Before you start
Before you start, sign in to the dashboard and select the organization you want to work in.- You need a knowledge base with at least one source. See Create a knowledge base.
- Your role must allow resource management to add, edit, or delete documents.
The Documents tab
Open the knowledge base from Knowledge in the sidebar. The page opens on the Documents tab. The left pane lists every document, and the right pane previews the selected one. A bar above the panes shows the last sync time and the page count. It also holds Manage media and, for roles that manage resources, Sync management. When edits or deletions leave a retriever behind, an amber alert reads A retriever needs reindexing. With several, it counts them. Its Reindex button switches to the Retriever tab. While an import runs, a progress banner reports the scraped page count. See Import a website for the import flow, the page review, and the sync schedules.The document list
The list header reads Pages.
When the knowledge base has no documents and no import, the pane shows No content yet with an Add source button.
Each row shows the document title and its page address, Uploaded file, or Document. A row carries a status badge only while it needs attention. A document without a badge is ready.
During the first import of a website, the list shows Importing pages and fills in as pages are scraped. During a later sync, the list keeps its current rows until the crawl finishes.
The preview pane
Select a row to load it on the right. The header shows the title and the source address. After a document has changed, an activity line gives the last update time, the cause, and the revision number. A sliding toggle switches the content view:
A document still parsing shows Parsing this document…. A failed one shows This document failed to parse. with the stored error.
Roles that manage resources also see actions in the header:
Edit and delete are disabled while a retriever builds. The menu then notes Locked while a retriever is building.
A scraped page that you edited and did not pause shows an amber warning above the content. It states that a website sync can replace your edit with the live page.
Edit a document
1
Select the document
The preview loads on the right.
2
Open More actions and select Edit
The pane switches to an editor with the title field above the full text.
3
Change the title, the text, or both
The body cannot be saved empty.
4
Select Save changes
Retrievers keep answering from their current index until the next reindex.
Pause website sync for one page
Select Pause website sync in the preview header of a scraped page. The row badge switches to Sync paused, and future fetches of that website skip the page. Select Resume website sync to let syncs update it again. Pausing protects one page. To change a whole source’s re-fetch schedule, use Sync management. See Import a website.Delete a document
Open More actions in the preview header and select Delete. The Delete document? dialog explains that the document keeps answering until the next reindex applies the removal. Confirm with Delete. A toast confirms Document removed, and the row leaves the list. The Apply pending changes dialog counts the removal until a reindex runs.Manage media
Select Manage media above the document list to open the Manage media dialog. The Media search switch at the top embeds the images and videos of this knowledge base so the agent can find them. It is off by default, and turning it on uses extra vector storage. With media search on, the agent can show matching images and videos in its replies. Roles that cannot manage resources see an Enabled or Disabled badge instead. Turning the switch off stops new media from being embedded.
Each row shows a thumbnail, the media type, the alt text, the URL, and the documents that reference it. A No vector badge marks media that is not embedded yet. Hover over the badge to see the error, when there is one.
Describe an image or video
A short description helps search find media whose alt text is missing or vague.1
Find the image or video
Use the row in Manage media, or hover over the media in the document preview and select Add context.
2
Type a description in the manual context field
The field takes up to 4000 characters, but only the first 500 are embedded.
3
Select Save
With media search on, the affected documents are embedded again shortly.
Reindex after changes
Edits and removals reach live answers only through a reindex. Each affected retriever switches to Needs reindex, and the amber alert counts them on both tabs. A stale retriever keeps answering from the index built before your changes.1
Open the Retriever tab
Or select Reindex in the amber alert on the Documents tab.
2
Select Reindex on a stale retriever card
The Apply pending changes dialog opens.
3
Review the counts
The dialog counts Changed documents and Pending removals.
4
Select Reindex now
The retriever keeps answering from its current index until the new one is ready.