Run it your way.
Same rules everywhere.
Xema is one codebase. Your deployment decides how it runs: composed into a few processes, or split into separate services, on your Kubernetes cluster, on a developer machine or on an appliance with no network at all. Packages install from a catalogue, organisation by organisation, and every one of them answers to the same authorization engine. Where it runs changes how you operate it. It does not change the rules.
One codebase.
The deployment picks the shape.
Whether components run together in one composed process or as separate services is decided by the deployment, never by the author of a package. The code is the same either way: a service reaches its peers in the same way, whether they share its process or run somewhere else.
1Composed or split
Run components together to keep a small installation small, or split them into separate services to scale each one on its own. Changing the shape is a deployment decision, not a code change.
2Kubernetes, in your cloud or on your servers
Install on your own cluster with the Xema command-line tool and Helm charts.
3Local, for development
A compose setup runs the platform on a developer machine, from the same code.
4An appliance, with or without a network
Xema on a single on-prem machine, with local model inference, backup and restore. Without a network, it installs from a signed offline bundle, and nothing in the bundle is loaded until its signature checks out.
Install a package
like an app.
An organisation picks a package from a catalogue and installs it. Updates follow the policy you set: apply them automatically, or ask first when a new version would widen what the package is allowed to do. Asking first is the default.
1More permissions means a person agrees
A new version that asks for more than the installed one had waits for someone to agree to it. The version you have keeps running in the meantime.
2Not ready, and why
If a package needs something that is not there yet, it shows as not ready, with the reason and the action that fixes it.
3Everything else keeps serving
A package that is not ready is set aside. The platform and every other package keep running.
Each organisation
is its own cloud.
Platform packages serve every organisation on the installation. Organisation packages are installed by one organisation into its own space. Each organisation has its own people, agents, data, permissions and AI usage.
1Platform packages
Installed once for the whole installation. They are the shared base every organisation works on.
2Organisation packages
Installed by an organisation into its own space, with a data schema and fences of their own. Another organisation that installs the same package gets its own schema.
3Its own cloud
People, agents, data, permissions and AI usage belong to the organisation they are in.
Packages declare.
Xema enforces.
A package states its resources, the actions on them and what it requires. Xema enforces those declarations with one authorization engine, the same way for every package and every organisation.
1Declared by the package
Resources, actions and requirements are part of the package. Xema reads them and enforces them.
2Changes apply live
Grant access or add someone to a team, and it takes effect at once. No restarts.
3Approvals go to a person
When an action needs a human decision, it goes to a person through one decision service.
Now see what you can build on it.
Deploy decides where Xema runs. What runs on it is packages, and every one of them installs on the same foundation.
At a glance
- the deployment picks the shape
- Kubernetes, local or appliance
- install from a catalogue
- ask first when permissions widen
- not ready says why
- each organisation its own cloud
- one authorization engine