Query

lokf query loads a bundle into an in-memory GraphStore (backed by pyoxigraph) and runs SPARQL over it. Nothing is written; the store lives only for the length of the command.

uv run lokf query examples/acme-knowledge \
  'SELECT ?s ?name WHERE { ?s schema:name ?name } ORDER BY ?name'

The preset prefixes

You never write a PREFIX block. The store prepends the LOKF schema namespaces to every query, so the common prefixes are already bound:

lokf, schema, prov, dcterms, skos, foaf, rdf, rdfs, owl, xsd, and dcat.

Write schema:name, prov:wasDerivedFrom, or lokf:Metric straight away. You can still declare your own PREFIX/BASE lines at the top of a query if you need one that is not preset.

With --schema domain.yaml, a domain schema’s own prefixes are preset as well, and the graph is projected under it as lokf convert does.

Worked queries over the reference bundle

SELECT

Every named thing in the bundle:

uv run lokf query examples/acme-knowledge \
  'SELECT ?s ?name WHERE { ?s schema:name ?name } ORDER BY ?name'

s                                                     name
https://acme.example/knowledge/glossary/active-user   Active User
https://acme.example/knowledge/services/analytics-api Analytics API
...

ASK

Does the bundle contain a Metric? (boolean questions need a machine-readable --format, e.g. json):

uv run lokf query examples/acme-knowledge \
  'ASK { ?s a lokf:Metric }' --format json

{"head":{},"boolean":true}

CONSTRUCT

Pull a subgraph back out as Turtle (CONSTRUCT/DESCRIBE always emit RDF):

uv run lokf query examples/acme-knowledge \
  'CONSTRUCT { ?s ?p ?o } WHERE { ?s a lokf:Metric ; ?p ?o }'

Output formats

--format / -f controls how results come back:

Value	Applies to	Result
table (default)	SELECT	aligned columns
json / csv / tsv	SELECT / ASK	SPARQL results
ttl	CONSTRUCT / DESCRIBE	Turtle (the default for those forms)

The just query recipe wraps the command:

just query examples/acme-knowledge 'SELECT ?s WHERE { ?s a lokf:Metric }'

From Python

lokf query is a wrapper around GraphStore. The store gives you the same queries, plus rdflib_graph(), serialize_results(), and dump():

from lokf.store import GraphStore

store = GraphStore.from_bundle("examples/acme-knowledge")
len(store)                                        # triple count

store.select("SELECT ?s WHERE { ?s a lokf:Metric }")
# -> [{'s': 'https://acme.example/knowledge/metrics/weekly-active-users'}]

store.ask("ASK { ?s a lokf:Metric }")             # -> True
store.construct("CONSTRUCT { ?s ?p ?o } WHERE { ?s a lokf:Metric ; ?p ?o }")

The prefixes are preset on the store too — store.prefixes is the dict it prepends. To publish the same store over HTTP, see Serve.