We already have CI scaffolding that pulls generated artifacts at release time and lands them in docs (used for API specs and CRD schemas). The same pattern should work for Helm chart values documentation, but we haven't done it yet.
The main candidate is helm-docs, but two things have blocked it:
- It's Go-based, which adds a tooling dependency to the docs build that isn't there today.
- Its output format (Markdown tables of values with descriptions) would need post-processing to fit the site's style and structure.
This matters for the Enterprise configure guides in particular: we're currently maintaining hand-written field tables for the Enterprise Manager, Cloud UI, and Registry Server charts. These will drift. Enterprise also can't be handled by linking to the source repo since the chart is private — generated output checked into docs is the only viable path.
Scope: affects ToolHive OSS charts and Stacklok Enterprise charts. The pipeline approach is the same either way; the Enterprise chart just has the additional constraint of not being publicly linkable.
Starting points worth evaluating:
- Whether helm-docs output can be massaged in a post-processing step to match the site's style
- Whether a lighter alternative exists that's not Go-based
- Whether the existing CRD generation scripts (
scripts/extract-crd-schemas.mjs) could be extended or adapted
We already have CI scaffolding that pulls generated artifacts at release time and lands them in docs (used for API specs and CRD schemas). The same pattern should work for Helm chart values documentation, but we haven't done it yet.
The main candidate is helm-docs, but two things have blocked it:
This matters for the Enterprise configure guides in particular: we're currently maintaining hand-written field tables for the Enterprise Manager, Cloud UI, and Registry Server charts. These will drift. Enterprise also can't be handled by linking to the source repo since the chart is private — generated output checked into docs is the only viable path.
Scope: affects ToolHive OSS charts and Stacklok Enterprise charts. The pipeline approach is the same either way; the Enterprise chart just has the additional constraint of not being publicly linkable.
Starting points worth evaluating:
scripts/extract-crd-schemas.mjs) could be extended or adapted