URL Sources & Refreshing

2 min readUpdated April 6, 2026

URL Sources & Refreshing

When you add a URL to your knowledge base, your agent:

  1. Fetches the page content via Jina Reader (extracts clean text, strips ads and navigation)
  2. Compiles it into a structured markdown summary (~2K chars)
  3. Tags it with topic keywords for relevance ordering
  4. Stores the compiled summary — not the live page

When Sources Go Stale

URL sources are fetched once at the time you add them. They don't auto-refresh. If the page changes (new pricing, updated docs), your agent won't know until you refresh it.

Your admin coaching session will flag stale sources in its Knowledge status line when a URL source hasn't been refreshed in more than 7 days. It looks like:

Knowledge: 3 docs (8.2K chars) · ⚠️ pricing URL stale (12d)

Refreshing a URL Source

In admin chat:

"Refresh my pricing page"

"Update my docs source"

The agent calls refresh_knowledge, re-fetches the URL, and recompiles if the content has changed. If nothing changed, it says so and doesn't update.

Images Can't Be Refreshed

Image sources (diagrams, screenshots) can't be re-fetched — they were uploaded as files, not URLs. To update an image source, remove the old entry and re-upload the new version.

Tips

  • Add your most important URLs first — they get the highest relevance when matching visitor questions
  • Give entries descriptive titles so you can find them easily in list view
  • If a page has multiple sections you care about (e.g., pricing AND FAQ), you can add it as two separate entries with different titles and tags
  • For high-change pages (pricing, changelog), refresh every time you update them