Skip to content

Status

Your browser asks this origin two questions, and the answers appear below. GET /health says whether the API is answering and which corpus it loaded. GET /v1/meta says how far the corpus runs.

This readout is self-reported. Both answers come from the process that served this page, so it can report a working API only while the API is working. An outage shows up here as a failed request, or as this page not loading. Nothing outside this origin watches it yet.

Now

Your browser fetches these when the page loads.

The readout runs on JavaScript. Without it, both answers are one request away: /health and /v1/meta return JSON to curl as readily as to a browser.

What the readout covers

corpus_version is sha256: plus the hash of MANIFEST.json, the inventory of published partitions — one file per source per day. Two responses carrying the same value were answered from the same corpus, and a backfill that changes a historical cut changes this string.

The last partition date is where the rows end, per source. One source can be certified up to a later day than another, so the per-source rows below matter more than the corpus-wide span. sec.edgar and us.federal_register.pi carry the bulk of the corpus. missing_days counts days inside a source's span that we have not mined. Their row count is null, and a cut that needs one of them answers 409 with the day's key in coverage.missing.

The API state covers one request from your browser to one Cloud Run instance, of up to ten. It says that instance answered, and it says which corpus that instance had loaded. Ingest, Stripe, and the flat-file downloads are not probed here. The changelog records what shipped and when.