Operate

Once integrations are running, most of the work is seeing what happened and deciding what to do next.

Dashboard

Overview is computed from the message ledger: volume, delivery success, latency, and the operator work queue. Instance monitoring shows CPU, memory, disk, database, and runtime utilization. Retention windows live under Settings.

Messages and replay

Search the ledger. Replay a message when a downstream system was down — the original record stays put; Sirius writes a new, auditable attempt. The dead-letter queue supports retry, resolve, or a terminal close.

Logs and alerts

Structured logs search by level, component, integration, and related message. Alert rules fire on log patterns, failures, integration status, and cluster events, and deliver to email, Slack, Teams, or a webhook.

High availability

Run more than one node behind your load balancer. Failover and controlled handover are product features. Backups are checksummed; restores are verified. Size CPU, memory, and replicas from the resource guide in your deployment package and the /metrics Prometheus endpoint.

Step-by-step: Watch the runtime, Replay a message, Page on failure.

Agent

The in-product assistant uses the same APIs you do. Skills are playbooks with tool allow-lists. An MCP server exposes that surface to external agents under the same RBAC and audit rules. See Use the in-product agent.