sdc_dcat3_us

One Semantic Data Charter model, described in DCAT-US 3.0.

sdcdcatus reads a published SDC model's package and writes the data.json Catalog that DCAT-US 3.0, the United States federal data catalog standard, asks an agency to publish: one Dataset per model, with the model's schema cited as the data dictionary (describedBy) and as the standard the records conform to (conformsTo), by URL and SHA-256. The output is checked with GSA's own JSON Schema and validation script at a pinned commit of GSA/dcat-us, and with Data.gov's online validator.

1. What this is, and where it came from

The sample, samples/nhanes-participant/data.json, 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]"
sdcdcatus write --package samples/nhanes-participant --out data.json
sdcdcatus write --ct-id xy8upneajsb8vdcmnve01g6g --ct-id <another> --save-package pkg/ --out data.json   # a catalog of models, from the public catalog
sdcdcatus write --package DIR --catalog my-catalog.yaml --contact-email [email protected] --out data.json  # a publisher's own declared input

A catalog of seven. samples/fair-data-demo/data.json is the same writer run over the seven FAIR Data Demo models fetched from the public catalog (NHANES Participant and Medication, BRFSS Respondent, CMS Beneficiary, Inpatient Claim, Outpatient Claim and Prescription Drug Event), each package saved beside it: one data.json, one Dataset per governed model, each with its schema as the data dictionary. That is what a comprehensive data inventory entry per model looks like.

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, publisher, dates, language, licence, the schema's URL and SHA-256, the variable count. From the catalog's publisher, declared in catalog.yaml because the schema or the statute asks for it and the package cannot know it: the contact point, the catalog's own title, description and homepage, and what the publisher asserts about access. The package ships a default file for the sample (src/sdcdcatus/data/catalog.yaml); an agency supplies its own.

Why describedBy. OMB Memorandum M-25-05 requires every agency's comprehensive data inventory to carry "the names and definitions of all variables in the data asset," and GSA's crosswalk maps that requirement to describedBy: a Distribution pointing at a machine-readable data dictionary. A published SDC schema is that: every variable with its label, definition, datatype, units, permitted values with their codes, and the sixteen typed reasons a value may be missing, immutable once published and cited here by URL and SHA-256.

What the Dataset carries:

title, description	the model's, plus one sentence saying what a governed data record is and that the schema is the data dictionary
identifier	an Identifier: notation dm-xy8upneajsb8vdcmnve01g6g, schemaAgency the publisher
contactPoint	a Kind with fn and hasEmail, from the declared input
publisher, creator, landingPage, keyword, license, language en	from the package; the publisher is the model's dc:publisher when the modeler wrote one, else the declared one
subject, contributor, spatial, relation, rights	from the model's Dublin Core when the modeler wrote them: subject keywords, contributors, coverage as a named Location, a relation URL, the rights statement beside the licence URL
describedBy	a Distribution: the schema's pinned URL as downloadURL, application/xml, XSD, a SHA-256 Checksum, conformsTo the SDC4 reference model
conformsTo	a Standard for the same pinned URL
issued, modified	the model's publication date; inventoried the day the catalog was written
accessRights	public, as the publisher declares for the model
versionNotes	a published model never changes; a revision is a new model naming the one it revised; the SHA-256 and the permanence page

2. How to verify it

The tests run GSA's own validator from data/dcat-us-4996591/, where their 26 class definitions, their example corpus and their test_json_schema.py are copied at the pinned commit (build/snapshot_dcat_us.py verifies the pin first):

pip install -e ".[dev]"
python -m pytest tests -q

The suite first runs their script on the snapshot (their good and bad corpus: 510 pass, 0 fail at the pin), then validates the sample Catalog, its Dataset and the describedBy Distribution with their method, a referencing registry of every definition by $id and Draft202012Validator with the format checker: 0 errors.

Second witness, Data.gov's online validator at https://harvest.data.gov/validate/ (schema "dcatus3.0 catalog", paste or upload the sample): "No validation errors found" for the one-model catalog and for the seven-model catalog, 9 October 2026. The seven-model catalog is also checked in the tests, and the committed file is asserted equal to a fresh run over the saved packages.

3. What the projection could not say

Left out rather than filled in:

• No distribution of the records. The FAIR Data Demo's records are not published; the model is. The schema allows a Dataset without distributions, and the writer emits them when a publisher declares an access or download URL.
• Nothing an agency would have to determine. The M-25-05 open-data, FOIA, Federal Data Catalog, open-format and licensing determinations, the Title 44 statements, the access, use and CUI restrictions, bureau and program codes: all of these are an agency's assertions. The writer takes them only as declared input and defaults none of them. The sample, published by a company and not an agency, carries accessRights: public and no determinations.
• No temporal coverage, no theme, no quality measurements. The package does not carry them. Spatial coverage, subject, contributors, relation and a rights statement come from the model's own 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: its modeler left them at the defaults.

Declared rather than read from the package, and said so here: the contact point ([email protected] on the sample), the catalog's title, description and homepage, the publisher's name and identifier, accessRights.

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-US's provenance, wasGeneratedBy and qualifiedAttribution describe the dataset; the per-record facts are in the records, defined in the data dictionary.
• The schema is bound to each record. Every record names the schema it was validated against; the catalog can say that once, which conformsTo with the SHA-256 does.
• 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 data dictionary defines them; which one applies, and where, is in the record.

4. What we learned about DCAT-US 3.0

Implementer's notes from building this against GSA/dcat-us at 4996591. They describe how the artifacts behave, so the next implementer spends the day on their own catalog rather than on these.

• The validated form is JSON against JSON Schema 2020-12. The earlier draft's JSON-LD context and SHACL shapes (from the Interior Department's repository) sit under DEPRECATED/; a data.json is what the script and the online validator check. An implementer coming from DCAT-AP, which is SHACL over RDF, should not look for shapes.
• Format assertions need the extras. The definitions use format: date, date-time and iri. With plain jsonschema those are not checked, and six of GSA's own "bad" examples pass; with jsonschema[format], which their CI installs, all 510 corpus cases behave as labelled. The tests here require the extras.
• Unknown properties pass. additionalProperties is not set on the classes, so a misspelled or legacy property is accepted without a message. The validator tells you what is missing or malformed, not what it did not recognize.
• Null means "not provided." Every recommended property accepts null explicitly ("Null allowed when not required"); omitting the property is equally valid.
• language is two letters (ISO 639-1), a documented breaking change from v1.1's BCP 47; en-US fails.
• Dates take four forms: date-time, date, YYYY, YYYY-MM.
• identifier is a string or an Identifier object (notation, schemaAgency, version); contactPoint is one Kind or a list; hasEmail must match a mailto: pattern; license is a string (the crosswalk says a URL).
• describedBy is a Distribution, not a Document, so a data dictionary gets a downloadURL, a mediaType, a checksum and its own conformsTo. That is the right shape for a schema.
• conformsTo is a single Standard on a Catalog and a list of them on a Dataset. Both accept identifier as a string or an Identifier object.
• A Distribution has no mandatory property, and a Dataset has four: title, description, contactPoint, identifier. Everything else that the statute asks for is Recommended.
• The registry method. Each definition's $id is https://resources.data.gov/dcat-us/3.0.0/definitions/<class> and references are the path /dcat-us/3.0.0/definitions/<class>; a referencing registry built from the directory resolves them, as test_json_schema.py does. The online validator agreed with the script on the sample.

MCP server

sdcdcatus is also an MCP server, registered as io.github.SemanticDataCharter/sdcdcatus, so an assistant that can read a published model with sdcreader can ask for its DCAT-US 3.0 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 sdcdcatus mcp            # or: sdcdcatus-mcp, or sdcdcatus 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/sdcdcatus/: dcatus.py (the Catalog), cli.py, data/catalog.yaml (the declared input for the sample).
• data/dcat-us-4996591/: GSA's definitions, examples, validator and README at the pin.
• samples/nhanes-participant/: the model's package as fetched and the data.json written from it.
• build/snapshot_dcat_us.py: re-creates data/ from a read-only clone of GSA/dcat-us at the pinned commit.

Licences

Apache-2.0 (see LICENSE, NOTICE). The DCAT-US artifacts in data/ are a work of the United States Government (17 USC § 105) released under CC0 1.0 and keep that status here.