Train from a website

The Websites tab in Train reads the text of a public web page and saves it as a document Centrus can answer from. It suits information that already lives on a website, such as your own service and pricing pages, a supplier's product page or a manufacturer's support notes. Open Train from the menu on the left and click Websites, the third tab along. You need Train access to see it. Admins always have it, and an admin can give it to anyone else from Edit users and roles.

Use Websites when the page itself is the source and nobody has the information as a file. If you already have it as a document, such as a PDF manual or a Word procedure, upload it on Files instead, where Train reads its headings and tables and the file keeps its own name. Websites are kept apart from your files, so a page you add here appears on this tab and not in the main Files list.

Everything starts from one address box. Enter the address, then choose one of the two buttons underneath. They do different jobs, so it is worth knowing which one you need before you click.

• Use Crawl Single Page when you know the exact page you want, such as https://nfsfire.co.uk/faqs. The page becomes one document, named after its address.
• Use Scan Website when you want several pages from the same site, such as each of your service pages. You enter a page that links to them, usually the home page, tick the ones you want, and they become one combined document.

Always include https:// at the start of the address. An address such as nfsfire.co.uk on its own is rejected with the message Website or Webpage URL is not a valid url, and one that starts with www. but has no https:// can pass the check and then fail when it is scanned. Copying the address from your browser's address bar is the easiest way to get it exactly right.

1. 1Open Train and click the Websites tab.
2. 2Paste the page's full address into Website or Webpage URL.
3. 3Click Crawl Single Page.
4. 4Wait while the button shows a spinning circle. When the page has been read, a message at the top of the screen says Website scraped, the box clears, and the page appears at the top of Included Links.

Only the page at that address is read. Links on it are not followed, so adding your FAQ page does not bring in the pages it links to. Stay on the tab until the message appears, so you can see whether it worked.

If the page cannot be read, the message says Failed to scrape website and nothing is added. Check that the address opens in your own browser without signing in. The same message appears when your organisation has run out of storage, which is covered at the end of this guide.

1. 1Enter the address of a page that links to the pages you want, usually the home page, and click Scan Website.
2. 2The Website Scanner opens with your address already filled in. Click Scan.
3. 3Wait while Train reads the page. A list appears with a name and an address for each page it links to.
4. 4Click a page to tick it, and click it again to untick it. You can tick up to five.
5. 5Click Crawl Selected Pages. The number on the button shows how many pages you have ticked.

The scanner does nothing until you click Scan. Before then it already reads Scan complete and No pages found. Try scanning a different website., even though no scan has run, so click Scan before deciding the site has nothing to offer.

When the crawl finishes, the scanner closes and a message confirms how many pages were crawled. The pages arrive in Included Links as one document, named after the site and the number of pages, such as nfsfire.co.uk (3 pages). Because they are a single document, they share one set of tags and are removed together.

The scan lists the links on the page you entered, not every page on the site, and it only keeps links that begin with the same address. If you scan https://nfsfire.co.uk, a link written as https://www.nfsfire.co.uk is left out. When a page you want is missing, scan the page that links to it, or add it on its own with Crawl Single Page. If you try to tick a sixth page, Train shows You can select a maximum of 5 pages at a time. and leaves it unticked. To add more, scan again and pick the next five, which makes a second document. A failed scan shows Failed to scan website, and a failed crawl shows Failed to crawl website pages.

A crawl saves the words on a page as a text document, with its headings and lists kept in order. Centrus answers from that text in the same way it answers from a document you uploaded.

• Text, headings and lists are kept, so a page with clear headings gives Centrus clear sections to answer from.
• Pictures are not read. Wording that only appears inside an image, such as a price list saved as a picture, is lost.
• Menus, cookie banners and footers are usually left out, so the document holds what the page is about rather than the parts that repeat on every page.
• Pages from a scan are joined into one document, and each one starts under a heading that gives its address.

The document is a copy of the page as it was when you added it. Train does not go back to the site afterwards, so a page that changes needs adding again, as explained under updating a website further down.

Unlike the Files tab, there is no upload panel here. The button you clicked shows a spinning circle while Centrus reads the page, and a message at the top of the screen tells you how it went. Included Links lists the newest website first, so a page you have just added appears at the top.

Each row shows the website's name, its size, when it was added and who added it, followed by its tags. A page added with Crawl Single Page is named after its full address. A document made with Crawl Selected Pages is named after the site and its page count, and its row does not say who added it, because Train does not record that for scans.

On a screen narrower than 1024 pixels, such as a phone or a small tablet, each website is shown as a card, with its tags under the name and its size, date and bin underneath. Adding, tagging and removing websites work in the same way.

09:41

Centrus

Train

Add data sources to train Centrus

Included LinksSortFilterSelect All

https://nfsfire.co.uk/faqs

Primary tag

14.2kB08/10/2026 09:41JMJohn Martin

Tags sit under the name

nfsfire.co.uk (3 pages)

TechnicalSecond tag

38.6kB07/10/2026 16:12

https://nfsfire.co.uk/pricing

PricingSecond tag

9.71kB02/10/2026 11:20SPSarah Price

A new website has no tag, and this is the step that matters most. Centrus only answers from documents that carry a tag the chat is allowed to search, so an untagged website sits in Included Links without ever being used. Until it has a tag, it is also hidden from everyone except admins and the person who added it.

1. 1Find the website in Included Links and click Primary tag on its row.
2. 2Choose a tag from the list. Start typing to narrow the list down, or type a new name and choose Create to make a new tag there and then.
3. 3To add another tag, click Second tag, which appears once the first is set, and then Third tag if you need it.

Tags work here exactly as they do for uploaded files, including the limit of three tags and the way each extra tag counts the website's size against your storage again. Upload files explains tagging in full, and Control who sees what with tags explains how tags decide who can ask about what.

The controls in the Included Links heading help once the list grows. They work on every website you can see, so there is nothing extra to load first.

• The magnifying glass opens a search box that matches words in each website's name. For a single page, the name is its address.
• Sort orders the list by name, size, date added, document type or tag, each in either direction. Until you choose, the newest website comes first.
• Filter narrows the list by Document Type or by Tag. Choose Tag, then No Tag, to find every website Centrus cannot use yet. The tag filter only looks at each website's primary tag. A dot on the button shows that a filter is on, and Clear turns off the filters and the search together.
• Select All ticks every website the list is showing, and changes to Deselect All once anything is ticked. The arrow at the end of the heading folds the list away.

As soon as you tick a website, an Actions button appears beside a count of how many you have selected.

• Apply Tag sets the tag you choose as the primary tag on every selected website, replacing any primary tag they already had. Unlike the Files tab, this list has no option to create a new tag, so make the tag first from a row's tag picker or in Settings.
• Add to folder puts the selected websites into one of your folders. Folders are created on the Files tab, and a website you add to one is listed there when you open that folder.
• Delete removes the selected websites straight away. There is no confirmation step, so check the selection before you click it.

Train does not revisit websites. There is no button to crawl a page again and no schedule, so each website in Included Links keeps the text it had on the day it was added. When a page changes, for example after a price rise, add it again and remove the old copy.

1. 1Add the page again with Crawl Single Page, or scan and tick the same pages again. Train makes a new, untagged document and leaves the old one where it is.
2. 2Give the new copy the same tags as the old one.
3. 3Delete the old copy, which is the one with the earlier date.

Tag the new copy before you delete the old one, so there is no gap in what Centrus can answer. Adding the same address twice never merges or replaces anything, so look for older copies whenever you add a page again.

To remove a website, click the bin at the end of its row and confirm with Delete. It is removed from Train and from every tag it carried, so Centrus stops using it straight away. Removing it cannot be undone, but it has no effect on the website itself, and you can add the page again whenever you need to.

You can rename a website by double-clicking its name, but we recommend leaving website names as they are. Train treats everything after the last full stop in the original name as a file extension and adds it to the new name, so renaming https://nfsfire.co.uk/faqs to Northfield FAQs saves it as Northfield FAQs.uk/faqs.

A crawl reads a page as a visitor would on a first visit, without signing in, clicking anything or following links. That covers most public pages, but not everything on a website.

• Public pages only. Train reads a page as a visitor who has not signed in would see it. A page behind a login, such as a customer portal or an intranet, gives you the sign-in page or an error rather than its content.
• One page per address. A crawl reads the page you give it and none of the pages it links to. A scan finds the links on one page and adds up to five of them at a time.
• Files on a website. PDFs, Word documents and spreadsheets linked from a page are not followed. Download them and upload them on Files, where Train reads their headings and tables.
• Content you have to click to see. Text that only appears after you open a tab, press a button such as Load more, or run a search may not come in. If answers seem incomplete, add the page where that text appears on its own, or paste the text into Text / Audio.
• Pictures. Words inside images are not read.
• How often you can add. Each person can run Scan Website and Crawl Selected Pages up to ten times a minute, and Crawl Single Page up to thirty times. Beyond that, the request fails with the usual error message, so wait a minute and try again.

Admins see every website in Included Links. Everyone else with Train access sees two kinds of website: those whose primary tag is one they have been given, and untagged pages they added themselves with Crawl Single Page.

John Martin

Admin

Sees every website

• https://nfsfire.co.uk/faqsNo tag
• nfsfire.co.uk (3 pages)Technical
• https://nfsfire.co.uk/pricingPricing
• https://nfsfire.co.uk/standardsSOPs
• https://nfsfire.co.uk/contactNo tag

DH

Dylan Hughes

Engineer, with Train access

Tags: Technical, SOPs

• nfsfire.co.uk (3 pages)Technical
• https://nfsfire.co.uk/standardsSOPs
• https://nfsfire.co.uk/contactNo tag
• https://nfsfire.co.uk/pricingNot shown
• https://nfsfire.co.uk/faqs (no tag)Not shown

Two details differ from the Files tab. Only the primary tag counts here, so a website stays hidden from you if its second or third tag is one of yours but its primary tag is not. And because a scan does not record who added it, a document made with Crawl Selected Pages does not appear for someone who is not an admin until an admin gives it one of their tags. If you are not an admin, add the pages one at a time with Crawl Single Page, or ask an admin to tag the scan for you.

Websites count towards your organisation's storage limit like every other source in Train. A website's size is the size of the text Train saved, not of the page in your browser, so a typical page takes up kilobytes rather than megabytes. As with files, each extra tag counts the size again.

The Sources bar at the foot of the Websites tab shows how much is in use against your limit, and turns red at 90 per cent. When storage runs out, Crawl Single Page shows Failed to scrape website and Crawl Selected Pages shows Failed to crawl website pages, without saying why, so check this bar if a crawl that should work keeps failing. To make room, delete websites or files you no longer need. If you need more space, extra storage is available as an add-on. See Add-ons.