All products

Build Directus dashboards with AI

Put counts, time series and filtered lists together in Insights. The dashboard uses each viewer’s own access to the underlying collections.

To build Directus dashboards with AI, start with the decision the reader needs to make. A content lead checking whether the team is on schedule might want a count of published articles next to a three-month trend. The directus-insights skill inspects the collection before choosing the date field or guessing the names of its statuses.

A panel can save successfully and still show nothing. The instructions explain which settings each chart accepts, how much grid space it needs and what to check when a colleague sees a different total.

What it does

  • Count published records or calculate an aggregate over a filtered collection.
  • Group a time series by the dates that matter to the team.
  • Add a product selector which filters related panels by record ID.
  • Arrange a list and a headline metric without overlapping the panels.
  • Diagnose a blank chart using its options and the viewer’s permissions.

Install and use

  1. Install the skill and connect an authorised Directus instance through REST or the official SDK. You need permission to build dashboards and read the source fields.
  2. Ask, “Show dispatched orders this week and a trend for the past quarter. Use our actual dispatch timestamp.”
  3. Open the result as an intended viewer. Compare one total against the records available to that account.

FAQ

Source

Made by Directus (Monospace Inc.). Original skill: directus-insights, published on GitHub under the MIT licence.

View on GitHub