update dockerfile for hf

Dockerfile

@@ -27,14 +27,46 @@ RUN pip3 install --no-cache-dir -r requirements.txt
 # Copy application code
 COPY . .
 
-# Create startup script
+# Create startup script with health checks and retries
 RUN echo '#!/bin/bash\n\
+\n\
+# Function to check if Ollama is responsive\n\
+check_ollama() {\n\
+    curl -s http://localhost:11434/api/version &>/dev/null\n\
+}\n\
+\n\
 # Start Ollama server\n\
 ollama serve & \n\
-sleep 5\n\
 \n\
-# Pull the model if not exists\n\
-ollama pull deepseek-r1:1.5b\n\
+# Wait for Ollama to be responsive (up to 60 seconds)\n\
+count=0\n\
+while ! check_ollama && [ $count -lt 60 ]; do\n\
+    echo "Waiting for Ollama server to start..."\n\
+    sleep 1\n\
+    count=$((count + 1))\n\
+done\n\
+\n\
+if ! check_ollama; then\n\
+    echo "Failed to start Ollama server"\n\
+    exit 1\n\
+fi\n\
+\n\
+# Pull the model with retry logic\n\
+max_retries=3\n\
+retry_count=0\n\
+while [ $retry_count -lt $max_retries ]; do\n\
+    if ollama pull deepseek-r1:1.5b; then\n\
+        break\n\
+    fi\n\
+    echo "Failed to pull model, retrying..."\n\
+    retry_count=$((retry_count + 1))\n\
+    sleep 5\n\
+done\n\
+\n\
+if [ $retry_count -eq $max_retries ]; then\n\
+    echo "Failed to pull model after $max_retries attempts"\n\
+    exit 1\n\
+fi\n\
 \n\
 # Start the Gradio app\n\
 exec python3 -u app.py\n\

app.py

@@ -6,6 +6,8 @@ from typing import List, Optional, Tuple, Dict
 from ollama import Client
 import re
 import os
+import time
+import requests
 
 # Set up logging configuration
 logging.basicConfig(
@@ -18,6 +20,30 @@ logging.basicConfig(
 )
 logger = logging.getLogger(__name__)
 
+# Initialize Ollama client with retry logic
+def init_ollama_client(max_retries=5):
+    client = None
+    for i in range(max_retries):
+        try:
+            client = Client(host='http://localhost:11434')
+            # Test the connection
+            response = requests.get('http://localhost:11434/api/version')
+            if response.status_code == 200:
+                logger.info("Successfully connected to Ollama")
+                return client
+        except Exception as e:
+            logger.warning(f"Attempt {i+1}/{max_retries} to connect to Ollama failed: {str(e)}")
+            if i < max_retries - 1:
+                time.sleep(2)
+    raise Exception("Failed to initialize Ollama client")
+
+# Global client initialization
+try:
+    client = init_ollama_client()
+except Exception as e:
+    logger.error(f"Failed to initialize Ollama: {str(e)}")
+    client = None
+
 def calculate_wer_metrics(
     hypothesis: str,
     reference: str,
@@ -97,19 +123,11 @@ def calculate_wer_metrics(
     
     return measures
 
-# Initialize Ollama client
-client = Client(host='http://localhost:11434')
-
 def extract_medical_terms(text: str) -> List[str]:
-    """
-    Extract medical terms from text using Qwen model via Ollama.
-    
-    Args:
-        text (str): Input text
-        
-    Returns:
-        List[str]: List of extracted medical terms
-    """
+    """Extract medical terms from text using Qwen model via Ollama."""
+    if client is None:
+        logger.error("Ollama client not initialized")
+        return []
     prompt = f"""Extract all medical terms from the following text. 
     Return only the medical terms as a comma-separated list.
     Text: {text}"""
@@ -121,19 +139,15 @@ def extract_medical_terms(text: str) -> List[str]:
             stream=False
         )
         
-        # Get the response text
         response_text = response['response']
         
         # Remove the thinking process
         if '<think>' in response_text and '</think>' in response_text:
-            # Extract everything after </think>
             medical_terms_text = response_text.split('</think>')[-1].strip()
         else:
             medical_terms_text = response_text
         
-        # Parse the comma-separated response
         medical_terms = [term.strip() for term in medical_terms_text.split(',')]
-        # Remove empty terms and clean up
         return [term for term in medical_terms if term and not term.startswith('<') and not term.endswith('>')]
         
     except Exception as e:
@@ -323,11 +337,14 @@ def create_interface() -> gr.Blocks:
 
 if __name__ == "__main__":
     logger.info("Application started")
-    app = create_interface()
-    # Explicitly configure Gradio to be accessible from outside the container
-    app.launch(
-        server_name="0.0.0.0",  # Bind to all interfaces
-        server_port=7860,
-        share=False,  # Don't create a public URL
-        debug=True    # Enable debug mode for more information
-    )
+    try:
+        app = create_interface()
+        app.launch(
+            server_name="0.0.0.0",
+            server_port=7860,
+            share=False,
+            debug=True
+        )
+    except Exception as e:
+        logger.error(f"Failed to launch application: {str(e)}")
+        raise