Kapitan generator vs Helm chart vs Kustomize

Kapitan offers several inputs for the same goal, files rendered from the inventory. They differ in where the knowledge lives.

You have	Use	Why
An upstream chart whose values are the real interface	`helm` input, chart pinned	The chart owner maintains the logic; a generator around it only renames values
A file that mirrors the inventory almost 1:1	`jinja2` or `copy`	A template is shorter than code and readers see the output shape
One entry that must become several consistent resources	kadet generator	Names, labels, ports and references are derived once, in code, and tested
The same manifests with per-environment patches	Kustomize, or classes in the inventory	Overlays express “the same, except”; a generator would hide the base
A format with structure YAML templating gets wrong (JSON dashboards, Terraform JSON)	kadet generator	Building data structures avoids quoting and indentation errors

Signs you picked the wrong one

Your jinja2 template has loops over loops and {% if %} on every field: the logic wants to be code; a generator is easier to test.
Your generator’s inventory block has the same keys as its output: it adds a second schema to learn without adding knowledge. Ansible playbooks are the classic case; they are YAML already.
You wrap a chart in a generator to set five values: use the helm input and keep helm_values in the inventory.

Mixing them

One target can use all of them: the helm input for the operator’s chart, a generator for the custom resources you configure through it, and jinja2 for a README. Generators on kadet-hub are fetched like any other OCI dependency, so they combine with the rest of your compile entries.