one row, one instant, no key
GET /v1/sample/news
$ curl -s "$PIT/v1/sample/news?example=svb&as_of=2023-03-10T23:59:59Z"
{
"as_of": "2023-03-10T23:59:59Z",
"visible_by": "published_at",
"results": [{
"id": "0001193125-23-067777_719739",
"title": "SVB FINANCIAL GROUP 8-K",
"published_at": "2023-03-10T23:59:59Z",
"acceptance_at": "2023-03-10T22:23:03Z",
"content_sha256": "8a95862ef3953dd206f1afeb5a0ab19014b57f7f332b764e948f6514fdd2d06a",
"source_locator": "https://www.sec.gov/Archives/edgar/data/719739/000119312523067777/0001193125-23-067777.txt"
}],
"coverage": { "touched": ["sec.edgar/2023-03-10", "identity.ticker/SIVB@2023-03-10"], "missing": [] }
}
01
A hash per row
content_sha256 covers the bytes we parsed and
source_locator points at the filing on sec.gov. An
auditor can fetch the publisher’s copy and compare it
against what your run was handed, without asking us for anything.
02
A certificate per day
coverage.touched lists every day the cut read, and a
complete day names the set of SEC form types it was measured
against. A cut that touches a day we never certified comes back
409 with that day named, so a hole never passes for a zero.
03
A version per answer
Each response names the corpus_version it was
computed from. Pin it next to a published result: if a later
answer differs, that version says the corpus moved underneath
it.