Status

GET /v1/status

Returns the current state of the API and of each product, the state of each for every UTC day of the last 90 days, and the uptime of each over the covered part of those days. The route requires no key.

Request

curl https://api.albastov.com/v1/status

import json
import urllib.request

req = urllib.request.Request(
    "https://api.albastov.com/v1/status",
    headers={"User-Agent": "my-app/1.0"},
)
status = json.load(urllib.request.urlopen(req))
print(status["status"])
for c in status["components"]:
    print(c["name"], c["status"], c["uptime_90d"])

const res = await fetch("https://api.albastov.com/v1/status");
if (!res.ok) throw new Error(`${res.status} ${await res.text()}`);
const status = await res.json();
console.log(status.status);
for (const c of status.components) console.log(c.name, c.status, c.uptime_90d);

Response

200, Content-Type: application/json, Cache-Control: public, max-age=30. The server computes the response at most once every 30 seconds; updated_at is the time of that computation. The example shows one component and two of its 90 days.

{
  "status": "operational",
  "updated_at": "2026-11-03T12:00:00Z",
  "components": [
    {
      "id": "api",
      "name": "API",
      "status": "operational",
      "uptime_90d": 99.98,
      "days": [
        {"date": "2026-08-06", "status": "unknown"},
        {"date": "2026-11-03", "status": "operational"}
      ]
    }
  ]
}

Status rules

Each component has health checks and, except api and events, a failure share: the share of its account key (abk_) requests answered with a 5xx status, and for text-to-speech and speech-to-text also the share of its finished jobs that failed on the service side or expired in the queue. A request or job refused for the caller's input counts as an attempt, not a failure; a cancelled job counts as neither.

Two health checks feed the components:

• API check: a request to the API's health route; the check fails while the API does not answer.
• Model service check: a request to the service that runs the models; the check fails while that service does not answer.

A day takes the worst state of its health checks and failure shares. uptime_90d is 100 × (1 − time with a failed health check ÷ time covered). The time covered runs from the first of the 90 days, or from the start of the status history when later, to updated_at; unknown days are outside it.

Only health checks count toward uptime_90d. A day that is degraded or down from its failure share alone does not lower uptime_90d, so a component can show 100 beside a degraded day.

Errors