sdc_dcat3_ap

One Semantic Data Charter model, described in DCAT-AP 3.0.1.

sdcdcatap reads a published SDC model's package and writes the dcat:Catalog that DCAT-AP, the DCAT Application Profile for data portals in Europe, asks a portal to exchange: one Dataset per model, the model's schema cited as the standard its records conform to (dct:conformsTo) and as the document that defines their variables (foaf:page), by URL and SHA-256. One rdflib graph, serialized as Turtle and as JSON-LD. The output is checked with SEMIC's own SHACL shapes at a pinned commit and with the Interoperability Test Bed's SEMIC validator.

1. What this is, and where it came from

The sample, samples/nhanes-participant/catalog.ttl (and catalog.jsonld, the same graph), is a Catalog with one Dataset: the records of NHANES Participant (dm-xy8upneajsb8vdcmnve01g6g) from the FAIR Data Demo, a demographic, examination and laboratory record of the National Health and Nutrition Examination Survey (CDC). The model is public:

• the catalog record: https://sdcstudio.axius-sdc.com/api/v1/catalog/dm/xy8upneajsb8vdcmnve01g6g/
• the schema the records conform to: https://sdcstudio.axius-sdc.com/dmlib/dm-xy8upneajsb8vdcmnve01g6g.xsd
• the Semantic Data Charter: https://semanticdatacharter.com

pip install -e ".[fetch]"
sdcdcatap write --package samples/nhanes-participant --out catalog.ttl --jsonld catalog.jsonld
sdcdcatap write --ct-id xy8upneajsb8vdcmnve01g6g --ct-id <another> --save-package pkg/ --out catalog.ttl   # a catalog of models, from the public catalog
sdcdcatap write --package DIR --catalog my-catalog.yaml --out catalog.ttl                                  # a publisher's own declared input

Two kinds of input, kept apart. From the model's package (its JSON-LD, its schema, the public catalog record and the published schema versions; no account needed; a model without a package is refused): title, description, identifier, creator, dates, licence, the schema's URL and SHA-256, the variable count, and the model's own Dublin Core when the modeler wrote it (publisher, subject keywords, contributors, coverage). From the catalog's publisher, declared in catalog.yaml because the profile's controlled vocabularies and range rules demand values the package cannot know: the contact point, the EU data theme, access rights, language, the publisher's ADMS type, the catalog's own title, description and homepage. The package ships a default file for the sample (src/sdcdcatap/data/catalog.yaml); a publisher supplies its own.

Every referenced node is typed and named inline. The range shapes require every object value to be an instance of its class, and the mandatory shapes require a name on every agent and a label on every concept, wherever they sit in the graph. So the publisher is a foaf:Agent with a foaf:name, the theme is a skos:Concept with a skos:prefLabel in its scheme, the licence is a dct:LicenseDocument, the language a dct:LinguisticSystem, and so on. A harvesting portal gets a complete graph.

What the Dataset carries:

dct:title, dct:description	the model's, English-tagged, plus one sentence saying what a governed data record is and that the schema is immutable
dct:identifier, adms:identifier	dm-xy8upneajsb8vdcmnve01g6g, as a literal and as an adms:Identifier with skos:notation
dct:conformsTo	a dct:Standard: the schema at its pinned URL (?sha256=), with the SHA-256 as dct:identifier
foaf:page	a foaf:Document: the schema at its stable URL, described as the data dictionary (153 variables), dct:format XML from the File Type NAL, itself dct:conformsTo the SDC4 reference model
dct:publisher, dct:creator	foaf:Agents with names; the publisher with its ADMS type and homepage
dcat:contactPoint	a vcard:Kind with vcard:fn and vcard:hasEmail, from the declared input
dcat:theme, dct:language, dct:accessRights	HEAL from the EU Data Theme vocabulary, ENG from the Languages NAL, PUBLIC from the Access Rights NAL, from the declared input
dct:license	the model's licence URL as a dct:LicenseDocument
dcat:keyword, dcat:landingPage, dct:issued, dct:modified	from the package
dct:provenance, dcat:version	a dct:ProvenanceStatement: a published model never changes, a revision is a new model naming the one it revised; the SHA-256 as the version

The Catalog carries its title, description, publisher, homepage, language, dates, the Data Theme vocabulary as dcat:themeTaxonomy, and the Dataset.

2. How to verify it

The tests run SEMIC's shapes from data/dcat-ap-3.0.1-4470b8e/, copied at the pinned commit (build/snapshot_dcat_ap.py verifies the pin first), with pySHACL:

pip install -e ".[dev]"
python -m pytest tests -q                   # the Interoperability Test Bed test runs too, over the network
python -m pytest tests -q -m "not network"  # offline, as CI runs it

Offline: the mandatory-property shapes (dcat-ap-SHACL.ttl) and the range shapes (ranges.ttl), both conforms, in Turtle and in JSON-LD. The committed samples are asserted isomorphic to a fresh run.

The second judge, the Interoperability Test Bed's SEMIC SHACL validator (https://www.itb.ec.europa.eu/shacl/dcat-ap/, v1.14.0-SNAPSHOT on 9 October 2026, through its REST API), on the sample:

Validation type	Result
dcatap.3_0_1_base0 (mandatory, no background knowledge)	conforms
dcatap.3_0_1_range0 (ranges, no background knowledge)	conforms
dcatap.3_0_1_codelists (controlled vocabularies)	conforms, 3 warnings: the publisher is not in the EU Corporate Bodies NAL (it is a company, not an EU body)
dcatap.3_0_1_rec (recommendations)	conforms, 19 warnings, listed in part 4
dcatap.3_0_1_full (everything, with background knowledge)	11 violations that are not in the document: see part 4

The network test asserts the first four.

3. What the projection could not say

Left out rather than filled in:

• No dcat:distribution of the records. The FAIR Data Demo's records are not published; the model is. The profile recommends a distribution and does not require one; the writer emits one when a publisher declares an access URL.
• No temporal coverage, no geometry, no frequency, no dataset series, no data service. The package does not carry them. Spatial coverage, subject keywords, contributors and the model's own publisher come from the model's Dublin Core when the modeler wrote them; SDCStudio's field defaults ("Universal" for coverage, "None" for relation, blanks) read as unset, which is why the NHANES catalog carries none of them.
• Nothing a publisher would have to decide. The data theme, access rights, language, the publisher's type and the contact point are declared input, said so here, and defaulted only for the sample.

In the other direction, what the record carries that a catalog entry has no place for, stated so a reader knows where to look rather than as a shortcoming of either side:

• The governance envelope is data in every record. Each Governed Data Record carries its own audit event, PROV activity and PROV agent as validated leaves. DCAT-AP's dct:provenance and prov:wasGeneratedBy describe the dataset; the per-record facts are in the records, defined in the schema the catalog points at.
• The schema is bound to each record. Every record names the schema it was validated against; the catalog says that once, in dct:conformsTo with the SHA-256.
• The reason a value is missing is typed in the record. A leaf may carry one of sixteen exceptional values in place of its value; the schema defines them; which one applies, and where, is in the record.
• There is no data-dictionary slot. DCAT-US 3.0 has describedBy for "the names and definitions of all variables"; DCAT-AP has dct:conformsTo and foaf:page, which is where the schema goes here.

4. What we learned about DCAT-AP 3.0.1 and its validators

Implementer's notes from building this against SEMICeu/DCAT-AP at 4470b8e and the Interoperability Test Bed on 9 October 2026. They describe how the artifacts behave, so the next implementer spends the day on their own catalog rather than on these.

• Type and name everything you reference. The ranges file checks sh:class on every object property, and the base file requires foaf:name on every foaf:Agent and skos:prefLabel on every skos:Concept in the graph, including vocabulary terms you merely point at (the ADMS publisher type, the data theme). SEMIC's own minimal example, with its publisher as a bare URI, passes the mandatory shapes and fails the ranges for exactly this reason.
• The full and base types report violations that are not in the document. An empty minimal catalog (title, description, a named publisher) submitted to dcatap.3_0_1_base returns the same five foaf:name violations on blank-node agents as our sample does, and full adds six dct:title violations on ADMS concept schemes (http://purl.org/adms/publishertype/1.0 and siblings). They come from the background knowledge those types load, not from the content. Read the four groups individually (base0, range0, codelists, rec), each of which reports only on the document; that is what the table in part 2 does.
• The recommendations type treats the Catalog as a Dataset. With DCAT loaded as background knowledge (dcat:Catalog is a subclass of dcat:Dataset), the Dataset recommendations fire on the Catalog too: contact point, distribution, keyword, theme, spatial and temporal coverage, licence. Of the 19 warnings on the sample, nine are those; the rest are the Dataset's missing distribution, spatial and temporal coverage, dct:type on the creator agent and on the licence document (ADMS licence type), and the publisher outside the Corporate Bodies NAL.
• ranges.ttl references property shapes it does not define. 206 sh:property triples point at nodes with no triples of their own. A Jena-based validator passes over them; pySHACL refuses to run ("not a well-formed SHACL PropertyShape"). The tests drop the undefined references before running; the 91 defined property shapes are what is checked.
• The REST API is simple. POST https://www.itb.ec.europa.eu/shacl/dcat-ap/api/validate with a JSON body of contentToValidate (the document as a string), embeddingMethod: STRING, contentSyntax, validationType and reportSyntax; the answer is a SHACL validation report. /api/info lists the validation types and the validator version. JSON-LD and Turtle gave identical results.
• dcat:version is a DCAT 3 property that rdflib's closed DCAT namespace does not know; build the IRI by hand.
• The language is a NAL concept, not a tag. dct:language takes http://publications.europa.eu/resource/authority/language/ENG, a dct:LinguisticSystem; a BCP 47 string fails the ranges.

MCP server

sdcdcatap is also an MCP server, registered as io.github.SemanticDataCharter/sdcdcatap, so an assistant that can read a published model with sdcreader can ask for its DCAT-AP 3.0.1 description. One tool, write_catalog, over the same function the CLI calls; stdio; read-only apart from the document it returns; no account; no tool takes a URL (the catalog host is fixed at start with --host). The protocol scaffold is sdcreader.mcp, shared by every SDC projection server.

uvx sdcdcatap mcp            # or: sdcdcatap-mcp, or sdcdcatap mcp

Layout

• The model's package is read with sdcreader (load_package, fetch_package, read_model): the record tree, its leaves, the enumerated values with their codes, the model's Dublin Core with SDCStudio's defaults as unset. The package format is documented there, once.
• src/sdcdcatap/: dcatap.py (the graph), cli.py, data/catalog.yaml (the declared input for the sample).
• data/dcat-ap-3.0.1-4470b8e/: SEMIC's shapes, context, examples and changelog at the pin.
• samples/nhanes-participant/: the model's package as fetched and the catalog written from it, in Turtle and JSON-LD.
• build/snapshot_dcat_ap.py: re-creates data/ from a read-only clone of SEMICeu/DCAT-AP at the pinned commit.

Licences

Apache-2.0 (see LICENSE, NOTICE). The DCAT-AP artifacts in data/ are © European Union, CC BY 4.0, and keep that licence here.