For AI agents: a documentation index is available at https://www.plungeai.com/docs/llms.txt. Append .md to any page URL, or send Accept: text/markdown, to get markdown. Setup instructions for agents are at https://www.plungeai.com/docs/agents.md. One ozk_ key opens every plane, models included.

Documentation Index: fetch the complete documentation index at /llms.txt. Use this file to discover all available pages before exploring further.

Status page

The inference gateway publishes a public status page at https://gateway.plungeai.com/status (opens in a new tab). It needs no API key. It shows one row per upstream provider the gateway can route to:

ColumnWhat it is
Stateoperational, degraded (under 95% of the last 24 hours' requests to that provider succeeded) or outage (the gateway has the provider marked out after repeated upstream failures and is routing around it)
Success (24 h)The share of the last 24 hours' requests to that provider that did not end in a 5xx error
p50 TTFT ms (1 h)The median time to first token over the last hour's requests that recorded one, in milliseconds

A provider with no traffic in the window shows a dash. The same data is available as JSON at https://gateway.plungeai.com/status.json:

{
  "generated_at": "2026-10-14T18:00:00.000Z",
  "providers": [
    { "provider": "openai", "state": "operational", "success_24h": 0.9998, "p50_ttft_ms_1h": 412 }
  ]
}

Both are aggregates across all customers. They never carry an organisation, key or request. The snapshot is recomputed at most once a minute.

Readiness and liveness

EndpointAnswers
GET https://gateway.plungeai.com/healthReadiness. 200 when the gateway can serve, 503 when it cannot
GET https://gateway.plungeai.com/healthzLiveness. Always 200 with a fixed body while the Worker runs; it touches nothing

/health runs three checks at once and reports each one:

CheckWhat it testsOn failure
billing_dbA query against the billing database that holds balances and request rowsstatus: "down", HTTP 503
kvA read from the key-value store that holds rate-limit counters and routing statestatus: "down", HTTP 503
ratecardThe age, in seconds, of the published rate card the gateway prices requests fromOlder than 26 hours, or no publish time: status: "degraded", HTTP 200

The body also lists providers_out: the providers currently marked out after upstream failures. A marked provider does not make the gateway unready; it shows outage on the status page.

{
  "status": "ok",
  "service": "inference-gateway",
  "env": "production",
  "checks": {
    "billing_db": { "ok": true, "ms": 4 },
    "kv": { "ok": true, "ms": 2 },
    "ratecard": { "ok": true, "age_s": 3600 }
  },
  "providers_out": []
}

Other health endpoints

Reporting a problem

If the status page shows every provider operational and your requests still fail, contact PlungeAI support with the x-request-id response header of a failing request.

Planned: TI-34

The docs assistant is not available yet. You can hand these docs to your own assistant instead.