Validate the model at the right level
Separate schema checks, whole-model semantic checks and validation of actual authored content.
Validation answers different questions at different stages. Be explicit about which checks a tool performs.
Model structure
Section titled “Model structure”The CMMF 1.2.0 JSON Schema checks the structure of an individual model: required properties, known type identifiers and conditional property rules.
The bundled source does not include a separate manifest schema. A project validator needs its own documented manifest and version checks.
Relationships across the set
Section titled “Relationships across the set”Whole-model checks inspect information a single-file schema cannot resolve. These include duplicate IDs, missing reference targets, incorrect object-target kinds, rich-text embed targets and missing editorial label fields.
Run checks on the complete set. A teaching excerpt may deliberately omit fields or targets and should not be treated as a standalone model document.
Declared content constraints
Section titled “Declared content constraints”Model fields can express content constraints such as character limits, numeric bounds, list limits and MIME types. These are intentions to carry into an implementation.
| Category | Common keys |
|---|---|
| Text | minLength, maxLength, regex |
| Number | min, max, integerOnly |
| Lists | minItems, maxItems, uniqueItems |
| Assets | mimeTypes, dimensions, fileSize |
The schema deliberately does not formalise every key/type combination. File-size units and dimension structures need an agreed translator convention. Reference item bounds can be expressed under relation; see the specification for the precise rules.
Authored content and implementation
Section titled “Authored content and implementation”Checking a model does not establish that real articles or assets satisfy its declared constraints. That requires content-instance validation or enforcement by the target system.
A translator should preserve supported constraints and record unsupported or lossy mappings. Editorial usefulness, relationship choices and platform fit still need a review.
The site’s production build checks both complete example sets for schema validity, manifest version, duplicate model/field IDs, editorial field references, relationship targets and coherent constraint bounds. The Customer Story project includes the same checks as a portable local validator. It does not perform content-instance validation or certify CMS mappings.