Fix and refresh Library items

Find items that failed or are out of date, see why, and extract them again.

Last updated 28 Oct 2026

Agents answer from the text extracted from your Library items. When an item fails, or the source behind it changes, here's how to find it and bring it up to date.

Find items that need attention#

On the Dashboard, Needs attention → Failed Library items opens Library filtered to Failed. The card appears only when something has failed.
In Library, filter by extraction status (Pending, Processing, Ready or Failed) and by type.
The Content Extraction column shows each item's status and when it was last extracted, such as "Extracted 3 days ago".

See what went wrong#

Open the item's row menu:

Action	What it shows
Extraction summary	The status and diagnostics of the last run, including Last error. For websites, it also shows crawl progress, such as discovered, processed and failed pages.
View content	The item and the text agents actually read from it

Common causes#

What you see	Likely cause	What to do
A website captured fewer pages than expected	The crawl reached Max pages (500 per link when blank) or Max depth, or a path rule left pages out	Raise Max pages (up to 5,000) or adjust the paths with Edit, then choose Extract content
Pages from one part of your site are missing	Your site's robots.txt disallows them, or they're behind a login	Allow them in robots.txt, or add the material as a File or Text Snippet
The crawl stays on one subdomain	Crawl boundary is Same hostname	Switch to Same domain, or add a link for each subdomain
A file won't upload	It's over 10 MB or not a supported type	Split or convert it. See Add files, Text Snippets and Q&A.
An item shows Failed after a long wait	A website crawl ran past 2 hours, or a file extraction past 30 minutes	Narrow the crawl with Max pages or paths, then choose Extract content

Extract again#

Choose Extract content from the row menu to refresh an item:

Websites: pages fetched in the last 24 hours are reused, and older pages are crawled again. If you changed a page right after a crawl, run it again the next day to pick up the change.
Files: a file extracted in the last 5 minutes keeps that result.
While an item is Processing, choosing Extract content doesn't start a second run. If the run has stopped making progress, it's marked Failed and a new run starts.

After you edit an item#

When you change a Website's links or settings, or the text of a Text Snippet or Q&A, saving with Edit extracts it again on its own. Changing only the title, description, tags or project doesn't. To refresh an item whose content you didn't change, choose Extract content.

Remove what's out of date#

Delete removes an item and everything extracted from it, so agents stop answering from it. Use it for material you've replaced, such as last year's price list.

Check Failed Library items on the Dashboard after you add a large website. A failed item doesn't stop your agents, but they can't answer from what's missing.

Was this helpful?

Still need help?

Ask the community