KitOps

61 posts in KitOps

Run Qwen3 4B on 60% of a GPU with HAMi and KitOps

Run a Qwen3 4B model on a fractional GPU slice on Kubernetes, with the vLLM server and weights shipped as one signed, reproducible OCI artifact. A walkthrough using Project HAMi for GPU virtualization and KitOps for model packaging.

ai - 10 min read

Why MCP servers and agent skills need an enterprise registry

MCP servers and agent skills change what agents do, yet they spread through copied folders and chat links. Here is why they need a governed enterprise registry, and how KitOps v1.13, Jozu Hub, and Agent Guard deliver one.

ai - 10 min read

Package the Agent, Not Just the Model: Native Skills Support in KitOps

Package agent skills, configs, and model weights as versioned ModelKits with KitOps v1.12.0. Push to Jozu Hub and serve locally with Jozu Rapid Inference Containers.

ai - 1 min read

Business Wire – Jozu Assessed Awardable for Department of Defense Work in the P1 Solutions Marketplace

Jozu has been assessed as Awardable in the U.S. Air Force’s Platform One (P1) Solutions Marketplace, enabling streamlined acquisition by Department of Defense customers across connected, on-premises, and air-gapped environments.

ai - 17 min read

Running a Local Coding Agent with OpenCode and Jozu Rapid Inference Container (RICs)

Learn how to package a quantized LLM as a ModelKit, deploy it locally with a Jozu Rapid Inference Container, and connect it to OpenCode to run a fully private AI coding agent on your own hardware.

ai - 7 min read

Signing Is Not Enough: Why AI Artifact Provenance Needs to Be a Graph

Signing your AI models isn’t enough. Learn why fine-tuned model provenance requires graph traversal, not just attestations, to close the supply chain gap.