Text this: Evaluating a large language model's ability to answer clinicians' requests for evidence summaries.