Update app.py

app.py

@@ -1,15 +1,32 @@
-# app.py - Main Gradio application
+# app.py - MedGemma with Authentication
 import gradio as gr
 import torch
 from transformers import AutoProcessor, AutoModelForImageTextToText
 from PIL import Image
 import os
 import logging
+from huggingface_hub import login
 
 # Configure logging
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
+# Authenticate with Hugging Face
+def authenticate_hf():
+    """Authenticate with Hugging Face using token"""
+    try:
+        hf_token = os.getenv('HF_TOKEN')
+        if hf_token:
+            login(token=hf_token)
+            logger.info("✅ Authenticated with Hugging Face")
+            return True
+        else:
+            logger.warning("⚠️ No HF_TOKEN found in environment")
+            return False
+    except Exception as e:
+        logger.error(f"❌ Authentication failed: {e}")
+        return False
+
 # Model configuration
 MODEL_ID = "google/medgemma-4b-it"
 
@@ -18,46 +35,63 @@ model = None
 processor = None
 
 def load_model():
-    """Load model and processor with error handling"""
+    """Load model and processor with authentication"""
     global model, processor
     
     try:
+        # First authenticate
+        auth_success = authenticate_hf()
+        if not auth_success:
+            logger.error("❌ Authentication required for MedGemma")
+            return False
+            
         logger.info(f"Loading model: {MODEL_ID}")
         
         # Check if CUDA is available
         device = "cuda" if torch.cuda.is_available() else "cpu"
         logger.info(f"Using device: {device}")
         
-        # Load model with appropriate settings for Spaces
+        # Load model with authentication
         model = AutoModelForImageTextToText.from_pretrained(
             MODEL_ID,
             torch_dtype=torch.bfloat16 if device == "cuda" else torch.float32,
             device_map="auto" if device == "cuda" else None,
             trust_remote_code=True,
-            low_cpu_mem_usage=True
+            low_cpu_mem_usage=True,
+            token=True  # Use authenticated token
         )
         
-        processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
+        processor = AutoProcessor.from_pretrained(
+            MODEL_ID, 
+            trust_remote_code=True,
+            token=True  # Use authenticated token
+        )
         
-        logger.info("Model loaded successfully!")
+        logger.info("✅ Model loaded successfully!")
         return True
         
     except Exception as e:
-        logger.error(f"Error loading model: {str(e)}")
+        logger.error(f"❌ Error loading model: {str(e)}")
         return False
 
 # Initialize model at startup
 model_loaded = load_model()
 
 def analyze_medical_image(image, clinical_question, patient_history=""):
-    """
-    Analyze medical image with clinical context
-    """
+    """Analyze medical image with clinical context"""
     global model, processor
     
     # Check if model is loaded
     if not model_loaded or model is None or processor is None:
-        return "❌ Model not loaded. Please try refreshing the page or contact support."
+        return """❌ **Model Authentication Issue**
+
+MedGemma requires authentication. Please ensure:
+
+1. **HF_TOKEN is set**: The Space owner needs to add their Hugging Face token to Space Settings → Repository secrets
+2. **Model access approved**: Make sure you have access to MedGemma at https://huggingface.co/google/medgemma-4b-it
+3. **Space restart**: After adding the token, restart the Space
+
+**Current Status**: Authentication failed - model cannot load without proper token."""
     
     if image is None:
         return "⚠️ Please upload a medical image first."
@@ -103,7 +137,6 @@ def analyze_medical_image(image, clinical_question, patient_history=""):
         
         # Move to appropriate device
         device = next(model.parameters()).device
-        dtype = next(model.parameters()).dtype
         inputs = {k: v.to(device) for k, v in inputs.items()}
         
         input_len = inputs["input_ids"].shape[-1]
@@ -129,7 +162,7 @@ def analyze_medical_image(image, clinical_question, patient_history=""):
         
         # Add structured disclaimer
         disclaimer = """
-        
+
 ---
 ### ⚠️ MEDICAL DISCLAIMER
 **This analysis is for educational and research purposes only.**
@@ -162,6 +195,13 @@ def create_interface():
             padding: 16px;
             margin: 16px 0;
         }
+        .auth-warning {
+            background-color: #fffbeb;
+            border: 1px solid #fed7aa;
+            border-radius: 8px;
+            padding: 16px;
+            margin: 16px 0;
+        }
         """
     ) as demo:
         
@@ -179,17 +219,30 @@ def create_interface():
         - 🩺 **Dermatology** - Skin lesions and conditions
         """)
         
-        # Warning banner
-        with gr.Row():
+        # Authentication status
+        if not model_loaded:
             gr.Markdown("""
-            <div class="disclaimer">
-            ⚠️ <strong>IMPORTANT MEDICAL DISCLAIMER</strong><br>
-            This tool is for <strong>educational and research purposes only</strong>. 
-            Do not upload real patient data or use for actual medical diagnosis. 
-            Always consult qualified healthcare professionals.
+            <div class="auth-warning">
+            🔐 <strong>AUTHENTICATION REQUIRED</strong><br>
+            MedGemma model requires authentication. Please:
+            <ol>
+            <li>Ensure you have access to the model at <a href="https://huggingface.co/google/medgemma-4b-it">MedGemma page</a></li>
+            <li>Add your HF_TOKEN to Space Settings → Repository secrets</li>
+            <li>Restart the Space</li>
+            </ol>
             </div>
             """)
         
+        # Warning banner
+        gr.Markdown("""
+        <div class="disclaimer">
+        ⚠️ <strong>IMPORTANT MEDICAL DISCLAIMER</strong><br>
+        This tool is for <strong>educational and research purposes only</strong>. 
+        Do not upload real patient data or use for actual medical diagnosis. 
+        Always consult qualified healthcare professionals.
+        </div>
+        """)
+        
         with gr.Row():
             # Left column - Inputs
             with gr.Column(scale=1):
@@ -221,8 +274,12 @@ def create_interface():
                     analyze_btn = gr.Button("🔍 Analyze Image", variant="primary", size="lg")
                 
                 # Model status
+                auth_status = "✅ Authenticated" if model_loaded else "🔐 Authentication Required"
+                model_status = "✅ Loaded" if model_loaded else "❌ Not Loaded"
+                
                 gr.Markdown(f"""
-                **Model Status:** {'✅ Loaded' if model_loaded else '❌ Not Loaded'}  
+                **Authentication:** {auth_status}  
+                **Model Status:** {model_status}  
                 **Model:** {MODEL_ID}  
                 **Device:** {'CUDA' if torch.cuda.is_available() else 'CPU'}
                 """)
@@ -289,18 +346,10 @@ def create_interface():
         # Footer information
         gr.Markdown("""
         ---
-        ### 📚 About MedGemma
+        ### 🔬 About MedGemma
         
         MedGemma is Google's specialized medical AI model trained on medical imaging and clinical text. 
-        It excels at:
-        - Multi-modal medical image analysis
-        - Clinical reasoning and differential diagnosis
-        - Structured medical reporting
-        - Educational medical content generation
-        
-        **Supported Image Types:** JPEG, PNG, TIFF, DICOM (converted)  
-        **Max Image Size:** 10MB  
-        **Optimal Resolution:** 896x896 pixels (auto-resized)
+        **Note**: This model requires authentication and access approval.
         
         ### 🔒 Privacy & Data Policy
         - **No data storage**: Images and text are processed in real-time and not saved
@@ -321,5 +370,5 @@ if __name__ == "__main__":
     demo.launch(
         server_name="0.0.0.0",
         server_port=7860,
-        share=True
+        show_error=True
     )