Industries
Not a slide about a sector — the object types, the link types and the counts that are in the database right now. Everything below this line that claims to be built can be counted, and the count comes from the store, not from this page.
Modelled today
Both packs seed from ATLAS_SEED=20261016 under ontology v1.0.1, so the figures below reproduce on another Mac.
21 object types · defense pack
Ships, berths, units, crews, missions, sensor payloads, spares and work orders in one graph. The demonstration journey breaks a platform in a maintenance record, watches the readiness margin fall to −6, asks the copilot why a named mission is short of platforms, then commits a reserve vessel and watches the margin return to +6 without a page reload.
21 object types · commercial pack
Depots, stock lots, order lines, inbound shipments, suppliers and transfer lead times, with the distinction that makes control towers honest: on hand versus planned supply. A 48-hour port delay moves the shortfall from 20 units to 220, and two reviewed interventions close it to 0 at a committed SGD 1,380.00.
Object coverage
42 distinct types across both packs. Read live from http://127.0.0.1:18180/sql on 2026-10-18 with SELECT segment, objectType, count() AS n FROM entity GROUP BY segment, objectType ORDER BY segment, objectType; and cross-checked against the API’s own per-workspace counts.
| Object type | Commercial | Defense |
|---|---|---|
| Asset | 24 | 34 |
| Capability | — | 12 |
| Component | 40 | — |
| Contract | 18 | — |
| Customer | 36 | — |
| Decision | 14 | 16 |
| Depot | — | 6 |
| Evidence | 76 | 97 |
| Incident | — | 24 |
| InventoryLot | 53 | 50 |
| InventoryTransfer | 13 | — |
| MaintenanceWorkOrder | 24 | 35 |
| Observation | 36 | 55 |
| Order | 33 | — |
| OrderLine | 65 | — |
| Organization | 12 | 3 |
| PersonnelRole | — | 28 |
| Plant | 5 | — |
| Product | 34 | — |
| ResponseMission | — | 18 |
| Risk | 14 | — |
| Route | 26 | 30 |
| Sensor | — | 30 |
| Shipment | 28 | 30 |
| Site | — | 28 |
| Supplier | 24 | — |
| SupplyItem | — | 40 |
| Team | — | 26 |
| Unit | — | 10 |
| Vehicle | — | 36 |
| Vessel | — | 7 |
| Warehouse | 8 | — |
| e2e_inspection_mvdoeyya | 1 | — |
Column totals are 584 and 615 entities — the same two figures printed in the header stat bar, arrived at by a second query path. If they ever disagree, the build fails rather than publish.
Transferability
A sector is a shape of question, not a logo. These are the primitives a sector needs and the honest state of each one in this build.
Capacity against demand, with the shortfall named
BUILT
Readiness margin and shortage-by-deadline are stored functions, and the tile prints its own subtraction. Two packs exercise it: platforms per mission, units per order deadline.
Time-to-arrive deciding whether a number is real
BUILT
Timely versus late arrivals are separate arithmetic. A transfer in transit is planned supply and never inflates on hand — the commercial pack holds 80 on hand while planned supply reads 140.
Provenance for every claim on screen
BUILT
81 evidence artefacts in the commercial pack and 97 in defense, joined by EVIDENCES edges and walked in the Evidence screen.
A write path that can be refused
BUILT
Seven governed action types with blocking and advisory preconditions, a preview that writes nothing, and an idempotency key on commit.
Coordinates, movement and live sensor feeds
PARTIAL
Entities carry an optional spatial slot and the scene renders it, but there is no AIS or telemetry ingest yet. Position is asserted, not received.
Optimisation across many variables
NOT BUILT
Costed interventions and what-if branches exist; a solver that proposes the cheapest feasible plan is after M6 and is not built.
Multi-user collaboration and audit workflow
NOT BUILT
Auth is a single-user loopback session. Roles and scope enforcement exist and are tested, but there is one session, not a team.
Where it fits next
Two sectors are built and demonstrated. The rest state what would have to change before anyone quoted you a delivery date.
Built and demonstrated
This is where the defense pack lives: berths, platforms, missions, readiness capacity and maintenance events, with a margin formula printed on the tile.
Live AIS feed and real port-call history would replace the synthetic arrivals.
Built and demonstrated
The commercial pack models depots, stock, order lines, inbound shipments and transfer lead times, and distinguishes on hand from planned supply.
An ERP extract with order, on-hand and in-transit tables is enough to reproduce the story.
Target sector · not built
Fleet readiness, spool availability and dock scheduling share one shape with what is already modelled: capacity against demand, with a named shortfall.
Would need a classified-data deployment; the POC runs on synthetic records only.
Target sector · not built
Substations, feeders, generation and work orders are entities with locations and dependencies; outage propagation is a traversal, which is the primitive this platform is built on.
Would need a link-type review against your asset registry, plus a historian feed for time series.
Target sector · not built
Pre-positioned stock, routable warehouses and demand across a disrupted network — the timely-versus-late distinction matters more here than anywhere.
Would need lead-time realism per corridor and a donor-visible export of the assumptions.
Target sector · not built
Most stalled twins are visualization projects without a write path. Here a commit is a first-class action with preconditions, a journal and a branch you can compare.
Would need one decision that is worth closing, chosen before any modelling starts.
Send one table of entities, one table of relationships and the decision you make weekly from them. We will say within a week whether the existing primitives carry it or whether it needs a new link type.
No matches. Try shortfall, SSE or ontology.