Technology · Ebook

KitOps: Packaging AI/ML as OCI Artifacts with ModelKits

by Shriira Press

KitOps is an open-source, CNCF-governed DevOps tool that solves a quietly painful problem: the assets of an AI project — model weights, datasets, inference code, prompts, and configuration — usually live in scattered, incompatible places with no shared versioning or security. KitOps consolidates them into a ModelKit, a standardized, immutable, OCI-compliant package defined by a simple Kitfile manifest and built with a Docker-like kit CLI. Because a ModelKit is just an OCI artifact, it lives in any container registry, versions and signs like a container image, and unpacks selectively so pipelines pull only what they need. This book builds KitOps up from the packaging problem through ModelKits, the Kitfile, the CLI, layer internals, distribution, Cosign signing, the ModelPack spec, and real MLOps integration.

Contents

1. 1Preface
2. 2Chapter 1 — The Packaging Problem in AI
3. 3Chapter 2 — ModelKits and OCI Artifacts
4. 4Chapter 3 — The Kitfile Manifest
5. 5Chapter 4 — The Kit CLI Workflow
6. 6Chapter 5 — Inside a ModelKit
7. 7Chapter 6 — Registries, Tags, and Distribution
8. 8Chapter 7 — Security and Provenance
9. 9Chapter 8 — KitOps in Practice