Wikipedia
Wikimedia publishes complete database dumps of Wikipedia and its sister projects (Wiktionary, Wikiquote, Wikisource, and the rest) at dumps.wikimedia.org, refreshed roughly twice a month. You can take a single language edition or every one of the nearly 300 languages, and choose raw wikitext, parsed HTML, or the SQL tables behind the site. The compressed English article text is around 22 GB.
The quickest route for RAG is to skip the dump parsing entirely and pull a pre-processed, one-article-per-record version from the Hugging Face Hub, where the wikitext markup has already been stripped back to clean prose. If you do work from the raw XML, budget time for a parser: wikitext is fiddly, with templates, infoboxes, and citation markup that all need handling before you get usable text.
Wikipedia is the natural first corpus for almost any general-knowledge RAG system. It is broad, consistently structured, cleanly licensed, and its article sections chunk neatly into passages. It is an excellent way to stand up and test a retrieval pipeline before you move on to messier, domain-specific data.
Two things to watch. First, currency: a dump is a snapshot, so anything edited after it was taken will be missing, which matters for fast-moving topics. Second, the licence. The text is Creative Commons Attribution-ShareAlike (CC BY-SA 4.0), which means commercial use is fine but you must credit Wikipedia and share any derived text under the same terms. Some embedded media and older content carry different or additional licences, so check before redistributing.
For structured facts rather than prose, pair Wikipedia with Wikidata, its sister project, which stores the same knowledge as machine-readable statements and stable identifiers you can use to ground and disambiguate the entities your documents mention.
Related sources
DBpedia
A knowledge graph built by pulling the structured parts of Wikipedia, mainly the infoboxes, into machine-readable data. It holds billions of facts about people, places, organisations, and more as RDF triples, small subject, predicate, object statements, which you can search with the SPARQL query language.
Freebase
A collaborative knowledge base once run by Google, now retired but still available as downloadable data dumps. It holds structured facts about millions of entities, and much of its content has since moved into Wikidata.
OpenCyc
The open release of Cyc, one of the oldest attempts to hand-build common-sense knowledge for machines. It holds hundreds of thousands of concepts and millions of assertions about how the everyday world fits together. Now archived, but the data is still available.
Structured Wikipedia
Wikipedia rendered as pre-parsed, machine-readable JSON: abstracts, short descriptions, infoboxes, sections, parsed tables and references, with links to Wikidata entities. A beta from Wikimedia Enterprise covering nine languages, also mirrored on Hugging Face. The section-segmented shape a RAG pipeline actually wants.