Website Training

Open Knowledge → Add source → Website.

Paste your addresses

Use the address field for one website, a sitemap URL or several page URLs, one per line. You can paste addresses with or without https://.

For one homepage, Also add other pages from this website is selected by default. Uncheck it to add only the homepage. For one specific page, Also add subpages is off by default. Turn it on to discover only that page and its descendants: /help/start belongs to /help, but /helpful does not.

With either option enabled, Boei looks for a sitemap first and otherwise discovers URLs through the website crawler. A selected subpath remains the boundary on later syncs too.

For several pasted links, sitemaps and file links, no scope checkbox is shown. Several links add only those addresses, while an explicitly pasted sitemap contributes its listed pages. Documents use document processing. Duplicate addresses are removed.

Discovery lists addresses without fetching the contents of every page. For pages that need a password or access token, turn on Access to protected pages and follow Protected pages. Pages missing from your sitemap can also be added by address.

Choose what to train

Open the website source in Knowledge.
Expand a folder if needed, or search and filter the pages you want to include.
Select the pages, or select the source to include its contents.
Click Train chatbot.

Reading and training continue in the background. You can leave the screen. Boei trains what fits your available space and pauses when you need to free space or upgrade. It retains measured content for resuming. See Training credits.

A page already added individually can be adopted into its website source when discovered there. It should not require a second copy to be trained.

Automatic page selection

When setting up a chatbot, Boei starts with pages that explain your business: the homepage, company information, services, contact, prices and help. It uses a fixed order based on the page addresses.

Sync now, available when your account includes Auto Sync, uses the same order for new pages. On Starter without Auto Sync, add and train pages individually. If some pages need more training space, the result and sync email explain this. Free up space and train those pages when you are ready. It shows whether the update is queued, checking the website or finished, and refreshes the results automatically. If part of the website cannot be read, open View error details for the full address and reason. Pages that did work remain available. Use the close button to hide the warning for now. The error details remain saved, and a new failure will show a warning again.

Websites in more than one language

Train the pages in each language you want your chatbot to use. Each version uses its own training space. A link to another language version does not add or train that page automatically.

When choosing a source for an answer, Boei can group language versions of the same page. It uses language links supplied by your website. For older pages without this information, it can recognise matching addresses such as /nl/pricing and /de/pricing.

The chatbot prefers a suitable version in the visitor's conversation language when that version was found for the question. Only trained pages the chatbot can access are considered. A question about a particular country or a specific page can need a different version. The source link below an answer may point to that language version.

Prices, currencies, availability and conditions can differ between versions. Boei uses the facts from the selected version. For example, a Dutch page can list prices in dollars. Keep each country's information clear and up to date, and train the pages for the countries you serve.

Existing pages gain the saved language information the next time they are trained or synced. To refresh a page now, select it and choose Train chatbot. Use Sync now to refresh a website source.

PDF and other file links

PDFs and other downloads found through a sitemap remain visible, but are not automatically trained. Select them and choose Train chatbot when you need their contents. Documents you explicitly selected before continue to receive updates.

Supported online document URLs use document processing. A document discovered as a website URL stays in that website source. Boei does not automatically download every PDF merely linked somewhere in an HTML page; its address must be discovered or supplied.

Find and manage content

Website pages show their full address. Hover over an address to see the page title.

Within a website, Boei skips a repeated domain folder and combines empty folder levels, such as docs / articles / payments. A folder with just one page shows that page directly, unless it has a saved automatic-training setting. Folders with their own index page or saved setting stay accessible. Separate domains remain distinct.

Folder counts include all matching pages, including pages beyond the first screen. Select a folder to select its matching pages. Turn off View → Group pages into folders for a flat list. This choice is saved for the chatbot.

Selecting a folder selects its current matching pages. To choose what happens to future pages, open the folder’s ··· menu and use Automatically train new pages in this folder. This setting requires Auto sync. It applies when the website next syncs, and does not change pages already trained. If the website uses manual sync, choose a sync schedule for automatic updates. PDFs are always left for you to select.

New web pages are included by default. You can turn this off for a folder. Changing a parent folder also changes the setting for its subfolders; you can then give a subfolder its own setting. If Auto sync access ends, saved automatic folder rules cannot start training new pages.

You can select more pages and choose Train chatbot while training is running. Boei adds the new pages to the same training. Pages already in that training are not added twice. If training is paused because your space is full, the new pages wait too. Free up space, then resume training.

When you free training space, pages that were stopped only because they did not fit show Ready to train if there is now enough room for that page. Select them to start training. Real reading errors remain Action needed.

Use the source's search field and View filters. Open a page to inspect the available extracted content and actions. Use ··· for deletion and update settings. Pages accepted into a running training show Waiting to train until processing starts. During processing, Training has a small pulsing dot. The dot stays still if your device uses reduced motion. A paused training shows the reason in the progress bar above the list. Other status labels distinguish content ready to use, content not trained and items needing action. Open an item needing action for its reason.

Website groups and individual pages have automatic-update settings. If your account includes Auto Sync, use Sync now for a whole-source check. Starter without Auto Sync supports individual page updates. Use page training when you need to read and process a particular page again.

Improve what is extracted

Boei removes common boilerplate. Set CSS filters in the source's ··· → Website settings if you need more control. If reading fails, inspect the error before retrying; see Crawl errors.

Check missing answers to find real customer questions that need better content.