Update app.py

app.py

@@ -24,24 +24,30 @@ limiter = Limiter(key_func=get_remote_address, default_limits=[config.RATE_LIMIT
 # --- Lifespan Management (Model Loading/Unloading) ---
 @asynccontextmanager
 async def lifespan(app: FastAPI):
-    # Startup: Load models
+    # Startup: Assign essential state FIRST
+    app.state.limiter = limiter # <<<--- MOVE ASSIGNMENT HERE
+    logger.info("Rate limiter assigned to app state.")
+
+    # Now attempt to load models
     logger.info("Application startup: Loading models...")
     try:
-        generation.load_models()
-        app.state.limiter = limiter # Add limiter to app state for dependency injection
+        generation.load_models() # This might still fail
         logger.info("Models loaded successfully.")
     except Exception as e:
-        logger.error(f"FATAL: Model loading failed during startup: {e}", exc_info=True)
-        # Optionally, prevent app start or handle gracefully
-        # For now, we let it start but endpoints needing models will fail.
-    yield
+        # Log the error, but the app continues. Endpoints needing models might fail.
+        logger.error(f"ERROR: Model loading failed during startup: {e}", exc_info=True)
+
+    yield # Application is now ready to serve requests
+
     # Shutdown: Clean up (optional, less critical in stateless services)
     logger.info("Application shutdown.")
-    generation.model_cache.clear()
+    if "generation" in globals() and hasattr(generation, 'model_cache'):
+         generation.model_cache.clear()
     # Add any other cleanup here
     logger.info("Resources cleaned up.")
 
 
+
 # --- FastAPI App Initialization ---
 app = FastAPI(
     title="AI Content Generation API",
@@ -51,7 +57,7 @@ app = FastAPI(
 )
 
 # --- Middleware ---
-# Rate Limiting Middleware
+# Rate Limiting Middleware - Now it can safely access app.state.limiter
 app.add_middleware(SlowAPIMiddleware)
 
 # CORS Middleware (Uncomment and configure if needed for browser-based clients)