components/src/dynamo/sglang/request_handlers/handler_base.py (1)

253-297: Missing error handling — inconsistent with existing handler pattern.

The existing release_memory_occupation and resume_memory_occupation methods wrap their logic in try/except and return structured error dicts ({"status": "error", "message": ...}). All five new methods propagate exceptions unhandled. If UpdateWeightFromDiskReqInput(**body) receives unexpected keys or a required field is missing, the caller gets an opaque Rust-level error instead of a structured JSON response.

Consider wrapping each method body in a try/except for consistency, e.g.:

Proposed pattern (applied to one method as example)

     async def update_weights_from_disk(self, body: dict) -> dict:
         """Update model weights from disk without restarting the server."""
         from sglang.srt.managers.io_struct import UpdateWeightFromDiskReqInput
 
-        req = UpdateWeightFromDiskReqInput(**body)
-        success, message, num_paused_requests = (
-            await self.engine.tokenizer_manager.update_weights_from_disk(req, None)
-        )
-        return {
-            "success": success,
-            "message": message,
-            "num_paused_requests": num_paused_requests,
-        }
+        try:
+            req = UpdateWeightFromDiskReqInput(**body)
+            success, message, num_paused_requests = (
+                await self.engine.tokenizer_manager.update_weights_from_disk(req, None)
+            )
+            return {
+                "success": success,
+                "message": message,
+                "num_paused_requests": num_paused_requests,
+            }
+        except Exception as e:
+            logging.error(f"Failed to update weights from disk: {e}")
+            return {"success": False, "message": str(e)}