Nested sitemaps, blocked-page detection, and URL suggestions when training.

Training from a website follows sitemap indexes all the way down, tells you when a page was blocked rather than silently learning the block page, and suggests the right URL to start from.

New: Sitemap indexes are followed into every child sitemap, so sites that split their sitemap by type or month are trained in full. A Review pages step lets you exclude pages before training starts.
New: Pages that returned a firewall or bot-protection block page are flagged as Blocked instead of being trained as content. Filter by them, retrain, or delete them from the Sources page.
New: When you add a website, HelpJet suggests the best URL to train from (for example the WordPress site root rather than a single post).
Fixed: Links in bot replies now open on sites that run click-blocking optimizers such as NitroPack.
Fixed: The widget header menu (new chat and related actions) works again.