Discover

Discover is where analysis happens in Celiq. You choose a reveal (a curated starting point in your semantic model), pick the attributes and metrics you care about, optionally add filters and calculations, then click Run. Celiq compiles the SQL, runs it against your warehouse, and renders the result as a chart or table. Because every query is built from your Forge definitions, it automatically respects your Data Keys, Data Gates, and role permissions — you never have to remember to scope a query to your access level.

Discover is also the editor you use inside Mosaics. When you edit a Mosaic tile, the same Discover surface opens pre-loaded with that tile's query, so there is one editing experience to learn, not two.

Overview

The Discover editor is split into three working areas:

• Left — the FieldPanel. Select a reveal, then browse and search its fields. Fields are grouped into TIME, DIMENSIONS, METRICS, ML, and CALCULATED sections. This is also where you set the grain and create calculated (derived) fields.
• Top — the action bar. Shows the reveal name, row limit, row count and query time, plus Ask Orion, Run, and a ⋯ menu for saving and exporting.
• Center/right — results and configuration. A tabbed area showing Results (chart or formatted table), SQL, History, Comments, and Notes. The ChartConfigurePanel opens on the right to style the chart, pick a chart type, or ask the Orion chart copilot.

Everything you build is captured as query state in the URL (a short query ID), so a refresh restores your work and the link is shareable.

When to use it

Reach for Discover when you want to:

• Answer an ad-hoc question against a reveal ("orders by week for the last quarter").
• Build a chart you intend to save as a Prism or pin to a Mosaic.
• Add a derived metric (AOV, margin) without changing your CML model.
• Drill from a summary chart into the detail behind a single point.
• Edit an existing Mosaic tile — the tile editor is Discover.

If you only need to read an existing dashboard, open a Mosaic instead. If you need to change the underlying data model (add a metric to the semantic layer permanently), use Forge.

Concepts

Getting started

• You have the Tracer, Weaver, or Luminary role (the editor requires the run_discover permission).
• Your workspace has at least one reveal. Reveals are authored in a domain in Forge — if none exist, the FieldPanel shows a "No reveals yet" prompt linking you to Forge.
• The reveal's project has an active warehouse connection.

Step-by-step

Examples

1. Select the Orders reveal.
2. Add the time field Created At and the attribute Channel.
3. Add the metric Revenue (leave the aggregation as SUM).
4. Set the grain to Week.
5. Add a filter: Created At → is in the last → 90 days.
6. Click Run. Open the chart picker and choose a Line chart, with Channel on color.

To add average order value without touching your model, create a calculated field in the FieldPanel using existing field names in square-bracket syntax:

Label:      Avg order value
Expression: [revenue] / [orders]
Format:     Currency ($)

Calculated-field expressions are evaluated per row in the browser and only allow safe arithmetic (+ - * / ( )); non-numeric values resolve to 0. For cumulative, ranked, or period-over-period math computed in the warehouse, add a window calculation from the Calculations section instead (Running Total, Rolling Average, Rank, Lag/Lead, % of Total, and more).

The generated SQL for the example above looks similar to this (your dialect and column names will vary):

SELECT
  DATE_TRUNC('week', created_at) AS created_at_week,
  channel,
  SUM(revenue)                   AS revenue
FROM orders
WHERE created_at >= CURRENT_DATE - INTERVAL '90 days'
GROUP BY 1, 2
ORDER BY 1
LIMIT 1000;

Best practices

• Start from the right reveal. The reveal determines which nodes (and therefore which fields and joins) are available. Switching reveals is faster than fighting the wrong one.
• Keep the row limit sensible. The limit control offers 100 / 500 / 1,000 / 5,000 / 10,000. Use a small limit while iterating, raise it only for exports.
• Filter before you raise the limit. Filters reduce work in the warehouse; a higher limit does not.
• Define reusable math in CML. Browser calculated fields and table calculations live with this query only. If you'll reuse a derived metric across Prisms, Mosaics, or Signals, promote it to a metric in Forge (the ⚙ Save to Forge action on a window calculation generates the CML for you).
• Edit Mosaic tiles in place. Because the tile editor is Discover, you get the full FieldPanel, chart picker, and SQL view — there is no reduced "tile editor". Use Save to tile to write changes back.

Tips

Common mistakes

Troubleshooting

Related pages

• Chart types — the full catalogue of charts the picker offers
• Query IDs — how Discover saves and restores query state in the URL
• Filters — operators, relative dates, and comparison periods in depth
• Prisms — save a Discover query as a standalone, shareable Prism
• Mosaics — pin a Discover query as a live tile on a dashboard
• Forge — define the reveals, nodes, and metrics Discover builds on