> ## Documentation Index
> Fetch the complete documentation index at: https://www.docusnap.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Having a Conversation

> Ask a question, reuse a response, handle errors, and start a new chat.

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.

<Warning>
  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.
</Warning>

## 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.

<Note>
  If *New* triggers a confirmation prompt, a running playback continues
  unaffected until you confirm the prompt.
</Note>

## Errors, retrying and stopping

A blocking error appears below the history with a *Retry* button. The text
depends on the cause:

| Cause | Displayed text |
| - | - |
| Timeout | "The response took too long." |
| Too many requests | "Too many requests – wait a moment and try again" |
| Assistant unavailable | "The assistant is currently unavailable." |
| Assistant reports a processing error | "The request could not be processed." |
| Connection disrupted during the response | "Response could not be loaded." |
| Chat could not be created | "Chat could not be started." |

*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.

<Note>
  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."
</Note>

## 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](/docs/en/assistant/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](/docs/en/assistant/overview)), this is final.

<Warning>
  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.
</Warning>

## Related

What the assistant may read and change is described under [Permissions and
Data](/docs/en/assistant/permissions-and-data).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.