fix: Improve error handling for NVIDIA model component (#7484)

fix: ignore ConnectionError in  NVIDIA model component in case there's no internet connection

- Added logging for connection issues when accessing the NVIDIA API.
- Implemented exception handling for ConnectionError, MaxRetryError, and NameResolutionError to improve robustness and user feedback.

src/backend/base/langflow/components/models/nvidia.py

@@ -1,5 +1,9 @@
 from typing import Any
 
+from loguru import logger
+from requests.exceptions import ConnectionError  # noqa: A004
+from urllib3.exceptions import MaxRetryError, NameResolutionError
+
 from langflow.base.models.model import LCModelComponent
 from langflow.field_typing import LanguageModel
 from langflow.field_typing.range_spec import RangeSpec
@@ -19,6 +23,12 @@ class NVIDIAModelComponent(LCModelComponent):
     except ImportError as e:
         msg = "Please install langchain-nvidia-ai-endpoints to use the NVIDIA model."
         raise ImportError(msg) from e
+    except (ConnectionError, MaxRetryError, NameResolutionError):
+        logger.warning(
+            "Failed to connect to NVIDIA API. Model list may be unavailable."
+            " Please check your internet connection and API credentials."
+        )
+        all_models = []
 
     inputs = [
         *LCModelComponent._base_inputs,