Fleet Monad — Sea Trial Acceptance Report

Infrastructure verified

✓
rock64 fully retired from the public path.
Router forwards 80/443 directly to Granite (192.168.0.100); no intermediate proxy hop remains.
✓
Real HTTPS, terminated on Granite itself.
Let's Encrypt certificate issued via Caddy automatic HTTPS for cameronlampley.com; plain HTTP redirects (308) to HTTPS.
✓
Bare domain and /monad/ prefix both serve the app directly.
cameronlampley.com/ and cameronlampley.com/monad/ both return 200 with no redirect chain between them.
✓
Firewall active, scoped to required ports only.
ufw allows SSH (22), HTTP (80), HTTPS (443) inbound; everything else denied by default.
✓
caddy, fleetcore-serve, and monad-lan-web all running as durable systemd units.

FleetCore Live Backend verified

✓
WebSocket path confirmed end to end, not just the handshake.
A real client connection to wss://cameronlampley.com/monad/fleetcore-ws/ws received a "connected" message followed by a live, continuously updating world snapshot stream.
✓
Command path verified: spawn, despawn, set-route, set-escort-mode all round-trip correctly against the live server.
✓
World reset to a clean baseline for testing: flagship + 3 scouts, escort mode off.

Escort Mode — Real Sim-Engine Motion verified

What changed

Escort Mode was previously client-side cosmetic drift, disabled outright in live mode. It is now genuine per-tick station-keeping in the FleetCore sim engine — scouts chase a continuously recomputed formation target relative to the flagship's live position and heading, using the same route-following physics as Set Waypoint and passive traffic.

Mode	Behavior verified	Measurement
Loose	Scouts converge and hold at range	~1200m radius
Tight	Scouts hold a close screen	exactly 350m radius, all 3 scouts
Patrol	Continuous bearing sweep, even vs. a stationary leader	145.2° → 129.4° → 115.6° → 105.9° over 4 samples

✓
Backend test suite green (cargo test, incl. determinism/replay test).
✓
Real headless-browser pass (Playwright/Chromium): page connects to FleetCore Live automatically, Escort Mode button cycles all four states correctly, zero console errors, visual render confirmed via screenshot.

Bridge Station Consolidation verified

✓
Bridge Station 2.0 and 2.1 fully retired — source removed, ad hoc dev-server processes killed, no dangling links.
✓
Bridge Station 3.0 rebuilt with relative asset paths and the public FleetCore WebSocket URL, deployed under web/toys/ like every other public toy — no separate port.
✓
Linked from the homepage and Command Deck, distinctly labeled to avoid confusion with the original (unrelated) Bridge Station "Live Console."

Live Course-Change Validation verified

What this adds

A structured VesselEvent stream (WaypointReached / RouteReplaced / RouteCompleted / Holding) now exists specifically to answer a question the client previously had to guess at: was an arrival a genuine autonomous completion, or did an operator replace the route first? Each Vessel carries a route_id, bumped on every real SetRoute, so events can be tied to the route they belong to. Escort Mode's own per-tick station-chasing deliberately does not bump route_id or emit an event — that's continuous station-keeping, not an operator order.

✓
Clean natural leg transition observed and validated against full acceptance criteria: exact position snap, route 4→3 legs, course jumped 83.5°→92.4° on the same tick, status held Underway throughout, WaypointReached emitted, zero unexpected events.
✓
Correctly distinguished manual reroutes from natural arrivals under real conditions.
Several rapid manual course changes during live testing correctly produced RouteReplaced events rather than being misread as arrivals — the validator flagged each one rather than false-passing on a route-length coincidence.
✓
Fleet Motion's ETA display bug found and fixed.
formatEta() was computing live-mode ETA against a local-simulation-only design constant (180 km/h) instead of the vessel's real live speed (72 km/h) — a ~2.5x error. Fixed and verified exact against server ground truth at both 1x (91.9km @ 72km/h → 1h17m) and 10x (91.5km @ 720km/h → 8m).
✓
Scout speed correction: escorts were slower than the flagship (16-18 m/s vs. 20 m/s).
Since the escort station target moves with the flagship every tick, a slower escort can never close the gap during sustained transit — discovered live, on a 12-leg flagship course. Fixed to 1.4-1.6x the flagship's speed (32.0/30.0/28.0 m/s) in the seed, the Rust initial_conditions(), and patched directly into the live world without disturbing the in-progress course or event history. Verified: all three scouts closing real distance afterward, where they previously could not.

Human Acceptance Pass — Captain's Chrome Test confirmed

✓
Homepage status strip (FLEET STATUS: GREEN / SERVED FROM: GRANITE) confirmed matching live source, byte for byte.
✓
Fleet Motion confirmed reading "FleetCore Live," not "Local Simulation."
✓
Observed flagship position west of the Strait of Hormuz, cross-checked against server state — confirmed genuine scripted transit (seed route, course 270°), not a rendering fault.

Open Items — Not Blocking

!
Two staff portraits missing on the roster (Commander Claude, Lt. Reyes) — files never existed on disk. Cosmetic; falls back to placeholder, not broken.
!
/monad/portainer/ is publicly reachable, contradicting the original "Portainer is not exposed publicly" doctrine note. Flagged, not yet resolved.
!
Generated 3D assets (.glb) are large (30–57MB each), served uncompressed over a residential upload link. Works, but not efficient at any real scale.

Wardroom Debrief

Sea trial concluded. Three real defects were found and fixed through live testing rather than assumption: the ETA display using the wrong speed source, escort formation targets recomputing correctly but escorts too slow to ever reach them, and — earlier in the trial — Fleet Motion silently stuck on Local Simulation instead of the live feed. All three were caught by actually exercising the running system against ground truth, not by inspection alone. The three open items below remain open by choice, not oversight — Portainer's public exposure specifically deferred at the Captain's order, not resolved this trial.

SEA TRIAL ACCEPTANCE REPORT