# Asking the docs a question

> In short: Every page has a question box that answers from this documentation only, using AI, and lists the pages it used. If no page covers your question, it says so and links to support instead of guessing.

Applies to: v0.9 · Last reviewed: 2026-10-09 · Canonical: https://docs.zymcos.com/product/ask-the-docs/

Every page of this documentation carries a question box. Type a question in your own words and
you get an answer assembled from these pages, with links to the ones it used.

**You are interacting with an AI system.** Answers are generated by AI from the documentation
published on this site. They are not written by a person, and they are not checked by one
before you see them — which is why every answer names its sources, so you can read the
original page and judge for yourself.

## Jumping to the box from the keyboard

On any page, press **`/`** to jump to the question box and start typing. It does nothing while
you are typing in a field, so a `/` in your question is just a `/`. Press **Escape** in the box to
leave it. If you would rather not have the shortcut, use the **Keyboard shortcut** button under
the box to turn it off; your choice is remembered in your browser.

## Asking CODA instead

You can also ask CODA, in the chat, how something works — "how do I cancel a take that is still
generating?". CODA looks the answer up in these same published pages and tells you in its own words,
naming the page it used. It reads only the public documentation, and if the pages do not cover your
question it says so plainly instead of guessing.

## What it can and cannot answer

It answers from the published documentation, and from nothing else. It has no access to your
account, your projects, your billing, or anything you have created. If you ask about those,
it will tell you it cannot answer rather than guess.

## When it says it does not know

If the documentation does not cover your question, you get that answer plainly, with a link to
support. This is deliberate: an assistant that produces a confident, plausible answer with no
source behind it is worse than one that says nothing, because you have no way to tell the two
apart. The rule is simple — **if it cannot cite a page, it does not answer.**

That is also a useful signal for us. A question the docs cannot answer is usually a page we
have not written yet.

## Reading the sources

Every answer is followed by the pages it drew on. Open them when the answer matters:

- the source page is the authoritative version — the answer is a summary of it;
- the source page may cover conditions or exceptions the summary left out;
- if the answer and the page disagree, **the page is right**. Please tell support, because
  that disagreement is a bug worth fixing.

## Where it works

The question box answers questions asked on this site. It does not answer requests sent from
other websites, or by search engines and other automated crawlers: they are welcome to read
every page, but each answer is composed by a paid AI service, so the box is kept for people
reading the documentation.

## Your questions

The text of your questions is not stored. We count how many questions are asked and how often
we could not answer one, so we know which documentation to write next — but the questions
themselves are not kept.

## If it is not working

- **"Too many questions just now"** — there is a limit on how quickly questions can be asked
  from one place. Wait a moment and try again.
- **"The answer service is busy — please try again later"** — there is also a limit on how
  many questions are answered in a day, across everyone. When it is reached, the box answers
  again from midnight UTC. The documentation itself is unaffected.
- **"The answer service is unavailable"** — the question box is temporarily down. The
  documentation itself is unaffected: every page still works, and the navigation
  links are unchanged.
- **An answer that looks wrong** — open its sources. If the page is right and the answer is
  not, tell support and include the question you asked.
