There is no exception handling with the usage of set_model_response_tool with usage of output schema and tools combined.

model validation exceptions are swallowed resulting stuck llm process without refinement turns.

currently it requires the usage of callbacks or plugins to prevent this exception