Service status

Release health

What open tabs report through release-watch: tabs heard from in the last 12 minutes on the release each service serves and on older ones, and since that release started, how many were prompted, updated in place or reloaded, deferred (and why) or failed. Drain is how long after a release went live the last older tab left; tabs report every 5 minutes, so it reads true only 6 minutes after a start. Watched since 2026-11-13T08:01:11.253Z.

Developer path

Once a day a test developer account goes from sign-in to a sandbox project and app, a Media upload, an Events publish and pull, an AI answer, the project export, a Services release, a credential rotation and revocation, and cleanup, through the public API and the SDK. Last run 2026-11-13T05:22:49.404Z: passed (13/13 steps).

• account: ok
• discovery: ok
• project: ok
• app: ok
• grants: ok
• media: ok
• events: ok
• ai: ok
• play: ok
• export: ok
• release: ok
• credentials: ok
• cleanup: ok

Tools job proof

Every six hours a service client submits a converter job to OpenVibe.Tools through the SDK, drops and reattaches its progress stream, downloads the result from OpenVibe.Media and finds the job's events in OpenVibe.Events. Last run 2026-11-13T06:25:24.905Z: passed (6/6 steps).

• token: ok
• cursor: ok
• submit: ok
• reattach: ok
• result: ok
• events: ok

SLO categories (proposals)

Every target below is a PROPOSAL for discussion. None is agreed, none is enforced, and no alerting exists for any of them. 'measured_by' names what can be measured today; 'not instrumented' means nothing measures it yet. Correctness-sensitive systems (billing, authentication) prioritise integrity over availability; live transport prioritises continuity, with eventual control-plane convergence. JSON: /api/v1/status/slo.

Identity login and token exchange

Share of POST /oauth/token and sign-in requests that answer without a 5xx; p95 latency of token issuance.
Proposed: availability 99.9% of requests over 30 days; latency p95 seconds 0.5 · priority: integrity · measured

Public page availability

Share of GET requests to server-rendered public pages that answer without a 5xx.
Proposed: availability 99.5% of requests over 30 days · priority: availability · measured

Control APIs

Share of authenticated write/control API requests (stream settings, object writes, subscriptions, jobs) that answer without a 5xx; p95 latency.
Proposed: availability 99.5% over 30 days; latency p95 seconds 1 · priority: availability · measured

Chat message persistence and delivery

Share of accepted chat messages that are persisted and delivered to connected viewers; delivery delay.
Proposed: persisted 99.99% of accepted messages; delivery p95 seconds 1 · priority: integrity · not instrumented: Live exposes live_ws_connections only; no per-message persistence or delivery counter exists yet

Ingest acceptance

Share of publish attempts (RTMP, WHIP, WebRTC, SRT) with a valid key that are accepted.
Proposed: accepted 99.5% of valid publish attempts over 30 days · priority: continuity · not instrumented: no publish-attempt counter exists yet

Active stream continuity

Share of live streams that end by the broadcaster's action rather than by a server restart, deploy or crash.
Proposed: unplanned interruptions fewer than 1 per 1000 stream-hours · priority: continuity · partly instrumented: the number of live streams is measured; the cause of each stream end is not recorded

Media upload and processing

Share of uploads and recordings that reach ready (never silently lost); time from upload or stream end to ready.
Proposed: reach ready or explicit failure 100%; processing p95 seconds 600 · priority: integrity · partly instrumented: states are counted; time-to-ready is not measured

Ledger settlement correctness

Share of settlements whose journal balances and reconciles with the provider; duplicate monetary effects.
Proposed: unbalanced or duplicate settlements 0 · priority: integrity · not instrumented by this change (Billing is owned by another track)

Event delivery latency

Time from an event's acceptance to its successful delivery to each subscriber; dead-letter depth.
Proposed: delivery p95 seconds 30; dead letters unreviewed after 24h 0 · priority: integrity · measured

Developer API availability

Share of public developer API requests (registry, JWKS, events pull, object API) that answer without a 5xx.
Proposed: availability 99.9% of requests over 30 days · priority: availability · measured