diff --git a/components/src/dynamo/sglang/request_handlers/handler_base.py b/components/src/dynamo/sglang/request_handlers/handler_base.py
index 524fc282e219..005b023b6f09 100644
--- a/components/src/dynamo/sglang/request_handlers/handler_base.py
+++ b/components/src/dynamo/sglang/request_handlers/handler_base.py
@@ -413,6 +413,18 @@ async def load_lora(self, request: Optional[Dict[str, Any]] = None):
                                 else:
                                     lora_worker_type = WorkerType.Aggregated
                                     lora_needs = []
+
+                            # Reuse the base-model metadata builder so LoRA
+                            # cards advertise the same token-overflow policy,
+                            # parser configuration, and routing capabilities.
+                            # Lazy import: static test collection lacks parts of SGLang.
+                            from dynamo.sglang.register import _get_runtime_config
+
+                            runtime_config = await _get_runtime_config(
+                                self.engine,
+                                self.config.server_args,
+                                self.config.dynamo_args,
+                            )
                             await register_llm(
                                 model_input=ModelInput.Tokens,
                                 model_type=lora_model_type,
@@ -424,6 +436,7 @@ async def load_lora(self, request: Optional[Dict[str, Any]] = None):
                                 base_model_path=self.config.server_args.model_path,
                                 worker_type=lora_worker_type,
                                 needs=lora_needs,
+                                runtime_config=runtime_config,
                                 # Publish the worker's per-worker LoRA slot budget so the frontend
                                 # allocator sizes placement against real capacity instead of the
                                 # hard-coded default.