What is different here
Four things this static build cannot do, and what it does instead.
This is the real dashboard, with real components, running real queries against a database that was invented on purpose. What deploys is a directory of static files: every page was rendered at build time and no server runs behind this URL.
What that buys#
A demo anybody can click, without anybody hosting a resolver, and with no possibility of a real household’s browsing history being published. It costs four things.
1. The range picker is fixed at 24 hours#
The range arrives as ?range=7d, and a static page cannot read a query string: reading one forces per-request rendering, which is the thing this build does not do. The alternative was prerendering six copies of the whole dashboard behind a path segment.
The fixture still holds three weeks. Every page here is reading the same twenty-four hours of it.
2. Control and Alerts are gone#
Both write to a resolver, and there is no resolver behind this URL. Rather than mock nine settings screensthe settings screens and a notification engine that cannot send anything, they were removed.
In the product, Control is where you change what the network resolves, behind confirmations that measure the blast radius against your own history first; Alerts is a rule engine with six notification transports and a delivery log that records what was suppressed and why.
3. The explorer filters a sample#
2,500 of the most recent rows, filtered in your browser, rather than one indexed query against the whole table. It was kept rather than cut because every summary elsewhere links into it pre-filtered, and a demo where forty of those links go nowhere is a demo of a different product. The page states which number you are looking at.
4. The live stream is a replay#
The arcs on the map and the rows in the live log are drawn from a pool of real fixture rows, stamped with the moment they appear rather than when they were generated. The components doing the drawing are the product’s own, unmodified: same coalescing, same pace handling, same arc lifecycle, same reconnect logic. Only the transport is different.
What is real#
Every query on these pages is the product’s own SQL, running against the product’s own schema, through the product’s own rollup pass. The charts, tables, attribution, geolocation, anomaly detectors and friction analysis are unmodified. If something looks wrong here, it is wrong there.