Manage your generative AI content
Last updated: 09 December 2024
To manage your generative AI knowledge, head over to Knowledge → Generative AI.
You can add (or remove) content from different sources:
Documents
To add knowledge from documents, expand the first dropdown and click + Add document.
Generative AI supports .DOC, .DOCX, and .PDF documents.
Select your document from the modal (or import new ones).
There is a delay between clicking the document and the modal closing. Please be patient!
Websites
To add web-based content, expand the second dropdown and click + Add URLs.
You can add content from three sources:
Single URL: Your platform will pull all the text content from the single URL.
XML sitemap: Your platform will pull content from all the pages it finds in your sitemap.
HTML sitemap: Your platform will start by pulling content from the URL, then follow all the links it finds and pull content from each subsequent page.
Your platform automatically re-crawls all web-based content every Sunday to ensure your chatbot always has access to up-to-date content. To manually refresh the content (e.g. mid-week), click on a URL’s View button, then hit Re-crawl.
FAQs
Nothing happens when I click to import a document.
There is a delay between clicking the document and the modal closing. It is, in fact, importing. Please wait a few seconds.
There is content missing from my website import.
The sitemap importer may only import content it is able to access. If content is missing, it is usually due to one of the following reasons:
The sitemap you have provided is not valid. Make sure it is a public
.xml
sitemap.Access to the sitemap is throttled or prevented (e.g. CAPTCHA, location-based, hidden behind a login page, etc.). Make sure your sitemap is public and that robots are allowed to crawl it.
Your content is hidden under Java syntax. For example, a web page with dynamic pricing tables that displays localised pricing to the user. To achieve this, a website may use Java variables such as
${location.pricing}
which get populated upon page load. This variable is therefore ‘empty’ during the crawl, returning empty content.