Data Quality Agent

The Data Quality Agent audits the health of your data so problems surface before they reach an answer. It runs an exhaustive check across every metric in your semantic layer, and every dimension within each metric, rather than sampling a few and assuming the rest are fine.

Where the Analysis Agent answers "what is happening in the business?", the Data Quality Agent answers "can I trust the numbers behind that answer?".

Where to find it

Start a conversation by clicking New chat in the sidebar and selecting Data Quality Agent. The full report it produces is saved as an artifact in the Library, so you can revisit or share it.

What it checks

For each metric, evaluated at daily grain unless the metric is defined at a coarser grain, the agent looks for:

• Freshness: is the metric up to date, or has it stopped updating?
• Recent breaks: has the metric's series moved in a way that signals a broken pipeline rather than a real business change?
• Dimension breaks: do individual dimension values look broken (missing, collapsed, or spiking)?
• Definition and dimension changes: has the metric's definition or its set of dimension values shifted since the last run, which can silently change what a number means?

Every metric and every dimension gets its own verdict backed by real values, nothing is skipped, sampled, or summarized as "covered elsewhere."

What you get

The agent produces a structured data-quality report: which metrics passed, which have issues, and, where it can, the likely root cause traced back through the pipeline. Use it as a recurring health check on the semantic layer, and as a starting point for what to fix in the Context Engine or with the Data Modelling Agent.

The check depends on your pipeline having run successfully to produce its data-quality inputs; if those aren't available, the agent tells you what's missing instead of guessing.