The health dashboard
The admin surface where you triage your knowledge base’s worst problems in minutes — score, read, act.
Open Evergreen AI from the Confluence app menu to reach the dashboard. It has three tabs — Overview, Findings, and Scans — plus a Scan now button and a link to Settings.
Overview tab
A fast read on the health of everything in scope.
- Stat cards — open findings, confirmed this week, pages scanned, and coverage (how many of your spaces are in scope).
- Spaces table — one row per space, sortable by health. Each row shows the health score, open-finding count, the worst finding type, the last scan time, and a link to view its findings.
The health score is transparent: hover the score to see the exact formula. In short, it starts at 100 and subtracts a penalty for each open finding, weighted by the finding’s confidence and its type, then normalizes for the number of pages. No black box — a low score always traces back to specific findings you can open.
Findings tab
The working list. Every finding across your scanned spaces, with the tools to cut it down to what matters.
- Filter by space, type, status, and minimum confidence, and search within results. Filters are reflected in the page URL, so you can bookmark or share a filtered view with another admin.
- Expand any row to see the verbatim evidence quote, the reasoning, the related page (for contradictions), and the triage actions.
- Pagination is server-side at 50 findings per page, so the list stays fast even with thousands of findings.
Permission-safe by design. If a finding sits on a page you don’t have permission to view, the row shows only its type and space with the note “A finding exists on a page you can’t view.” No title, no quote, no link. Evergreen AI never reveals page content through a finding to someone who couldn’t already read that page.
Triage actions
From an expanded finding (or in bulk — see below), you can take one of four actions:
| Action | What it does |
|---|---|
| Confirm | Marks the finding as a real problem worth fixing. Moves it to CONFIRMED. |
| Dismiss… | Opens a short modal: add an optional reason (up to 200 characters) and, if you like, tick “This page is fine — reduce future flags like this.” Moves it to DISMISSED. |
| Assign… | Opens a user picker (licensed users only) to hand the finding to a page owner. Their avatar then shows on the row. |
| Mark resolved | Records that the underlying page has been fixed. Moves it to RESOLVED. |
Bulk actions: select multiple findings with the checkboxes and assign or dismiss them all at once — useful for clearing a batch of the same false positive.
The suppression learning loop
This is what keeps Evergreen AI from crying wolf. When you dismiss a finding and tick “reduce future flags like this,” the app remembers your reason and feeds it into future scans of that page. It won’t raise the same thing again. Dismiss once, with context, and the noise goes away for good. You can review and undo these in Settings → Suppression list.
Scans tab
A ledger of every scan the app has run, drawn from its own records:
- Scan date, scope, and status — Running, Partial, Complete, or Paused — budget.
- Pages planned versus completed (a progress bar while a scan is running), findings produced, tokens used, and a cost-band indicator.
- Expand a row to see any per-batch errors — useful when a scan came back partial.
For how scans are scheduled and how incremental scanning works, see Scans & scheduling.
The four states you’ll see
Every dashboard surface is designed for the messy real world, not just the happy path:
- Loading — skeleton cards and tables, never a bare spinner.
- All clear — a designed empty state that celebrates a clean space and shows coverage and the next scheduled scan.
- Error — an inline message with a retry and a ledger reference id, never a blank screen.
- Degraded — a clear banner when scanning is paused at budget (“triage remains available”) or when the latest scan covered only part of a space and will resume next window.