Wikibase Entity Extractor

Point at any Wikibase instance and extract its items and properties through the modern Wikibase REST API v1 — the typed successor to the MediaWiki Action wbgetentities API. One runner for every conformant install: Wikidata, MaRDI, Wikimedia Commons, wikibase.world, and the growing population of self-hosted institutional / GLAM / research knowledge graphs.

Every Wikibase answers the identical REST v1 grammar and returns the identical typed envelope, so a single actor covers them all — no per-site scraper.

What it does

Give it one or more instance URLs (e.g. https://www.wikidata.org) and a list of entity IDs (Q… items, P… properties), pick a mode, and it emits clean flat rows.

Mode	One row per…	Columns
entity (default)	entity	id, type, data type, labels, descriptions, aliases, a statements summary (count + property IDs used), sitelink count
statements	statement / claim	property, property data type, rank, value type + content, qualifiers, references count
labels	entity	just the multilingual labels + descriptions (cheapest way to translate / label thousands of IDs)

Optional languages filters labels / descriptions / aliases to a subset. Each entityIds value may be a bare Q42 / P31 or a full URL — a REST URL (…/entities/items/Q42) or a human entity page (…/wiki/Q42) — from which the instance and ID are parsed automatically.

Every row keeps a lossless _raw copy of the untouched API object plus _instanceUrl provenance.

Endpoints used

{instance}/w/rest.php/wikibase/v1/entities/items/{Qid}
{instance}/w/rest.php/wikibase/v1/entities/properties/{Pid}
                                              …/statements
                                              …/labels
                                              …/descriptions

Example input

{
  "instanceUrls": ["https://www.wikidata.org", "https://portal.mardi4nfdi.de"],
  "entityIds": ["Q42", "P31", "Q1"],
  "mode": "entity",
  "languages": ["en", "de"],
  "maxEntities": 1000
}

Honesty & edge cases

• A host that is bot-gated, returns an HTML shell, or is not a Wikibase REST endpoint is skipped with a warning — its entities are simply not emitted. (Note: the hosted *.wikibase.cloud fleet often bot-gates the REST path; the addressable population is self-hosted Wikibase + MediaWiki installs.)
• If every requested instance is unreachable, the run fails fast (non-zero exit) so an empty dataset can never masquerade as success.
• A reachable instance on which a requested entity does not exist emits 0 rows for that entity and exits cleanly — never an invented row.

Authentication

None required for public read. An optional bearer token and/or extra headers are supported for a private deployment and are never logged.

Pricing

Pay-per-event: one entity charge per Wikibase entity successfully returned.

Budget-safe: if you set a maximum cost per run, the actor checks before each entity's request that your budget still covers it, and stops cleanly when it does not. You are never sent (or billed for) entities beyond your budget, and entities that are not found are free. In statements mode one entity is one charge, however many statement rows it has.