While the assistant is working, you now see a brief description of what it’s doing, instead of just a loading indicator. When a request triggers tool calls—such as reading a document, searching your knowledge base, or browsing the web… the interface displays these intermediate steps in real time (e.g., “Searching documents…,” “Reading your_doc.pdf…,” “Searching the web…”), making the wait for the first tokens clearer and more transparent.
