Build a dashboard
Build a real-time, filterable dashboard straight from a SQL database — the kind of thing you'd reach for Superset, Metabase, or Grafana to do.
1. Connect your database
Add a Postgres or MySQL connection in the Data panel (encrypted, read-only). See Databases & SQL.
2. Add datasets
For each metric or table, add a SqlData brick with a SELECT and a target key — for example
sales.byMonth, orders.recent. Aggregate in SQL (GROUP BY, date_trunc) so the database does the work
and only summarized rows cross the wire.
3. Add filters
Add Select / Input controls with bindKeys (e.g. filter.region, filter.from), and reference them in
each query's params ({ from: "filter.region" }, WHERE region = $1). Changing a filter re-runs the queries
and updates every chart — see Datasets & filters.
4. Visualize
Bind charts to the datasets — StatCard for KPIs, LineChart / TimeSeries for trends, BarChart / PieChart for breakdowns, DataGrid for tables — and lay them out in a Grid. See Charts & visualizations.
5. Go real-time
Set a dataset's refresh interval (say, every 5 seconds) and the bound charts update live, Grafana-style. TimeSeries is ideal for streaming metrics.
Just describe it: "a sales dashboard from my Postgres — revenue by month, top regions, and a recent-orders table, with a region filter, refreshing every minute." The agent connects the pieces; you refine.