Update

backend/app.py

@@ -352,13 +352,17 @@ try:
         else:
             print("⚠️ Warning: Hugging Face authentication failed, using local model only.")
     else:
-        print("⚠️ HF_TOKEN not found in environment — skipping authentication.")
+        print("⚠️ HF_TOKEN not found in environment — Path Foundation model cannot be loaded.")
+        print("   To use histopathology predictions, set the HF_TOKEN environment variable.")
 
     # Load Path Foundation model
+    print("Attempting to load Path Foundation model...")
     if classifier.load_path_foundation():
         print("✅ Loaded Path Foundation base model.")
     else:
-        print("⚠️ Could not load Path Foundation base model, continuing with local weights only.")
+        print("⚠️ WARNING: Could not load Path Foundation base model.")
+        print("   Histopathology predictions will not work without this model.")
+        print("   Set HF_TOKEN environment variable to enable this feature.")
 
     # Load trained histopathology model
     model_path = "histopathology_trained_model.keras"
@@ -367,10 +371,13 @@ try:
         print(f"✅ Loaded local histopathology model: {model_path}")
     else:
         print(f"⚠️ Model file not found: {model_path}")
+        print("   Histopathology predictions will not work without the trained weights.")
 
 except Exception as e:
     classifier = None
     print(f"❌ Error initializing histopathology model: {e}")
+    import traceback
+    traceback.print_exc()
 
 def predict_histopathology(image):
     if classifier is None:
@@ -381,7 +388,17 @@ def predict_histopathology(image):
             image = image.convert("RGB")
         image = image.resize((224, 224))
         img_array = np.expand_dims(np.array(image).astype("float32") / 255.0, axis=0)
+        
+        # Check if path_foundation model is loaded
+        if classifier.path_foundation is None:
+            return {"error": "Histopathology prediction failed: Path Foundation model not loaded. Ensure HF_TOKEN is set and the model can be downloaded."}
+        
         embeddings = classifier.extract_embeddings(img_array)
+        
+        # Check if model is loaded
+        if classifier.model is None:
+            return {"error": "Histopathology prediction failed: Classification model not loaded."}
+        
         prediction_proba = classifier.model.predict(embeddings, verbose=0)[0]
         predicted_class = int(np.argmax(prediction_proba))
         class_names = ["Benign", "Malignant"]

backend/model_histo.py

@@ -392,6 +392,10 @@ class BreastCancerClassifier:
             >>> # Process with smaller batch size for memory-constrained environments
             >>> embeddings = classifier.extract_embeddings(images, batch_size=8)
         """
+        # Check if model is loaded
+        if self.path_foundation is None:
+            raise RuntimeError("Path Foundation model is not loaded. Call load_path_foundation() first.")
+        
         print(f"Extracting embeddings from {len(images)} images...")
         
         embeddings = []