Core concepts

A short glossary for the terms used throughout these docs.

Connector

A connection to one external database, warehouse, or API - the credentials and options needed to reach it (host, user, database name, and so on). Every query runs against exactly one connector. Access to a connector is granted per group, so a group only sees the connectors it’s been given.

Query

A saved SQL statement against a connector. A query can take parameters (named placeholders filled in at run time), can hold reusable snippets, and can be put on a schedule so its result refreshes automatically instead of only on demand.

Queries have version history: every save creates a new version, and a query can be in a draft state - edited but not yet published - so you can iterate without changing what dashboard viewers currently see.

[!NOTE] Running a query can return an immediate cached result or hand you back a job to poll, depending on whether a fresh run was needed. See the async query model if you’re calling this through the API rather than the editor.

AI features

Optional features backed by an organization-configured AI provider. Ask with AI is an agent that turns a plain-language question into SQL: it drafts, runs a read-only preview, checks the result, retries on errors, and hands you the final query to review (AI-assisted SQL generation). Auto Visualize picks a visualization type and configuration for a query result. A data source’s AI profile (table and column descriptions, relationships, column statistics) gives these features context. Suggest checks proposes data-quality monitors, and Why did this fire? adds an explanation to alerts. Monitors can also be threshold-free (baseline), which is plain statistics with no AI call. Most of what AI features send is schema-level metadata; sampled column values and alert-investigation query results (the model is asked to aggregate them) are the exceptions, and AI Settings can switch them off. Data handling & privacy lists exactly what each one sends. All AI features are off until an admin adds a provider key in AI provider configuration.

Visualization

A rendering of a query’s result: a table, a line/bar/area/pie/scatter chart, a counter, or a funnel. A visualization always belongs to one query, but the same query can have several visualizations.

Dashboard

A grid layout of visualizations (called widgets once placed on a dashboard), optionally with its own parameters that filter every widget on the page at once, regardless of which query backs each one.

Alert / Monitor

A threshold check against one column of a saved query’s result - for example, “notify me if this value goes above X.” Alert and monitor are the same thing: the API and the object itself are always called a monitor, while “alert” is the everyday name for what it does. Notifications go out over email, Slack, webhook, PagerDuty, Mattermost, Teams, Discord, Google Chat, or Datadog.

There’s no separate “data observability” mechanism - freshness, row count, null rate, and duplicate checks are just this same monitor mechanism pointed at a query written to answer one of those questions. See Data observability monitors for the query recipes.

Sharing

Every query and dashboard has its own access control: per-user or per-group permissions, a public secret-token link (org-only, or open to anyone when created as anonymous), or an iframe embed for pasting into another page.