Bring container-like rigor to AI: package models, datasets, code, and configuration together as a single versioned, signable OCI artifact you can push to any registry.

Welcome to KitOps: Packaging AI/ML as OCI Artifacts with ModelKits.

KitOps is an open-source DevOps tool, now governed by the CNCF, that solves a quietly painful problem: the assets that make up an AI project — the trained model, the datasets it learned from, the inference code, the prompts, the deployment configuration — usually live in scattered, incompatible places, each with its own versioning and security story, none of them tied together. KitOps consolidates them into a ModelKit: a standardized, immutable, OCI-compliant package defined by a simple YAML manifest called a Kitfile, built and moved with a Docker-like command-line tool called the kit CLI. Because a ModelKit is just an OCI artifact, it lives in any container registry you already run, versions like a container image, signs like one, and unpacks selectively so a pipeline can pull only the model weights or only the prompts. This book opens with the packaging problem AI teams face, then builds KitOps up piece by piece: what a ModelKit is and why OCI was the right foundation, how the Kitfile describes a project, the everyday CLI workflow of init, pack, push, pull, and unpack, the layer-and-digest internals that make selective unpacking and deduplication possible, how registries and tags handle distribution, how Cosign signing and the CNCF ModelPack specification give you provenance, and finally how KitOps slots into real MLOps and DevOps pipelines. By the end you will understand not just how to use the kit CLI, but why packaging AI the OCI way changes how teams ship it.

This title is part of the ShriIra library and is free to read in full, right here — our small contribution to making world-class knowledge easy to reach.

A note on reading it: open the Contents menu at the top of the reader to jump between chapters, use the Aa menu to set a comfortable text size, theme (light, sepia, or night), and single- or two-page layout. Your place is saved automatically, so you can always pick up where you left off.

We hope it serves you well.

— Shriira Press

Contents

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