UNOFFICIAL/EXPERIMENTAL
This guide does not redesign the catalog. Treat the catalog and chart pages as evidence to read, not as pages to rewrite.
Product Boundary
The public front door supports work -> OCI, OCI -> work, and OCI -> work -> OCI without requiring a ConfigHub account. ConfigHub begins when a user claims the configuration so it can be saved, changed, approved, promoted, or rolled out.
Treat work as a reusable operation, not a product or a mandatory first step. It may mean render, inspect, explain, test, scan, compare, or edit. The operation may sit before an OCI is built, after one is pulled, or between an input OCI and an output OCI. Local and GitHub Actions use are proven. A hosted anonymous service is planned and must not be described as available.
Inside an existing delivery flow, distinguish ConfigHub's three roles:
- congruent pass-through: keep the specs and user-supplied metadata, add only the recorded
confighub.com/originprovenance annotation, and publish a new OCI artifact; - transformation: create a reviewed variant and publish the changed objects;
- fan-out: publish specific outputs or deliver one reviewed release to several targets.
Require input and output digests and a field comparison for each pass-through claim. Do not infer congruent output from a successful import, or fleet rollout from one controller becoming healthy.
Do not treat every OCI artifact as interchangeable. Check whether a link is a source or installer package, a literal bundle for cub variant upload, a portable deployment bundle, or a ConfigHub release. The consumer and receipt must match the artifact.
For source-neutral examples, start with config-catalog/program.yaml and docs/reference/config-catalog-doctrine.md. The AICR example keeps separate receipts for generation, OCI packaging, ConfigHub upload, required-approval behavior, and promotion. Do not use one of those receipts to claim a later stage.
For the no-account boundary, use runs/serverless-oci-gitops-proof/receipt.yaml. It proves one public installer OCI input, isolated local rendering, one temporary output OCI, and Flux reconciliation. It does not prove hosted output or ConfigHub operations.
For the CI execution mode, use runs/anonymous-oci-ci-proof/receipt.yaml. It proves anonymous public OCI pull, rendering, object inspection, local OCI-layout creation, and pull-back comparison in GitHub Actions. The output is a workflow artifact, not a public registry package.
For the first continuous upgrade path, use runs/redis-upgrade-app-proof/receipt.yaml. It proves one Redis package upgrade, one retained post-render replica change, two sequential environment promotions, and the same portable OCI digest on two Argo CD clusters. Keep its three product limits: the promotion dry-run printed no mutations, the portable registry was temporary, and the observations were not submitted to ConfigHub observation storage.
What Not To Touch In This Pass
Do not edit these unless the user explicitly asks for catalog work:
site/charts/*.htmlsite/charts/index.htmldata/master-catalog-matrix/*recipes/*/*/*/CATALOG.mdpackages/*/*/*/README.md- generated chart evidence under
data/
If a catalog output is wrong, fix the source generator or source data in a separate catalog task.
Fast Read-Only Routes
| Question | Start with |
|---|---|
| Can I use this chart today? | data/chart-use-guide/summary.md |
| Where are this chart's evidence files? | data/chart-evidence-router/summary.md and data/chart-evidence-router/router.csv |
| What is the broad row status? | data/master-catalog-matrix/matrix.html or data/master-catalog-matrix/matrix.csv |
| What render-intent file exists for a real base? | data/helm-render-intents/summary.md and data/helm-render-intents/intents.csv |
| Which claims are backed or refused? | data/claims-register/summary.md |
| What known gaps affect the row? | data/model-gap-workdown/summary.md, data/target-prerequisite-workdown/summary.md, and data/live-parity-decisions/summary.md |
Inspect One Chart From The Terminal
Use rg before opening large files:
rg -n "bitnami/redis|redis" data/chart-use-guide data/chart-evidence-router data/helm-render-intents
Then open the smallest matching source:
- chart-use row for the short recommendation;
- evidence-router row for paths to receipts and decisions;
- render-intent row for the base config;
- matrix row for lane status.
Read A Render Intent
Start with:
rg -n "bitnami/redis|redis" data/helm-render-intents/intents.csv
Then open the matching YAML under data/helm-render-intents/intents/.
A render intent records the chart version, base, values profile, namespace, release name, capability profile, source lock, lifecycle routes, target prerequisites, and evidence links for a real base row. It is not a production promise.
How To Report Catalog Findings
When summarizing a chart, include:
- chart and version;
- recommended or inspected base;
- current route: use now, review, improve base, blocked, or refused;
- strongest evidence lane;
- known target prerequisites or lifecycle routes;
- next action and source file.
Do not claim more than the source row claims.