You type a question into the AI Assistant’s input field and get the response
directly below it.
Asking a question
Enter sends the question; Shift+Enter inserts a line break instead.
Copying a response
The Copy button copies the unformatted text of the response, not the
rendered view: a table or list is copied to the clipboard as Markdown.
Rating a response
The two thumb buttons below a completed response (Helpful Response,
Unhelpful Response) report your rating to product analytics. Only the
rating and the chat’s identifier are sent, never the content of the question
or the response. The rating is not saved with the chat itself.
Attachments
If the assistant generates a document during a response, it appears as an
attachment under the response – with an icon, file name and, if reported,
file size. Clicking the row downloads the file; a link the assistant writes
into its own response text never triggers a download.
If you abort a running response – with Stop Response, with New, or by
closing the assistant – an attachment generated up to that point is lost,
even if the assistant had already created the file. An attachment from an
already completed response is not affected.
Read aloud
Next to Copy, if the browser supports speech synthesis, is Read Aloud.
Clicking it reads the plain text of the response aloud; clicking Stop
Reading stops it. Only one response is read aloud at a time – a new
response, sending the next question, or closing the assistant stops a
running playback.
If New triggers a confirmation prompt, a running playback continues
unaffected until you confirm the prompt.
Errors, retrying and stopping
A blocking error appears below the history with a Retry button. The text
depends on the cause:
Retry repeats the last question sent; if the chat could not be created in
the first place, it starts a new one instead. Stop a running response with
Stop Response; the text received up to that point remains.
If a single processing step fails in the middle of a response, the response
does not abort – the assistant tries again. If it compensates for the error,
no note appears. Otherwise, a note appears below the completed response: “A
processing step failed – the response may be incomplete.”
Used-up quota
When the AI quota is used up, the assistant removes the started response from
the history and shows the message “The AI quota for this month is used up.”
below the history. Your question remains. There is no Retry in this case,
and the input field stays locked – even after you close and reopen the
assistant.
The lock ends as soon as quota is available again after a purchase. New
also lifts the lock; if the quota is still empty, the assistant rejects the
next question again and the lock returns. How the quota is made up is
described under Quota.
Export
Once a history exists, export it via the Export button as JSON,
Markdown, CSV or PDF – without involving the assistant again. With the
PDF format, an export can fail with “Busy right now – wait a moment and try
again”; trying again after a short wait usually resolves it. Export
always downloads the whole history; Copy affects only a single response,
an attachment only one file the assistant generated.
Starting a new chat
Once a history exists, discard it via the New button in the assistant’s
header and start a new chat – this also deletes all files generated so far.
Unlike simply closing the assistant (see
Fundamentals), this is final.
If a running response or an undownloaded file would be lost, New asks for
confirmation first. A history that is already complete with no open
downloads, however, is discarded without confirmation – export it first if
you want to keep it. Once confirmed, the history and its files cannot be
restored.
