Add a website as a knowledge source
Use Web Sitesi Ekle on the Dökümanlar tab to submit an address. The crawl runs in the background, so judge it by the row badge and chunk count.
Chatinbox looks for sitemap.xml first, and reads only the home page when it finds none. There is an upper limit on how many pages are read. Pages are fetched as plain HTML; content loaded later by JavaScript and parts such as the top menu and the footer are left out, so add your address and phone number as SSS entries. The crawl is a one-time copy; when a price or opening hour changes on your site, delete the WEBSITE row and add the site again. Pressing Siteyi Tara closes the dialog at once, but that is not a finished crawl. The work carries on in the background. Nothing in the panel shows progress, a page counter or a crawl log, so the Hazır badge and the chunk count on the new row show that processing finished. They do not guarantee that the whole site or a given piece of information was captured.
Before you start
- An account that can open Bilgi Bankası; this area draws no admin and agent split.
- A publicly reachable domain served over http or https, such as ebrumoda.com.
- A robots.txt that does not block the crawler, because disallowed paths never reach the list.
Step by step
These screens sit inside the Intent AI section. Sign in, open Intent AI; the menu names below are in that section's left menu.
Open Web Sitesi Ekle from the Dökümanlar tab
Open Bilgi Bankası from the left menu. With the Dökümanlar tab selected, press Web Sitesi Ekle.
Type the address and confirm the detected domain
Type the domain into the Web Sitesi URL field. A line below reads ebrumoda.com taranacak; press Siteyi Tara once the domain shown is yours. Leave Başlık empty and the record takes the host name.
Read the chunk count on the new WEBSITE row
A row tagged WEBSITE appears in the Dökümanlar list once the dialog closes. When the badge turns Hazır, read the parça figure next to it, which counts the text chunks the crawl produced. A Hata badge means the site source could not be processed.
Result checkIf it did not work
Check these
- URL gerekli or Geçerli bir domain girin (örnek: ebrumoda.com) means the address could not be read. Enter one bare domain with no spaces.
- URL resolves to a private or loopback address and URL hostname could not be resolved arrive in English. Internal addresses are refused on purpose, so use an address that opens from outside your network.
- A Hata badge means the site source could not be processed. Possible causes: the site does not open from outside, its pages are drawn by JavaScript, or robots.txt blocks the crawl. Fix the cause, delete the row and add the site again; if no cause can be found, write to support.
- A price or opening hour you change on your site does not reach the knowledge base by itself. Delete the WEBSITE row and add the site again. When the new row turns Hazır, try price and hours questions from a test number.
Does this step look different in your panel?
We can review the screen and exact error text together.
This guide was checked against the product UI and code inventory on 4 October 2026.
