Knowledge base articles are stored as a custom post type with their own category taxonomy. They are public and have their own archive, so each article gets a real, linkable URL you can send to customers.
Adding an article:
- Open Knowledge Base and use Add New Article.
- Give it a title, a short excerpt and a body. The excerpt is what appears in search results and in the suggestion box on your submit form, so write it as a summary rather than leaving it blank.
- Save. The article is published and indexed immediately.
How the index works:
On save, the article's title, excerpt and body are stripped of markup and split into chunks of roughly 400 words. Each chunk is stored with a keyword list built by lowercasing the text, dropping words shorter than three characters, removing common stopwords and keeping the 40 most frequent terms.
Searches are scored by counting how many chunks contain each of your search terms, then ranking articles by the total. It is keyword matching, not embeddings, which means it needs no external vector service and no extra API spend. The practical consequence is that your articles should use the words your customers actually use, including the error text they will paste in.
Where the index is used:
- The resolver's knowledge base answers, top five articles.
- The drafter's context, top four articles.
- Pre-submit suggestions on the contact form, top three articles.
The existing articles table lists the 50 most recently modified articles, each linking to its public page, with a delete button. Re-saving an article always rebuilds its index entries.