Documentation

Files, text, Q&A and website pages

Everything your AI reads that is not a knowledge base or a documentation set — uploaded files, pasted text, question-and-answer pairs and pages of your website. It all belongs to the brand, and it lives on the lower half of the Knowledge screen.

Where these live

Open Brand → (your brand) → Knowledge. The upper half of that screen holds the brand's knowledge bases and documentation sets — the material that is also published for your customers to read. The lower half, headed Other sources, is this page's subject, and the screen states plainly what it is: "Everything else your AI reads. None of this appears on your public site — it is knowledge only, used to answer in chat, on the phone and in social posts."

The lower half has three tabs, each showing a count:

  • Website — pages of your site and sitemaps we read for you. These are the only sources that change on their own, so they get their own table and their own columns; they are covered on their own page.
  • Sources — files you upload, text you paste and Q&A pairs you write. Things that never change unless you change them.
  • API tools — endpoints the AI can call for live data, covered on their own page.

Everything here belongs to the brand, not to the AI agent. Replace or delete the agent and these sources stay exactly as they are, and website pages keep being re-read on their normal schedule.

What each tab shows

Website pages and sitemaps sit apart from the rest because they behave differently: we re-read them for you every week, they can fail, and they need columns that say when each was last read and when it will be read next. The Website tab has those; see the Website pages article below for what each column means.

The Sources tab is one searchable table for everything that is simply stored as you left it, with a coloured Type badge on every row:

  • Text — something you pasted in.
  • Q&A — a question paired with the answer you want given.
  • File — a document you uploaded.

You can search either table by name and filter it by type and by status. On the Sources tab each row shows Type, Status, Indexed and the date it was Added.

Row actions appear only when they apply to that row. On the Sources tab: See what was stored, Edit (text and Q&A only) and Delete. On the Website tab: See what was stored, Read again now, Pages found (sitemaps), Pause or Resume automatic re-reading, Open the page and Delete.

Adding a source

Each tab offers what belongs on it.

On the Website tab, Add takes a Website page or sitemap, and Smart import brings in a whole site at once. On the Sources tab, the Add menu offers three options:

  • Paste text
  • Question & answer
  • Upload files
Deleting a source removes it and everything the AI learned from it. The confirmation says so: "Delete “the source's name”? Everything your AI learned from it is removed at the same time."

Status — is this source actually working?

The Status column is the answer to "why isn't my AI using this?".

On both tabs it can read:

  • Ready — the AI can answer from this right now.
  • Not indexed — nothing was stored from it, so the AI cannot answer from it. Usually an empty file, or a page that could not be read.
  • Reading… — we are working through it at the moment.
  • Warning — it was read, but something went wrong along the way.

Files, text and Q&A can also show Failed — the last attempt to store it did not succeed.

Website pages and sitemaps have one of their own instead: Stopped. A web source that fails is retried on the following runs, and only after three failures in a row is it stopped — at which point automatic re-reading is switched off for it until you use Read again now. Your AI keeps answering from the last copy that worked, and the account owner is emailed when it happens.

Hover the information icon next to a badge to see the reason. On the Website tab an Issue column also spells the problem out in plain words. If anything has stopped, a warning appears the moment you open the page, so you are not left to discover it by scrolling.

Indexed — how much a source contributes

The Indexed column counts the searchable pieces the AI stored from that source. Longer sources become more pieces. A dash means nothing was stored, which is the same thing the Not indexed status is telling you.

Two things surprise people here. One article usually becomes several pieces, so the number is always larger than the number of documents you added. And for knowledge bases and documentation sets, only published articles are stored — a draft contributes nothing.

Smart import

Smart import is the fast way to bring in a whole website. You give it your website address; it reads the site's navigation or its sitemap, picks about twenty-five pages that actually help a support AI, and imports them in the background. That takes a few minutes, and you can leave the page while it runs.

Re-running it later updates a page it has already read instead of adding a second copy of it.

If your plan does not include Vex AI, Smart import tells you up front that the website pages will be skipped.

Product feeds are not here

A product feed is not listed among these sources, deliberately, so the same thing cannot be deleted from two different screens. Product feeds — their refresh status, their last error and how many products came out of them — live only on Brand → Products.

What language a source is searched in

A source's language decides which dictionary is used when the AI searches its text. It does not translate anything.

Knowledge bases and documentation sets each carry their own language, set on the row in the upper half of the Knowledge screen. The sources on this page — files, website pages, pasted text and Q&A — carry no language of their own and fall back to the brand's language, set at Brand → Brand settings → General.

The AI searches in every language the brand's content is written in, up to three at once, so a brand holding an English and a Hungarian knowledge base keeps full accuracy in both.

The indexed-item allowance

On the free plan your organisation can store 250 indexed items in total, and the Knowledge screen prints how many are used: "212 of 250 indexed items used". The pool is account-wide — every help article, documentation page, uploaded file, website page and Q&A pair across every brand draws from the same 250. It applies to module-only ("à-la-carte") accounts too: buying a module buys features, not content volume. At the cap, publishing new articles and adding new sources is paused; the buttons are not greyed out, so you find out through an error message when you try. See Free plan limits for the full breakdown.

Who can see this half of the screen

The Knowledge page opens for anyone holding Vex AI, Knowledge base or Documentation. The Other sources half below only renders for team members holding Vex AI — so a colleague who only manages help articles sees the upper half and nothing else.

Website pages

Add a single page or a whole sitemap, or let Smart import choose. Covers what makes a good page to add, what to leave out, and how re-reading works.

API tools

Let the AI fetch live data while it answers — order status, stock, account details. Supports forms that collect what the call needs, and handles failures gracefully.

Best practices

Give sources clear names. The table is searchable by name, and a list of files called export-final-2.pdf is a list you cannot manage.
Quality over quantity. Focus on high-quality, relevant material rather than adding everything you have. Better sources make better answers, and every piece counts against your allowance.
Check the Status column after adding anything. A source sitting at Not indexed contributes nothing, and nothing else on the screen will tell you.