You can add various types of documents to become the knowledge of the Bot.

  1. Files: Supports multiple formats such as .doc/.docx/.pdf/.md/.txt.
  2. Spreadsheets: Supports formats such as .xls/.xlsx/.csv.
  3. URL: Provide the URL of the website, and the system will crawl the web content as Bot knowledge.
  4. TXT: Directly enter content online as knowledge.
  5. Q&A: Knowledge type structured as "Q&A (question and answer pair, that is, one question and one answer)".


Different types of documents provide different slicing methods. You can preview the slicing effect on the page.

If you don't understand slicing, it is recommended to use the system default values.


For files, URLs, and TXT types, you can choose two different slicing methods:

  • By Token number: Slice based on the set Token number. For example, if it is set to "1000", a slice is generated for the document every 1000 Tokens.
  • By identifier: Slice based on the selected identifier. For example, if it is set to "\n", that is, a newline character, a slice is generated for the document every time a newline occurs.


For spreadsheet type, you can choose multiple rows as the header. The system will slice the table data based on the "header + row" method.


For Q&A type, there is no need to choose a slicing method, the system will automatically treat each Q&A as a slice.

Copy from other Bot


You can copy the document from other Bots within the same organization to the current Bot, making it the knowledge document of this Bot.



For documents that have been added, they can be edited online.

Documents of the spreadsheet type are not currently supported for online editing.

URL Update


For URL type, updates can be made to obtain the latest content of the webpage as knowledge.


If there are too many URL documents in the Bot, you can also set up "Scheduled Updates". The Bot will update all URL document content for you in full at the time you set.


You can choose to update your URL document daily, weekly, or monthly.