Version 3.0
Dashboard

The Dashboard tab

Open a Data Collection and you land on the Dashboard. It is the owner's seat: build health, what is scoped, what is grounded, and what people have been asking the collection lately, all on one page.

Audience: builders

Where the Dashboard lives

Every Data Collection has three primary tabs: Dashboard, Setup, and Catalog. The Dashboard is the default. Open /data-collections/<slug> and you land there unless your URL pins another tab.

The first time you land after creating a collection with no sources yet, the Dashboard renders the first-run guide instead: a single call to action that opens Setup. The full Dashboard kicks in once at least one source is attached.

Read the attention banner

A banner sits at the top of the Dashboard whenever the catalog needs your attention:

  • Catalog build in progress. Our AI is building or rebuilding the catalog. The banner shows the phase message and links to Open setup so you can watch phase-by-phase progress.
  • Catalog build paused. The last run failed. Click Open setup to read the failure detail and re-run.
  • Pending ontology proposal. A build finished and a proposal is waiting on you. The banner offers Discard and Review proposal.

If nothing needs attention, the banner does not render. Quiet is good news here.

Read the summary tiles

A row of six tiles leads the Dashboard:

  • Entities. Catalog entity count, with a subline for relationships.
  • Sources. Total sources, broken down into connections and file uploads.
  • Selection. Scoped relations, with a subline for scoped columns.
  • Grounded. The percentage of scoped tables our AI placed in the catalog. Below 90% turns warn-toned.
  • Last build. Duration of the most recent build, with the status as a subline.
  • Proposals. Pending ontology proposals (zero, or one awaiting review).

Above the tiles, the page subtitle reads the summary sentence: for example, "12 entities, 8 relationships, 3 metrics across 2 sources, 14 scoped relations, 92 scoped columns." That sentence is your one-line state-of-the-collection.

Grounded versus scoped. Scoped tables are the ones you selected in Setup. Grounded tables are the ones our AI used in the live catalog. A grounded percentage under 100% means some scoped tables did not make it in: open Build the catalog for the coverage workflow.

Read the two-column body

Below the tiles, the Dashboard splits into a left column for activity and a right column for sources and health.

Left column.

  • Recent activity. The last six intent runs against this collection. Click a row to expand the summary, see badges, and (where present) click through to the task. Full reference: Recent activity and intent history.
  • Recent builds. The last five catalog builds with phase reached, duration, and outcome. Failed builds show the section that failed and the error message.

Right column.

  • Connected sources. Every source feeding the collection, with its alias and a status dot.
  • Source health. Profile signals from each scoped relation. Warnings sort to the top; healthy relations carry an OK badge.

Move between tabs

The Dashboard does not edit the catalog. Two natural jumps land you where editing happens:

  • Open Setup. From the attention banner or the Project's tab strip. Use Setup to add sources, change scope, or queue a build.
  • Open Catalog. Tab strip again. Use Catalog to read the map, review coverage, and walk the entities our AI proposed.

What to do when a build fails

The attention banner reads "Catalog build paused." Click Open setup to land on the catalog step. The step shows the failed phase, the section it failed inside, and the error message. Address the cause (more business context, drop a misnamed table, retry the upstream connection) and click Build catalog again. Nothing in your live catalog changes until a new build succeeds and you apply its proposal.