Knowledge Sources
Knowledge sources decide what the agent can answer. Manage them under Knowledge → Links and Knowledge → Documents.
Where content comes from
- Your site domain — when an agent is created, siteaiagent crawls the site starting from the home page and auto-discovers sitemaps from standard paths and robots.txt.
- Sitemap links — add sitemap URLs and siteaiagent indexes the pages they list.
- Page links — add individual page URLs from the Links screen.
- Uploaded documents — upload
.txtor.mdfiles from the Documents screen.
Sign-in, sign-up, login, and registration pages are never indexed.
Keeping content current
Use Update content on the Links screen:
- Incremental update only fetches newly added links and documents.
- Full update re-fetches every indexed page to pick up content changes. Pages whose content has not changed are detected by a content hash and skipped.
Each link shows its status while an update runs: Pending, Updating, Failed, or Skipped.
What indexing stores
Expand a link to see the raw indexed content plus Retrieval metadata: a per-chunk summary, extracted facts, suggested questions, and search terms. siteaiagent enriches every indexed page automatically so the agent can find the right passage — it builds search metadata in both English and Chinese, extracts key facts, and prepares the questions each passage can answer. No configuration is needed.
Content hygiene
- Remove outdated prices, policies, and guarantees.
- Avoid internal-only content.
- Prefer short, focused documents over large mixed files.
- Run an update after important website or policy changes.