Omni
Omni is the intelligence and orchestration layer of QTM OS. It observes the system, interprets what it sees, and produces recommendations. It does not control anything.
What Omni is
Omni sits alongside the execution system. It reads from the same data — tasks, jobs, service events, records, signals — and uses that information to help human administrators and operators make better decisions.
Think of Omni as an analyst that has read everything in the system. It can summarize what’s happening, flag problems, and suggest next steps. But it cannot approve, execute, or mutate anything on its own.
What Omni sees
Omni can observe:
- Tasks — what is open, stuck, or aging
- Jobs — what is queued, active, or incomplete
- Service events — what is in progress or blocked
- Records — what has been completed
- Trust signals — how many signals an operator has, when the latest was emitted
- Access states — whether operators are active, limited, or suspended
- KPI data — system-wide health metrics
When a request is routed through the system, Omni can attach a trust advisory — a summary of the operator’s trust signals — to the routing context. This advisory is read-only and has no effect on routing decisions in v0.01.
What Omni does
- Interpretation — turning raw system state into understandable summaries
- Recommendations — suggesting next actions based on what it sees
- Anomaly detection — flagging queues that are stuck, work that is aging, exceptions
- Agent coordination — running bounded sub-agents (like a CEO agent or COO agent) and synthesizing their outputs into a single recommendation
Omni produces a Decision Packet — a structured recommendation that includes what it found, what it suggests, and what it cannot resolve. Human admins review this and decide what to do.
What Omni does NOT do
| Action | Omni’s status |
|---|---|
| Approve a task | Cannot |
| Create a job | Cannot |
| Reassign an operator | Cannot without approval |
| Complete a service event | Cannot |
| Change access state | Cannot |
| Settle a payment | Cannot |
| Bypass policy | Cannot |
| Mutate records | Cannot |
All mutations require human approval and go through standard system paths.
Where Omni appears
Omni operates within the Admin Desk at /admin/desk/omni. This is a visibility and decision-support interface. Admins can see Omni’s recommendations and choose to act on them.
Omni also attaches context to agent routing decisions. When the system routes a request, Omni can include a trust advisory for the relevant operator.
Trust advisory
When Omni processes a request for an operator, it can include a trust summary:
{
"operator_slug": "qtm-detailing",
"active_signal_count": 7,
"has_trust_signal": true,
"trust_state": "present",
"latest_signal_emitted_at": "2026-04-21T..."
}This is labeled trust_advisory_effect: "none" in the system. Omni can see this information but the system does not act on it in v0.01.
Multi-operator isolation
Omni is designed to operate across multiple operators without leaking information between them. Each operator’s data is scoped to that operator. Omni sees global abstracted signals, but operator-private data stays within each account’s boundary.
Current status
Partially implemented in the Planck pilot runtime. Available at /admin/desk/omni in the Admin Desk. Agent coordination and Decision Packet features are assigned but not yet live.
Guardrails
- Preserve the distinction between planned, partial, and live.
- Omni observes and recommends only. It has no execution authority.
- Multi-operator isolation must be preserved.
- Trust advisory data is read-only and does not drive routing.