Observability and console

Product guide: what the capability brings, for whom, its limits and the version that delivered it - no configuration syntax.

What it does	The gateway emits logs without personal data, metrics per operation, upstream and tenant (bounded cardinality), and W3C traces. The console reads them: overview, traffic, an operation's detail, upstreams and their health, health per replica, recent errors, the gateway's traces, consumption per tenant, agents, the API catalogue, Kafka clients
Why it matters	The console is enough without Grafana or a metrics backend: it becomes the gateway's collector and keeps 35 days of detail and 13 months of aggregates. With your backends, it reads them and stores nothing twice. A provoked upstream outage shows in it within a minute - chart, ranking, alert and a trace of the stage at fault
Alerts	Built-in rules (error rate, latency, circuit breaker, silent replica, new error), notified by signed webhook; a daily or weekly summary; a rule created from a chart
Catalogue	The APIs derived from the contracts, their owner, their lifecycle; a Compliance tab lists operations without an owner or deprecated without a withdrawal date
Access	Local accounts without an identity provider (first admin, admins' MFA), or OIDC alongside; roles and tenant per account; a reader of one tenant gets 403 on another, never a misleading empty list
Mobility	The whole console works on a phone and a tablet, for on-call
Limits	A console reads one gateway: the multi-gateway console is scoped, without a milestone; the developer portal is a separate product, planned for 1.11.0
Since	0.1.0 (read-only console), 1.0.0 (local accounts), 1.2.0 (redesigned screens, agents, compliance), 1.3.0 (autonomous observability, alerts), 1.4.0 (health per replica, phone and tablet)
Configure	Observing, Reading with the console