3.7 KiB
Kaptan Ground Suite — spatial design reset
This board is the implementation benchmark for the persistent operations scene. The supplied Kaptan images are treated as visual-language evidence, not as an information architecture.
Evidence board
| Reference | Borrow | Avoid |
|---|---|---|
| Company of Heroes 2 | Safe left-click selection, additive selection, numbered control groups, contextual command verbs, readable queued intent, tactical overview | Orders that fire from an ambiguous right-click; game-only pause semantics; selection implying command rights |
| Total War battles | Formation footprints, drag-to-orient placement, compact unit cards, group framing, large-force distributions | Screen-filling unit-card chrome; hidden morale/state meanings; instant consequential placement |
| QGroundControl setup | Capability coverage, explicit calibration state, parameter metadata, preflight blockers, geofence concepts | A separate page per subsystem, long forced setup wizard, scalar-only telemetry, forms that hide geography |
| Supplied semantic-zoom image | Strategic → operational → tactical information density; selected/observed/planned separation | Three separate map instances or page transitions to represent zoom |
| Supplied mission overview | Graphite/navy palette, compact status strip, authority/presence language, direct map labels | Top navigation, oversized attention columns, cards competing with the map |
| Supplied task/recovery images | Spatial route preview, exact recommendation comparison, audit timeline | Full-screen recovery page; acceptance before a final revalidation |
| Supplied design-system image | Cyan selection, violet planning, amber attention, red only for danger, non-color marker shapes | Color-only state, generic rectangular button walls, tiny low-contrast text |
Interaction state boundary
Select is the only safe resting state. Map pointer input is routed through one typed interaction controller. A mode owns its cursor, valid targets, pointer/keyboard semantics, draft geometry, completion, cancellation, validation, and output. Space/middle pointer temporarily owns pan. Escape cancels segment → draft → mode. Draft undo/redo never alters observed mission truth.
Selection, local control groups, authoritative mission groups, and authority leases are independent state domains. A command draft captures an immutable target set and is revalidated before the simulated adapter can acknowledge it.
Persistent compositions
All viewports use the same mounted scene and the same regions; only secondary density changes.
| Viewport | Map target | Composition |
|---|---|---|
| 1366×768 | ≥ 52% of useful body area | 42 px status strip; 48 px tool rail; 218 px roster; map center; 286 px inspector; 122 px unit deck; 32 px collapsed timeline |
| 1920×1080 | ≥ 60% | 48 px rail; 238 px roster; 320 px inspector; expanded spatial labels and telemetry charts |
| 3440×1440 | ≥ 68% | fixed secondary rails; map absorbs width; inspector becomes a two-column engineering surface; unit cards remain bounded |
Tools never replace the map. Fleet occupies the roster overlay; deployment and contextual abilities occupy the bottom deck; queue/events/telemetry expand upward; configuration and recovery dock from the right; settings is a compact drawer.
Required high-fidelity states
The live implementation itself is the high-fidelity composition source. Deterministic state presets and screenshot tests cover: default tactical control, large multi-selection, route/formation drafting, task-area drawing, deployment/discovery, calibration, parameter diff, and degraded recovery. Screenshots are captured at 1366×768, 1920×1080, and 3440×1440; no preset is allowed to unmount [data-testid="operational-map"].