Give users useful visibility into an assistant’s work without placing its trace ahead of the answer. Reasoning stays compact until someone chooses to inspect it.
Copy the command below and run it in your terminal.
Use streaming while the model is thinking; use duration when it completes.