Skip to content
SEMANTIX Docs

Data Explorer

The Data Explorer is where you get hands-on with the data before trusting it. It has three tabs: Columns, SQL Playground, and My Requests.

Columns

Section titled “Columns”

Every registered column with its type, its role, and its description.

Sample fetches the distinct non-null values of that one column — a SELECT DISTINCT … LIMIT 20 against it. It is not a preview of the table, and it is not row samples. It is exactly what you want when you need to know what values a filter can take: which regions exist, which statuses are used, how a status column is spelled.

Results are cached for about a minute per column, so re-sampling a column you have just looked at is fast.

SQL Playground

Section titled “SQL Playground”

Write read-only SQL against the entity. The editor comes pre-filled with a SELECT … LIMIT 50 against the entity’s qualified name.

  • Run executes it, capped at 200 rows.
  • Generate with AI writes a first draft from a plain-English description — you review and edit it before it is worth anything. Generated SQL is a draft, not an answer.

Writes are not possible: the statement whitelist and row limits described in read-only by design apply to everything run here.

A connector’s Data Explorer works the same way but is bound to the connector, so a single query can join across any tables it exposes.

My Requests

Section titled “My Requests”

The path from ad-hoc to governed.

  1. Find the query you want in the playground.
  2. Give it a label — the name people will recognise — and optionally a description of what it measures.
  3. Request metric.

An admin reviews it. Approving it creates a Registry metric, and your request links to the metric it produced. Rejecting it leaves the note explaining why.

This is the intended route for adding metrics. It means the person who found the number proposes it, and the person accountable for the number approves it.

Before you request

Section titled “Before you request”

Worth checking the columns are described and marked with sensible roles — see Registry. A query written against undocumented columns is a query that will be misread later, and a metric is meant to be reused.