Wikidata

Wikidata is the structured knowledge graph behind Wikipedia. wiki reads it two ways: entity lookups for a single item, and SPARQL for questions across the whole graph.

Entity lookups

Pass a Q-id (or P-id) to see an item's label, description, aliases, and claims:

wiki entity Q937              # Albert Einstein
wiki entity Q64 --lang de    # Berlin, labels in German

Restrict the claims to the properties you care about:

wiki entity Q937 --props P31,P569,P570   # instance-of, born, died

Do not know the id? Resolve it from a Wikipedia title with --title (wiki follows the article's wikibase_item):

wiki entity "Albert Einstein" --title --props P569,P570

The table view flattens each claim to a single value in the language you pick with --lang. The structured output keeps the whole entity: every language's labels, descriptions, and aliases, and every statement with its qualifiers, references, rank, and full typed datavalue, plus sitelinks with their badges and the entity's lastrevid and modified stamp. Nothing the API returns is dropped, so you can rebuild the original record from the JSON:

wiki entity Q937 -o json
wiki entity Q937 -o json | jq '.claims.P569[0].mainsnak.datavalue.value'

SPARQL

Run any query against the Wikidata Query Service. The query can be inline, read from a file with @path, or read from stdin with -. In the table view entity URIs are shortened to bare Q/P ids; -o json keeps each binding's full value along with its RDF term type, language tag, and datatype, so a literal is never confused with a URI:

wiki sparql 'SELECT ?city ?pop WHERE { ?city wdt:P31 wd:Q515; wdt:P1082 ?pop } ORDER BY DESC(?pop) LIMIT 10'

From a file, rendered as CSV:

wiki sparql @capitals.rq -o csv

From stdin:

echo 'SELECT ?p WHERE { wd:Q937 wdt:P800 ?p }' | wiki sparql -

Each SELECT variable becomes a column, so the result drops straight into a table, CSV, or JSONL just like every other command.