Knowledge Base

Creating Articles

Add and edit Knowledge Base entries for Keva's AI

Knowledge Base entries are internal reference material. Keva's AI uses active entries when drafting answers, while staff search and maintain the entries in the Knowledge Base.

Add an entry

  1. Open Knowledge Base for the workspace you want to update.
  2. Select Add to Knowledge Base.
  3. Paste or type content, load one .txt or .md file, upload one .pdf file, or select resolved tickets to import.

The Add panel currently uses manual entry. AI analysis is unavailable, so complete the entry details yourself.

Enter the details

For a manual or text/Markdown entry, complete:

  • Title - A clear name for the reference material.
  • Category - One category label used to organize and filter entries.
  • Content - The text Keva's AI and staff can use.

Select Save to create the entry. When editing an existing entry, change the title, category, or content and select Update. The editor accepts up to 500,000 characters of content.

PDF uploads

Select a PDF from the Add panel. PDF uploads are imported into the default category; you can change the category afterwards on the article. Keva extracts its text and uses the file name without its extension as the title. The PDF must contain readable text; an empty, corrupted, or password-protected file is rejected.

Imported ticket entries

Resolved and closed tickets can be imported from the ticket picker. Each imported entry contains the customer question and the reply, with personal information redacted. Imported entries start as AI Suggestions so staff can review them before accepting them as active reference material.

Entry status and indexing

Entries use the statuses active, archived, and suggested. Active entries are available to the AI and staff search. Suggested entries wait for staff review. Archiving keeps an entry but removes it from the normal active list.

The entry card also reports indexing status:

  • Searchable - A current semantic index is available.
  • Indexing pending - Indexing has not finished.
  • Indexing failed - The semantic index could not be created; keyword matches may still work.
  • Not searchable yet - No semantic index exists yet; keyword matches in the title or content may still work.

Keep entries useful

  • Use a specific title and put the answer in the content.
  • Keep instructions and dates current.
  • Remove personal or private information before saving reference material.
  • Check the indexing status after saving or updating an entry.