The generated household

Sixteen invented devices, and the stories each page was given something to tell.

everyoneAssumes a home network and that you have heard of DNS.page 21 of 24

Sixteen invented devices with the names somebody would actually give them, resolving the domains those devices actually resolve, at the hours they are actually awake.

Why it is not random#

A demo dataset made of device-01 through device-16 resolving example-domain-4.com proves the charts render and nothing else. The product’s whole claim is that it turns a log into things you recognise as belonging to your home, and a fixture that cannot be recognised cannot demonstrate that claim.

The household#

Living room Apple TV    Kitchen Echo Show     Hunter's MacBook Pro   Hunter's iPhone
Nintendo Switch         Bedroom Samsung TV    Front door Ring        Hue bridge
Roborock S8             PlayStation 5         Kitchen Sonos          Pixel 8 (guest)
Brother printer         Synology NAS          Work ThinkPad          Model 3

Each has an hourly rhythm curve and its own domain mix. The work laptop goes quiet at weekends; the console does the opposite; the doorbell never sleeps.

The stories it was given#

Set piece	Which page it makes work
The bedroom TV hammers its ad endpoint every evening, refused every time	Friction
The doorbell retries a blocked metrics host in short, frequent bursts	Friction, as a recurring shape rather than a loop
The guest Pixel looks up dns.google and the Firefox canary	Anomalies
The Model 3 stopped resolving three days ago	Anomalies, gone silent
The Pixel first appeared three days ago	Anomalies, new device
Four domains first appear in the last two days	Domains
One upstream is deliberately slower than the other two	Health
Endpoints pinned to real cities across five continents	DestinationsDestinations — which a real Pi-hole install does not have

That last row is the one place the fixture is more generous than your install would be. It carries answer addresses because it was generated to exercise every page; FTL does not report them, so on Pi-hole the destinations map, the country columns and the city labels are absent rather than sparse. What is different here says the same thing about the demo as a whole.

How it is built#

npm run demo:generate   # ~417k queries → real rollups → prune to 7 days of detail
npm run demo:snapshot   # extract the client JSON (stream, search, explorer)
npm run build           # prerender ~435 pages against the fixture, emit out/

Traffic is generated per device, per hour, against that device’s curve, with a weekend multiplier and jitter. Blocked verdicts come from a per-domain policy: some names are never blocked, some always, and a few are blocked most of the time but not all of it, because that middle is where a filtering resolver actually lives.

Then the product’s own runRollups() aggregates it hour by hour, and detail older than seven days is deleted. Seeded throughout, so the same commit produces the same database.