ModelOps User Guide

Register, version, build, deploy, monitor, govern, and optimize models across Kubernetes, Docker, hosted inference, and GitOps targets.

Who This Guide Is For

Where To Go

Page Use It For
/modelops/register-model Register a model from Hugging Face, Git, storage, upload, or Docker.
/modelops/registry Browse models and versions.
/modelops/builds/trigger Build serving images.
/modelops/deployments View deployments.
/modelops/deployments/create Create Kubernetes, Docker, or hosted deployments.
/modelops/monitoring Monitor serving health and performance.
/modelops/gates Configure pre-deployment gates.
/modelops/security Review security posture.
/modelops/promotions Request and review promotion workflows.

Core Concepts

Concept Meaning
Model A registry entry describing source, metadata, runtime, and ownership.
Model version A deployable artifact version with approval, aliases, build references, and lineage.
Build A container image creation and push process.
Deployment A running or managed serving endpoint.
Gate A policy-backed pre-deployment readiness check.

Common Workflows

Register and deploy a model

  1. Open Register Model.
  2. Choose source and review source analysis.
  3. Set identity, visibility, tags, and runtime.
  4. Submit registration.
  5. Build a serving image if required.
  6. Create a deployment and select target runtime.
  7. Configure resources, networking, health, compliance, and GitOps.
  8. Review and deploy.

Promote to production

  1. Confirm model version approval.
  2. Run required evaluations and scans.
  3. Review gate readiness.
  4. Resolve failures or create a justified override.
  5. Submit a promotion request.
  6. Deploy through GitOps where required.
  7. Monitor after rollout.

Best Practices