One place that says what is stuck
Nexview has around thirty settings pages. Not one of them answered the question “is everything actually running?” — you had to click through them in turn. The dashboard answers it in ten seconds: what is waiting for somebody, and what is broken.
Not a thirty-first page, but the way into the thirty
Four figures at the top, each of which asks for an action: requests nobody has decided on, requests in flight, open tickets, unanswered feedback. Below them the findings — and at the bottom three buttons into the areas they come from.
The point is explicitly not to build another page with things on it. It is to make the thirty existing ones unnecessary as long as nothing is wrong. When something is wrong, every finding leads from here to exactly where you fix it.
For administrators, nobody else. An approver never sees the dashboard — instance state and disk figures are not needed to decide on a request.
A finding is an exception, not a state
Anything permanently true is a statistic and belongs in the analysis — not in a list you look at daily. “You have 3,505 films” is true and always true, and that is exactly why it is not here: a list that always says the same thing stops being read after a week.
So the same rule holds for each of the twenty checks: if it does not apply, it says nothing. No “all good” tick, no empty table, no nudge that you could connect a media server. If there is nothing, nothing is there.
And every finding names the consequence, not the measurement. Not “callback silent for 9 days”, but: Nexview learns late that something has finished downloading — at the next round instead of straight away. Things still work, just more slowly. Anybody reading that knows, without prior knowledge, whether they have to do something tonight.
Not “Settings”, but Settings → Services → Sonarr
Every finding carries a target, and the target is never a bare page. Press “View” and you land on the tab where the problem sits — for a stuck request on the filtered list, for a silent callback at exactly the instance that is quiet.
It sounds like a detail and it is the difference between a notice and a help: a link that drops you on an overview page and leaves you searching costs more time than it saves.
A test holds that in place — no finding may point at a page without a tab. The addresses themselves are deliberately readable: ?reiter=dienste&unter=sonarr tells a human where they will land.
Requests that have been “searching” for two weeks
In the interface “searching” looks like work, not like a standstill. Which is precisely why the most common silent failure — indexer gone, request sitting there for weeks — has until now only surfaced when somebody complained.
Nexview counts along: if a request has said “searching” for more than fourteen days and nothing is in the queue for it, that is not searching any more. That is a finding.
The thresholds are fixed and reasoned in the source — fourteen days for stuck, three days for a waiting approval, ninety percent for the disk. Deliberately not a thirty-first settings page: anybody who would want to tune those numbers would have to have lived through them first.
Twenty checks across six areas
Services — is the instance answering, since when not, is the version old, does the callback arrive, does Radarr or Sonarr report a problem itself. Their own messages stay in their original English wording: it is their statement, not ours.
Supply — what is stuck, what is waiting for approval, what failed repeatedly in the last seven days, where the queue needs a hand.
Disk — how full, and how fast it is filling.
Library — entries with no file behind them, two instances sharing a download category, titles nobody has watched for months.
Source reconciliation — where Radarr, the media server and Nexview's own books disagree. In full on the Statistics & analysis page.
Operations — backups overdue, mail given up on, errors in the log, diagnostic level left on by accident, a new version available.
Three parts instead of two — and how long it will last
The ring separates media, everything else and free. “Used versus free” would be a half-truth: backups, photos and the operating system live on the same disk, and only “media” can be cleaned up through Nexview. Anybody who thinks they can free 36 TB because it shows as “used” is in for a disappointment.
Next to it, the question that actually matters: will it last? “91 percent full” tells an operator that they will have to do something soon — but not whether “soon” means next week or next year. Nexview measures once a day and draws a curve from that.
While there are too few readings, it says so: too early for a curve, two more daily readings needed. A line through two points would not be knowledge, it would be a claim.
A button on the home page
Administrators get an Admin dashboard button in the top right of the usual home page; the menu under your own name carries the same entry, with a badge as soon as a finding reaches the “error” level.
Nobody else sees any of it — for them the home page stays exactly as it was.
A tile for Homepage, Homarr and friends
If you use a home board as your start page, Nexview can sit on it with real numbers: open approvals, findings by severity, library size, the state of every instance. One call, one answer:
GET /api/v1/dashboard
Authorization: Bearer nxv_…The address lives under /api/v1, and that is a promise: as long as v1 is in the address, nothing disappears from this answer. The fields are listed one by one on the API page, and ready-made snippets for Homepage and Homarr are in docs/dashboard-tile.md.
⚠️ The token has to belong to an administrator — instance state and disk figures are an operator's business, an ordinary account gets 403 here. The may only read switch limits it to GET, but it may then also read the user list, the log and the settings. On your own machine that is fine. Before it goes on a screen other people walk past, think twice.
