Evaluation

Continuous re-evaluation and drift detection keep your context accurate as data changes.

Starter page — expand with the evaluation dashboard and drift workflow.

The open-source harness gives you a one-shot eval delta you run locally. The hosted product turns that into a continuous system so your context stays trustworthy as the underlying data and definitions change.

What the hosted evaluation adds

• Continuous re-evaluation — re-run the eval pairs on a schedule, not just once, and track accuracy over time.
• Drift detection — flag when an upstream dbt model a definition depends on changes, so a human re-confirms before the agent trusts stale context.
• Correction harvesting — turn analyst corrections into new labeled ground truth that feeds the next evaluation.

Together with observability, this is how data teams keep context correct as they scale self-service analytics. For a demo, requirements, or pricing, contact sales at [e-mail].