HELPERG Data · Operations

API status

Live, domain-level status of the public data surface — economics, weather, news — with how each domain’s data arrives, which feed is late when one is, when each domain last received data, and whether the corpus is held in durable storage. Scheduled ingestion and on-demand providers are reported separately, because only one of them can be late. Read from the public status endpoint when the page loads.

Scope: the data domains behind /api/public/v1 and /api/v1. Website availability is reported separately on the platform status page. A domain that has never recorded data is unknown, which is not the same as healthy.

Current status

Checking the public status endpoint…

How data arrives

Scheduled ingestion: An ingestion job fetches the data on a timetable and writes it to HELPERG's own store. Lateness is measurable here, so a feed that stops is a fault, and each feed's own freshness is listed beneath the domain. Economics and news arrive this way.
On demand: The provider is fetched when a reader asks for a reading. There is no timetable, so nothing can be late; what can be reported is whether the last contact with the provider succeeded. Weather arrives this way.

What the words mean

Operational: For a scheduled domain: data is arriving within its freshness window and the last ingestion of every required feed succeeded. For an on-demand domain: the last contact with the provider succeeded. Never inferred from the absence of an error.
Degraded — data is stale: Data is arriving but later than its freshness window allows, or the latest run left part of the corpus behind. A partial run is never reported as operational.
Unavailable: A required feed has failed repeatedly, the store cannot be read, or the last contact with an on-demand provider failed. Responses for that domain fail closed rather than serve an empty result as fact.
Not yet checked: An on-demand provider that nobody has asked for on this deployment. Nothing is wrong and nothing has been verified, and the page says so rather than showing a green badge or a red one. The record of provider contact is kept in the durable store bound to this environment, so this is a statement about the deployment and not about whichever server answered your request.
Unknown: A scheduled domain with nothing recorded — no ingestion has run here, or its state could not be read. Reported as unknown, never promoted to healthy.

The overall word is the worst of the data domains, with one exclusion: an on-demand provider that has never been contacted is left out of it. Nothing was scheduled and nobody has asked, so that is an absence of events rather than an absence of health — and counting it made a deployment with a perfectly healthy corpus report itself as not operational. A contact that FAILED is still counted. A line under each domain says why it has the word it has, in the endpoint's own closed vocabulary. For a scheduled domain, "last data received" is the instant its newest successful ingestion completed; for an on-demand domain, it is when the provider was last reached successfully — an instant HELPERG records without keeping any note of what was asked about. An on-demand domain is never degraded by age alone: a week with no reader is not staleness. Each reading also carries its own freshness on the weather page. "Durable storage" says whether this deployment keeps its corpus in a durable store — yes or no, and nothing about which store. Operational detail behind these words is available only with a credential and is not published here.