Live database dashboards, built into the IDE
Most database tools show you a database as it is right now. SnoutData's dashboards show you a database as it moves: named sets of SQL-backed widgets that poll on an interval and plot the trend, right next to the editor that wrote the query.
A dashboard is a named set of widgets, and each widget is a SQL query on a timer. Point one at whatever you want to watch (rows written per minute, replication lag, the size of a queue table) and SnoutData runs it on an interval and charts the result. No exporting to a separate observability stack, no leaving the app you already have your connection open in.
Two kinds of widget
A snapshot widget re-runs its query and replaces the chart each time, for when you only care about the current shape of the data. A trend widget samples a scalar or grouped metric over wall-clock time, accumulates the series, and plots how it moves. The polling and rendering are fully deterministic: no model runs on the hot loop, so a dashboard is cheap to leave open all day.
The AI is the analyst, not the poller
The loop that collects the numbers stays deterministic on purpose. The AI comes in on demand, as an analyst over the accumulated series: ask what a movement means, and it reads the trend you have been recording rather than guessing from a single snapshot. Deterministic signals do the watching; the model does the explaining.