The reader sees what the assistant is doing — loading, searching, reading a page, thinking, writing — and how long a reasoning model thought.
Status and reasoning
A wait without information feels broken. While a turn runs, one line under the question names the current step, and a reasoning model's thinking sits in a collapsible block that opens while it thinks and closes when the answer starts.
DocPilot
Why did my webhook stop firing?
The reader asks about deliveries stopping. The webhooks page says an endpoint is disabled after 8 failed attempts…
Loading the docs index
Ask about Acme
All docsAI-generated. Check the linked pages.
The steps
| Line | When |
|---|---|
| Loading the docs index | the first question, while the index downloads |
| Searching the docs | retrieval in the browser |
| Looking at the page list | the model asked for the list of pages |
| Reading guide/webhooks | the model opened a section |
| Checking Plan limits | one of the site's own tools is running; the line names its title |
| Thinking | a reasoning model before its first word |
| Writing the answer | the answer is streaming |
With ui.waitingEscalation, a turn with no answer yet says Still working at 8 seconds, and Still working — the answer has not started yet at 25. After a reasoning model finishes, the block reads Thought for 4s and stays available to open.
Options
| Key | Default | What it does |
|---|---|---|
chat.reasoning | 'auto' | Asks a reasoning model to think, and shows the thinking. false turns it off; a level or { effort, budgetTokens, visible } tunes it. |
ui.waitingEscalation | true | The Still working lines. |
More: chat.reasoning in the reference.