Provenance
Where the recaps come from.
Recap.at reconstructs events from public sources - encyclopedias, archived web, news indexes, structured knowledge bases, and the cultural record. Every claim on a recap page traces back to one of those, listed in that recap’s “Sources” section and aggregated here.
We hot-link to providers rather than re-host their work. Short excerpts are quoted under fair use; long-form content stays at the source. Where a fact has only one provider, we say so quietly. Where a claim has weight, we want it standing on more than one.
1,471 citations across 14 providers, drawn from 1,903 published recaps covering -130000–74000.
Data feeds
Where the recaps are assembled from.
The upstream APIs the build pipeline queries to compose each recap. The citation list below is what each finished recap names; this is the broader feed surface the platform pulls from.
43 feeds across 8 categories
Encyclopedic
DPLA
Digital Public Library of America - 47M items across US archives, libraries, museums.
HathiTrust
18M digitized historical texts from research libraries.
Inventaire
Wikidata-shaped book + author graph; SPARQL queries.
Open Library
Books, authors, ISBNs, and reading lists from the Internet Archive's bibliographic catalog.
Wikipedia REST + Action API
Article summaries, page content, and revision history.
Wikisource
Public-domain source documents: speeches, treaties, laws, transcripts.
Structured knowledge
Anthropic Claude API
LLM for narrative recap synthesis and entity extraction.
data.gov (US)
US federal open-data portal across 50+ agencies.
data.gov.uk
UK Government open-data portal - 50k+ datasets.
GeoNames
11M place-names with coordinates, admin hierarchy, and historic alt names.
OSM Nominatim
Reverse geocoding + place lookup against OpenStreetMap.
Our World in Data
Curated long-run datasets - pandemic, war, poverty, energy, with provenance.
OurAirports
Aviation reference - 80k airports, history of routes + aircraft incidents.
UN Data
UN statistical databases - refugees, treaties, peacekeeping, votes.
Wikidata SPARQL
Structured knowledge graph queries via SPARQL endpoint.
News + journalism
Apify (Actors)
Pay-per-compute scraping platform. Runs news/media-coverage Actors that seed heal-media-coverage with real contemporaneous press headlines. Dormant scaffold.
Chronicling America
Library of Congress historic US newspapers, 1777–1963.
GDELT 2.0 DOC API
Global news events and article metadata firehose.
Social + public posts
Bluesky (AT Proto)
Federated social network; public read endpoints + AT-Proto firehose.
Hacker News (Algolia)
Searchable HN comments + stories via Algolia, 2006–present.
Mastodon
Federated microblogging; per-instance public timelines.
Mastodon - Instance
Mastodon instance host (e.g. historians.social). Optional; defaults to historians.social when unset.
Reddit OAuth API
Subreddit posts, comments, and historical discussion threads.
Media
Archive.org TV News
Internet Archive's TV news archive: searchable transcripts of US broadcast news 2009-present.
Library of Congress
Digital collections, photos, and historical archives.
Library of Congress Photos
LoC's /photos/ index - image-only filter on top of the LoC search API. Public-domain anchor for the hero-finder cascade.
NASA Open APIs
NASA Image Library - searchable space/astronomy imagery for the heal-gallery cascade.
Pexels
Free stock photography library. Hero-finder fallback source.
Pixabay
Free stock photography + illustrations. Hero-finder fallback source.
Unsplash
Free stock photography with photographer attribution. Hero-finder fallback source.
Wikimedia Commons
Free media files, images, and structured Commons data.
By provider
Sorted by citation count.
Showing 1–14 of 14 providers
Wikipedia
1433Cited in 1429 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenOpen Library
8Cited in 5 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenWikidata
5Cited in 4 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenInternet Archive
4Cited in 4 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenWayback Machine
4Cited in 3 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenGDELT
3Cited in 3 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenHacker News
3Cited in 3 recaps
Most prominently in The September 11 Attacks (2001)
OpenLibrary of Congress
3Cited in 3 recaps
Most prominently in The Fall of the Berlin Wall (1989)
OpenWikimedia Commons
3Cited in 3 recaps
Most prominently in The Fall of the Berlin Wall (1989)
Open9/11 Commission
1Cited in 1 recap
Most prominently in The September 11 Attacks (2001)
OpenElectoral Commission
1Cited in 1 recap
Most prominently in The Brexit Referendum (2016)
OpenHouse of Commons Library
1Cited in 1 recap
Most prominently in The Brexit Referendum (2016)
OpenNASA
1Cited in 1 recap
Most prominently in Apollo 11 (1969)
OpenWikisource
1Cited in 1 recap
Most prominently in The Fall of the Berlin Wall (1989)
Open
Methodology
How sources land here.
- Cache before fetch. Every external API call goes through a TTL+upstream-signal cache (see
apps/web/src/lib/cache/types.ts). We don’t re-hit Wikipedia for a fact already known. - Hashed dedupe. Identical responses are detected by SHA-256 of the canonicalized JSON, so “refresh” doesn’t re-write a row that hasn’t changed.
- Storage policy. IDs and metadata stored. Short excerpts (<200 chars) only. Hashed handles for social posts unless verified or official.
- Provider attribution. Each source links out. Rate limits respected. User-Agent identifies as Recap.at with a contact mailbox.