1. Data feed description
This method relies on the use of AI to generate question/answer pairs.
Our solution analyzes and interprets the content of your articles, integrating text and images, to transform them into individual topic and content records, optimized for our Assistant. This information can be updated manually by either specifying a new URL or restarting the question/answer pair generation process.
2. Assistance in creating the knowledge source
- Give the knowledge source a name.
- Choose the content language.
- Paste the URLs of the pages to import, one per line (200 maximum).
What is read
- Only the page at each URL is read, not linked pages: add one URL per page.
- The header, navigation menu and footer are ignored. Most accordions and tabs are opened before reading.
- Content that only appears after another visitor action may be missed: check the generated content.
- Each URL starts with https://. A URL starting with http:// is rejected when saving, including in an existing source.
3. Error management
The web pages to be ingested must be publicly accessible on the internet, as no additional authentication methods are supported.
Depending on the format of the specified web page, our processing may encounter difficulties in extracting the data. In case of an issue, an error message will appear at the knowledge source level, specifically identifying the URL(s) responsible for the error.
A page that only shows a captcha, a cookie banner or an error page is reported with the Blocked page (captcha, cookie banner or error page) error: no content is extracted from it. Check that the page displays without these obstacles for a visitor who is not logged in.