Digital Public Library of America (DPLA)

The Digital Public Library of America gathers records from thousands of American museums, libraries, and archives into a single searchable collection that runs to millions of items. It is organised on two levels: items, meaning one digitised object contributed by a partner, and collections, meaning meaningful groupings of those items. Contributors include Harvard, the New York Public Library, ARTstor, and the David Rumsey Historical Map collection, so the coverage across American cultural heritage is genuinely wide.

There are two sensible ways in. The API returns item and collection records as JSON and is the right tool for targeted queries, though you register for a free API key first. For anything at scale, the entire DPLA repository is offered as zipped JSON files you can download, which makes it easy to process the whole catalogue offline and index it in one pass rather than paging through the API.

The point to grasp before you build is that DPLA hands you rich descriptive metadata (titles, creators, dates, subjects, rights statements, and links back to the holding institution) rather than the digitised objects themselves. For a RAG system that is often exactly what you want: clean, structured provenance sitting beside your other text, ready for grounding answers about American historical objects, powering a subject-searchable index of holdings, or enriching your own records with authoritative dates and attributions.

On rights, the metadata is released under CC0, so you can reuse it freely, including commercially, with no attribution needed. The digitised items it links to are a separate question: each stays under whatever terms its contributing institution set, which is why we flag that the content licence differs. In short, reuse the descriptions freely but not necessarily the underlying images or full texts. Bear in mind too that metadata depth and consistency vary by contributor, since the records come from many different cataloguing traditions.

For European material, Europeana is the direct counterpart and works much the same way, and Wikimedia Commons is worth pairing in when you need openly licensed media files rather than metadata alone.