test_accuracy', 'N/A'):.4f}</div>
-            <div><strong>Test Precision:</strong> {metrics.get('test_precision', 'N/A'):.4f}</div>
-            <div><strong>Test Recall:</strong> {metrics.get('test_recall', 'N/A'):.4f}</div>
-            <div><strong>Test F1 Score:</strong> {metrics.get('test_f1_score', 'N/A'):.4f}</div>
-            <div><strong>Test ROC-AUC:</strong> {metrics.get('test_roc_auc', 'N/A'):.4f}</div>
-            <div><strong>Validation F1 Score:</strong> {metrics.get('val_f1_score', 'N/A'):.4f}</div>
+        <h4>📊 Test Performance</h4>
+        <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 15px; margin: 15px 0;">
+            <div style="padding: 10px; background-color: white; border-radius: 5px;">
+                <strong>Accuracy:</strong> <span style="font-size: 1.2em; color: #2196F3;">{format_metric('test_accuracy')}</span>
+            </div>
+            <div style="padding: 10px; background-color: white; border-radius: 5px;">
+                <strong>Precision:</strong> <span style="font-size: 1.2em; color: #4CAF50;">{format_metric('test_precision')}</span>
+            </div>
+            <div style="padding: 10px; background-color: white; border-radius: 5px;">
+                <strong>Recall:</strong> <span style="font-size: 1.2em; color: #FF9800;">{format_metric('test_recall')}</span>
+            </div>
+            <div style="padding: 10px; background-color: white; border-radius: 5px;">
+                <strong>F1 Score:</strong> <span style="font-size: 1.2em; color: #9C27B0;">{format_metric('test_f1_score')}</span>
+            </div>
+            <div style="padding: 10px; background-color: white; border-radius: 5px; grid-column: span 2;">
+                <strong>ROC-AUC:</strong> <span style="font-size: 1.2em; color: #F44336;">{format_metric('test_roc_auc')}</span>
+            </div>
         </div>
         <hr>
-        <h4>Training Information</h4>
-        <p><strong>Number of Features:</strong> {metrics.get('n_features', 'N/A')}</p>
-        <p><strong>Training Samples:</strong> {metrics.get('n_train_samples', 'N/A')}</p>
-        <p><strong>Test Samples:</strong> {metrics.get('n_test_samples', 'N/A')}</p>
-        <p><strong>Best CV Score:</strong> {metrics.get('best_cv_score', 'N/A'):.4f}</p>
-    </div>
+        <h4>🎯 Training Details</h4>
+        <p><strong>CV Folds:</strong> {format_metric('cv_folds')}</p>
+        <p><strong>Search Iterations:</strong> {format_metric('search_iterations')}</p>
+        <p><strong>Best CV Score:</strong> {format_metric('best_cv_score')}</p>
+        <p><strong>Number of Features:</strong> {format_metric('n_features')}</p>
     """
+
+    # Add feature names if available
+    if 'feature_names' in metrics:
+        html += f"""
+        <hr>
+        <h4>📝 Features Used</h4>
+        <p style="font-size: 0.9em; line-height: 1.6;">{metrics['feature_names']}</p>
+        """
+
+    html += "</div>"
     return html
 
 
@@ -236,31 +377,20 @@ def create_gradio_app():
 
                 gr.Markdown("### Model Evaluation Visualizations")
 
-                # TODO: Check if pipeline has model_dir attribute to access images
-                model_dir = getattr(pipeline, 'model_dir', None) if pipeline else None
-
                 with gr.Row():
                     with gr.Column():
                         gr.Markdown("#### Confusion Matrix")
-                        if model_dir:
-                            confusion_path = os.path.join(model_dir, "evaluation_image_1.png")
-                            if os.path.exists(confusion_path):
-                                gr.Image(value=confusion_path, label="Confusion Matrix")
-                            else:
-                                gr.Markdown("*Confusion matrix image not found*")
+                        if model_info_cache and model_info_cache.get('confusion_matrix_path'):
+                            gr.Image(value=model_info_cache['confusion_matrix_path'], label="Confusion Matrix")
                         else:
-                            gr.Markdown("*Model directory not available*")
+                            gr.Markdown("*Confusion matrix image not available*")
 
                     with gr.Column():
                         gr.Markdown("#### Feature Importance")
-                        if model_dir:
-                            importance_path = os.path.join(model_dir, "evaluation_image_2.png")
-                            if os.path.exists(importance_path):
-                                gr.Image(value=importance_path, label="Feature Importance")
-                            else:
-                                gr.Markdown("*Feature importance image not found*")
+                        if model_info_cache and model_info_cache.get('feature_importance_path'):
+                            gr.Image(value=model_info_cache['feature_importance_path'], label="Feature Importance")
                         else:
-                            gr.Markdown("*Model directory not available*")
+                            gr.Markdown("*Feature importance image not available*")
 
         gr.Markdown(
             """