Custom Dashboards

Build any visualisation from your telemetry data by writing SQL and choosing a chart type. Dashboards are saved to your tenant's D1 database and shared across your team.

Creating a dashboard

Click + New Dashboard, type a name, and press Enter or click Create. The empty dashboard opens immediately — add your first panel.

Adding panels

Click + Add panel to open the panel editor. Give it a title, optionally a description (shown as a tooltip on the panel header), pick a starter template or write your own SQL, choose a chart type, and size it on the 12-column grid.

Quick-start templates

The panel editor has a row of one-click templates for common queries — click one to fill in the SQL and switch to a matching chart type, then tweak it however you like:

Schema browser

Click Schema next to the SQL editor to open a sidebar listing every queryable table and column — click any column to insert it into your query. The same list, kept in sync with the actual database schema, is reproduced below.

Panel SQL

Any SELECT query against your hot index tables works:

-- Request rate over time
SELECT
  (timestamp / 3600000) * 3600000 AS bucket,
  service,
  COUNT(*) AS requests
FROM recent_spans
WHERE timestamp > 1785150241875
GROUP BY bucket, service
ORDER BY bucket ASC
LIMIT 500

-- Error rate by service
SELECT
  service,
  COUNT(*) AS total,
  SUM(is_error) AS errors,
  ROUND(SUM(is_error) * 100.0 / COUNT(*), 1) AS error_pct
FROM recent_spans
GROUP BY service
ORDER BY error_pct DESC
LIMIT 100

-- p99 latency from metrics
SELECT
  service,
  AVG(p99) AS avg_p99,
  MAX(p99) AS max_p99
FROM metrics_samples
WHERE metric_name = 'http.server.request.duration'
  AND timestamp > 1785085441875
GROUP BY service
LIMIT 100

Chart types

Display options

Line, bar, and area panels support additional display config:

Saving

After adding or editing panels, click Save in the top right. The dashboard is saved to D1 and immediately available to all team members.

Full schema reference

Every table available to panel SQL, and every column on it: