Get startedOperate · Continuous assurance
Run conformity and monitoring
Check live behavior against a declared production contract and investigate drift or policy failures.
Source reviewed 2026-09-16. Availability depends on your installation, permissions, and compatible runtimes. A supported path is not a guarantee of model quality or production readiness.
Outcome
Know whether the live system remains within its approved quality and operating envelope.
Start here
What you need to know first
Check the inputs below before starting. If you are new, begin with the first-project guide. A dataset holds media and labels; a model produces results; a deployment makes a selected model version callable. Creating one does not create the others.
Bring these inputs
- — A released artifact and declared production contract
- — Reference and live evidence windows
Expected output and limits
- — A conformity run and pass/fail findings
- — Linked remediation evidence
Core workflow
Define the contract
Choose the release, benchmark, thresholds, slices, and operating constraints.
Run conformity
Launch the durable check and follow progress from Tasks or the model monitoring page.
Inspect evidence
Review failures, affected slices, drift signals, and linked production examples.
Resolve the finding
Accept, rollback, adjust policy, or route examples into annotation and a new version.
What this surface supports
Release conformity checks
Production health
Evidence-linked failures
Feedback routing
Expert section
Contracts, signals, and failure modes
Use this section when you are defining acceptance criteria, automating the surface, or reviewing whether its output is safe to promote downstream.
Quality and operating signals
- — Threshold breaches
- — Input and prediction drift
- — Critical-slice regression
- — Policy violation rate
Common failure modes
- — Stale reference distribution
- — Thresholds without operational meaning
- — Alerting without an owner or remediation path
Expert release checklist
□ Inputs and dependencies are pinned to immutable versions.
□ Acceptance metrics include critical classes and operating slices.
□ Failure, retry, cost, and rollback behavior are understood.
□ The resulting artifact has an owner and a downstream review path.
Engineering safeguards
- — Checks identify their release
- — Failures remain actionable
- — Remediation does not silently rewrite history