lazy model loading

app.py

@@ -103,25 +103,13 @@ def load_models_from_mlflow():
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
-    """Load models on startup"""
-    logging.info("===== Application Startup - Checking for models =====")
-    
-    if MLFLOW_AVAILABLE:
-        try:
-            # Try to load models but don't block startup if it fails
-            logging.info("Attempting to load models from MLflow...")
-            success = load_models_from_mlflow()
-            if success:
-                logging.info("✅ Models loaded successfully from MLflow")
-            else:
-                logging.warning("⚠️ No models found in MLflow. Train via /train endpoint.")
-        except Exception as e:
-            logging.warning(f"⚠️ Could not load from MLflow: {e}. Train via /train endpoint.")
-    else:
-        logging.warning("⚠️ MLflow not available. Please train via /train endpoint.")
-    
+    """Initialize application on startup"""
+    logging.info("===== Application Startup =====")
+    logging.info("⚠️ Models will be loaded on first /train or /predict request")
     logging.info("✅ Application ready to serve requests")
+    
     yield
+    
     logging.info("===== Application Shutdown =====")
 
 app = FastAPI(lifespan=lifespan)