Jozu Hub Overview
What is Jozu Hub?
The Jozu Hub is an OCI 1.1 compliant registry built to host the open source KitOps ModelKits for enterprise AI projects. Jozu Hub can be used as SaaS, installed on-premises, or deployed to a private environment.
Jozu Hub allows you to serve a curated set of models and datasets to your teams through a simple interface, but backed by your existing enterprise registry.
It provides greater security, privacy, and control than public registries like Hugging Face, making it ideal for organizations in any regulated industry or those that value data privacy and cleanliness in their AI/ML projects.
What are ModelKits?
ModelKits revolutionize the way AI/ML artifacts are shared and managed throughout the lifecycle of AI/ML projects. ModelKits:
- Encapsulate datasets, code, configurations, and models into a single, standardized unit.
- Can be stored in any container registry so you don't need additional storage or security reviews to start using them.
- Are based on the existing OCI standard which is the same standard containers use, so they're secure and tool-compliant.
The KitOps CLI allows you to only unpack what you need from a ModelKit - just the model and dataset for testing, or perhaps just the code for integration.
The ModelKits on Jozu Hub make a great starting point for enterprise AI/ML projects. The Discover page lists the most popular, trending, and newest models, or you can use the Browse page or the search bar to find what you need. If you don't see a model you'd like to use send us a note at [email protected] with your wishlist.
You can learn more about ModelKits and the Kitfiles used to define them in the documentation for the open source KitOps project.
Why use Jozu Hub?
You can store ModelKits in any container registry, but only Jozu Hub will allow you to see the contents of ModelKits, and auto-generate inference microservices for them across multiple platforms. Jozu Hub simplifies discovering, sharing, using, and tracking projects within teams or across organizations. It has ModelKits for many of the popular AI and ML models and datasets, from LLMs, to predictive ML, computer vision, and others.
How do I get started?
We have several short examples from our Getting Started page.
How do teams use Jozu Hub and ModelKits?
Typically operations teams will make the creation of a ModelKit a pre-requisite for any AI/ML project going to production. Pushing a versioned ModelKit to an OCI registry like Jozu Hub ensures that the operations team has everything they need for validation, deployment, rollback, or other operational tasks.
Sharing up-to-date changes in a project is also easier with Jozu Hub - everyone can see not only the list of ModelKits and versions, but even what files are inside each ModelKit making it easy to distinguish what has changed version-to-version. For example, in the event of a production issue and a needed rollback.
For organizations that subject to regulatory checks, the Jozu Hub makes it easy to audit AI/ML projects. The EU AI Act, for example, requires organizations training or deploying models be able to provide the data state as it was used to train the model for up to 10 years post-training. By saving each AI project version as an immutable ModelKit and storing them in the Jozu Hub, organizations can quickly go back to any version and see exactly what data and state was used for training.