Keeping your knowledge base answer-ready

The Auto-Responder and the Widget Answer agent are only as good as the articles they draw from. Helmdesk uses semantic search — matching on meaning, not just keywords — so a visitor who asks "how do I log in with Google?" finds your "Single sign-on" article even if it never uses the word "login."

How retrieval works

When you publish or edit an article, Helmdesk computes a vector embedding of its content.

A question is embedded the same way and compared against your articles by semantic similarity.

The most relevant published articles are passed to the agent, which grounds its answer in them and cites them.

If semantic search is unavailable, Helmdesk falls back to keyword search, so retrieval still works.

Only published articles are searchable — drafts (including agent-drafted articles awaiting review) are never surfaced to customers.

Writing articles agents can use

Answer one question per article, and make the title that question.

Be self-contained — the agent answers from a single article, so don't rely on the reader having read another.

Use clear headings and short paragraphs so the most relevant passage is easy to retrieve.

Keep them current — re-editing an article refreshes its embedding automatically.

The payoff compounds: every good article improves your public knowledge base, your widget answers, and your Auto-Responder drafts at the same time.