Operations

The operations blocks are what you run kis.ai with, rather than what you build on. They cover getting telemetry out of a deployment, storing the artifacts it runs, and keeping processes alive on the nodes that host them.

The blocks

Block	What it does
Observability	Collects logs, traces and metrics per tenant, and serves queries over them
Registries	Container and artifact storage for what you deploy
Supervisor	Keeps a declared set of services in their desired state on a node

How they relate

   service node
        │
        ├── Supervisor      keeps declared services running, restarts on
        │                   failure, restores the stack after reboot
        │
        ├── emits ──▶ Observability collector ──▶ storage ──▶ querier
        │
        └── pulls images from ──▶ Registries

Supervisor is the lowest layer — it is what makes a node’s services exist. Observability tells you what they are doing once they do. Registries hold what they run.

Tenancy

Telemetry is collected and stored per tenant, the same as everything else on the platform. That matters more in operations than elsewhere: an observability stack that mixes tenants is a data boundary violation, not just untidy.

See also

Two things that look like operations blocks and are not:

• Usage measures business time — human versus AI versus idle in a process. Observability measures system behaviour. Different questions, different blocks.
• Audit is a tamper-evident record for compliance, not a log stream for debugging. Do not reach for observability when you need audit, or the reverse.