FleetCommander — platform patterns (mine, pieces only)

Policies that keep the org coherent

dotnetgovernancenovolis

Status (2026-05-25): WEGO replay shipped as Novolis.Simulation.Replay (+ optional ISimulationEventStore<T> spike). FleetCommander repo should adopt via PackageReference in headless tests. Master plan: platform-import-plan.md.

Source: D:\repos\FleetCommander (also on D:\github\FleetCommander) Verdict: Product — do not import the repo. Extract patterns into novolis-simulation, novolis-transports, novolis-storage, novolis-machinelearning as needed.

What (worth borrowing)

FleetCommander area	Types / behavior	Novolis target
WEGO simulation	`RoundPlan` → lock → `RoundSimulator.Resolve` → `RoundPackage` / `GameEvent` list	`Novolis.Simulation.*` — deterministic round pipeline
Replay	`ReplayTimeline`, `RoundRecord`, `ReplayScrubber`, MessagePack `RoundPackageMessagePack`	New facet e.g. `Novolis.Simulation.Replay`
Interpolation	`Interpolate` / `InterpolateFromPackage` for render scrub	Simulation.View or app bridge
Transport	`FleetCommander.Transport.Abstractions` + InProcess `Channel` + SignalR	Inspiration for test transport; SignalR stays in apps
Scenarios	JSON scenarios under `.storage/blob/scenarios/`	`Novolis.Simulation.World.Builders` or scenario loader
Headless CLI	`FleetCommander.Headless` play-scenario	Dogfood pattern for simulation tests
AI	ONNX + heuristics; uses Frank.ML.Foundation.Neural NuGet	Already aligned with `novolis-machinelearning`; keep training in product
Rendering	Veldrid + Raylib adapters	Skip — use `novolis-rendering` + `novolis-raylib`; FC renderers are product-specific

Why

simulation-layer-policy.md lists replay under Simulation; `Novolis.Simulation.Replay` is implemented (WEGO-style timelines; racing still uses ticks).
FleetCommander is the most mature deterministic sim + replay + transport Frank product on disk.
Full import would violate boundaries (Veldrid, SignalR, Azure storage, game domain in platform).

How

P1 — Replay facet (shipped)

Status: Novolis.Simulation.Replay in novolis-simulation (SimulationTimeline, InMemorySimulationRecorder, ReplayPlayback, SimultaneousPlanBuffer).

Original spike steps:

~~Spike in novolis-simulation (new project Novolis.Simulation.Replay):~~

ISimulationRecorder / ISimulationPlayback with immutable snapshots (BCL + optional MessagePack).
Borrow data model from RoundRecord / ReplayTimeline, not FC namespaces or ship/combat types.

Unit tests: record 10 ticks of Novolis.Simulation.Racing or a minimal SimulationWorld → playback equality.
No PackageReference to FleetCommander.

P2 — WEGO planning (only if SCR/WEGO product needs it)

PlanBuffer + simultaneous commit phase — generic names in Novolis.Simulation.Abstractions.
Document difference vs real-time SimulationClock tick.

P3 — Scenario JSON loader

Extract schema style from FC scenario files for dogfood; keep game content in apps.

Skip (duplicates)

FC piece	Already in Novolis
Raylib rendering project	`novolis-raylib`
Neural foundation	`novolis-machinelearning`
General physics	`novolis-physics`
Aspire host	`novolis-aspire` (separate)

Acceptance

Replay facet documented in library-boundaries.md placement table when implemented.
Zero FleetCommander.* references in platform packages.