Member post by Gorkem Ercan, CTO, Jozu
The merry band of maintainers and contributors at the KitOps project is happy to announce the 1.0 release of KitOps. Jozu kicked off the KitOps project after observing the significant challenges due to the fragmentation of AI/ML packaging and versioning. KitOps, which has been submitted to the CNCF Sandbox, is an open-source packaging and versioning solution designed for speed, security, and consistency in a variety of computing environments. The tool outfits AI/ML practitioners with a standardized OCI artifact called ModelKit, leveraging the widely embraced OCI standard. This enables organizations to manage, secure, and audit AI/ML projects effectively.
KitOps has one grand ambition: to bolster collaboration among data scientists, application developers, and SREs. Therefore we have created tools around ModelKit to serve all these groups. KitOps releases a CLI named kit which is used on desktops, in pipelines, and by tool integrations. We also release a GitHub Action and a Dagger module for use in AI/ML workflows. For data scientists, the PyKitOps library and the new MLFLow plugin provides a streamlined mechanism to share their work as ModelKits on any OCI registry of their choosing.
Since launching, KitOps has been installed over 45,000 times, and is being used in production by private enterprises with global footprints and security-conscious public sector development teams.
New in 1.0: Hugging Face Import or Generated ModelKits
One common use case of KitOps is assembling a secure and curated set of base models for internal consumption from third-party sources like Hugging Face. To make this even more easy, the kit CLI now includes a shiny new import command. Point it at a Hugging Face repository, and—presto! it generates a ModelKit ready to be served from your enterprise OCI registry.
But your favorite model is not always hosted on Hugging Face. The new kit init command automatically scans a directory and generates a Kitfile recipe. The Kitfile instructs the CLI exactly how to package everything into a ModelKit so you can securely share your AI/ML project with the rest of your crew in minutes.
This 1.0 release is more than features and a version bump—it’s a statement. The core features KitOps pledged to deliver are complete, and production-proven. The community has hammered away at the interfaces until they were ready for heavy-duty production usage.
This release signals that KitOps isn’t just a great idea, it’s a critical tool for today’s enterprise AI/ML teams.
A Note of Gratitude to Early Adopters
We owe a hefty round of applause to our early adopters and the community that rallied around the project. KitOps benefited immensely from user feedback from across the globe and in varied industries. Early adopters found KitOps a breeze to integrate with existing infrastructure—no new clusters, no hidden subscription fees, no forced migration to the trendiest cloud du jour. Being able to utilize existing security, access controls, and audit trails in OCI registries has been a game-changer, especially for organizations that care about data security, and regulated industries that need tight provenance and attestation measures.
As we celebrate this milestone in the KitOps journey, we invite you to download KitOps 1.0 today and slot it straight into your AI/ML development pipelines. We welcome new contributors—troublemakers, dreamers, and coders alike—to help shape our shared vision. Check out the GitHub repo, open an issue, or join our Discord community to speak your mind. Together, we’ll drive KitOps toward the future it deserves.