Skip to content

Integration safe degradation

Authoritative source: AetherCloud. This page is mirrored into the unified AetherIoT documentation.

Use this runbook when an edge-owned integration reports invalid data, topology gaps, stale observations, receipt uncertainty, or repeated delivery failure.

  • Experimental Home Assistant topology and observation flows keep discovery and acquisition at AetherEdge.
  • Cloud projection queries are bounded, Tenant scoped, and labelled as edge-reported copies rather than live-state authority.
  • Topology generation and replay rules reject conflicting or out-of-generation observations and support a complete resynchronization.
  • Integration control is default off, fixed in scope, explicitly confirmed, and does not treat delivery as physical completion.
  • Webhook delivery uses bounded retries and dead-letter evidence without rolling back the committed business result.

Published alpha.4 consumption, production CloudLink and credential composition, supported-broker evidence, production control composition, signing-key lifecycle, a public agent service, and a governed control re-enable workflow remain gated or planned.

  1. Keep AetherEdge local collection, deterministic automation, and safety behavior independent of cloud availability.
  2. Mark the cloud projection stale or unavailable. Preserve its observation time and never guess missing device state.
  3. Remove command exposure for the affected integration while keeping bounded read-only discovery only when its identity and Tenant scope remain trusted.
  4. Request a complete edge-owned topology and observation resynchronization using the existing session and Runtime Manifest boundaries.
  5. Restore read access only after generation, digest, ordering, and freshness converge. Restore control only through explicit operator enablement and policy review.
  6. Keep Home Assistant addresses, tokens, credentials, and provider payloads at the edge.

Escalate repeated topology conflicts, non-converging sequence gaps, credential disclosure, unknown control outcomes, or physical safety risk. Dead-letter redrive and control re-enablement require explicit human review; an agent must not choose them merely to make a warning disappear.

Read Home Assistant integration, governed control, and audit and integrations.