How Onyxia catalogs map to Helm repositories and how to customize them.
Onyxia ships with a set of official service catalogs.
If you don’t configure anything, these are the defaults:
Interactive services (IDEs)
Databases
Automation
Data visualization (optional)
As an instance admin, you can heavily customize what users see and can do:
Change defaults for a service (resources, images, features).
Apply different policies per user group (example: who can request H100).
Fork our catalogs or build your own.
Turn any Helm-deployable software into a service.
Example: Doom launched as an Onyxia service.
If you already know Helm, most of this will feel familiar.
A Helm repository is a collection of Helm charts.
A Helm chart is a recipe to deploy software on Kubernetes.
Charts expose configuration via values.
Defaults live in values.yaml.
values.yaml
Example: values.yaml (jupyter-python).
When installing a chart, you can override any default value.
Charts can also ship a values.schema.json.
values.schema.json
This JSON Schema describes:
which options exist
the expected types / formats
constraints (min/max, enums, patterns, …)
Example: values.schema.json (jupyter-python).
You configure which Helm repositories Onyxia should load as catalogs.
If you don’t configure catalogs, Onyxia loads the defaults from catalogs.json.
catalogs.json
On the “Service catalog” page:
Each Helm repo becomes a tab (Interactive services, Databases, Automation, …).
Each chart becomes a service card (Jupyter, RStudio, …).
When a user opens a service:
Onyxia reads the chart’s values.schema.json.
It renders a form from the schema.
It generates a final values object that Helm will apply.
values
Onyxia can also inject user-specific defaults. For example, it can prefill S3 credentials:
You have two main customization paths:
Instance-level overrides (recommended for most setups)
Bring your own catalogs (or a fork of ours)
Was this helpful?
