MVP web client (Phases 1-30) is complete; reorganize planning + living docs around that. - PLAN.md kept as the staged MVP record (1-30) with a status block + pointers; removed the 31-36 stages, regression scenarios, and deferred-TODO section (moved out); fixed a stale cross-machine plan path. - ui/PLAN-finalize.md (new): active web-finalization plan in 8 stages (visual system, a11y, i18n, error UX, PWA, build hygiene, docs, owner manual-QA loop); absorbs former Phases 33 and 35. - ui/ROADMAP.md (new): post-MVP (Wails, Capacitor, realistic projection, acceptance + regression scenarios) and triaged deferred follow-ups. - ui/docs/README.md (new): grouped topic-doc index. - De-archaeologized all 20 ui/docs topic docs + ui/README.md + ui/core/README.md: stripped Phase-N build history, rewritten as current-state; deferred work now points at ROADMAP.md / PLAN-finalize.md. Docs-only; no code change. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
7.7 KiB
UI Client — Post-MVP Roadmap
The MVP web client (Phases 1–30 of PLAN.md) is complete. The near-term web finalization pass — visual system, accessibility, localisation, error UX, PWA, docs — is tracked separately in PLAN-finalize.md.
This roadmap holds the work deliberately deferred until after the web client is finalized: the native platform wrappers, the realistic multi-turn projection (a feature, not polish), the cross-platform acceptance pass, and a set of non-blocking follow-ups. Items keep their original PLAN.md phase numbers for continuity; their full original specs live in the PLAN.md git history.
Phase 31 — Wails Desktop Wrapper
Goal: native desktop app (macOS, Windows, Linux) running the same
frontend bundle, with ui/core embedded as Go instead of WASM.
Key artifacts: ui/desktop/ Wails entry + IPC bindings; per-OS secure
storage (macOS Keychain, Windows DPAPI, Linux Secret Service with a
0600 file fallback); modernc.org/sqlite cache; WailsCore /
WailsKeyStore / WailsCache adapters; OS icons; ui/Makefile
desktop-* targets; ui/docs/desktop-secure-storage.md. Re-evaluate
Wails v2 vs v3 at phase start.
Depends on: Phase 6 (KeyStore/Cache interfaces) and the web form of
Phases 7–30. Acceptance: all three OS binaries launch, log in, and
persist the keypair on a fresh profile from one codebase; Linux file
fallback activates without libsecret and writes 0600.
Phase 32 — Capacitor Mobile Wrapper
Goal: native iOS and Android apps running the same frontend bundle,
calling a gomobile-compiled ui/core.
Key artifacts: ui/mobile-bridge/bridge.go gomobile façade;
ui/Makefile gomobile/ios/android targets; Capacitor project +
custom galaxy-core plugin (Swift + Kotlin); CapacitorCore /
CapacitorKeyStore / CapacitorCache adapters (secure-storage +
sqlite community plugins); app icons + splash; ui/docs/mobile-bridge.md.
Depends on: Phase 6 and the web form of Phases 7–30. Acceptance: iOS
Simulator and Android Emulator launch, log in, and persist the keypair
across restarts (manual smoke); same Core/Storage TS interfaces as
web/desktop. Full Appium automation lands in Phase 36.
Phase 34 — Multi-Turn Projection (realistic planet forecast)
Goal: a realistic multi-turn planet projection, surfaced in the planet inspector and the calculator's planet area. (Reach circles already shipped in Phase 30; this phase no longer owns them.)
The Phase 30 planet area is single-turn (MAT-only). This phase makes it
realistic and multi-turn by extracting the planet economy into pkg/calc
(single-sourced, engine delegates): PlanetProduction,
ProducePopulation, UnpackColonists, UnpackCapital, reusing
ProduceShipsInTurn; plus a ProjectPlanetBuild projector ("K ships in
M turns" under guaranteed per-turn supply) mirroring MakeTurn steps
09/12/14/15. CAP and COL only affect future turns (post-production
unpacking), so they become meaningful here and join MAT as supply inputs
in the calculator's planet area.
Key artifacts: the pkg/calc economy extraction + bridges + Core
typings; planet-inspector forecast section (next-turn population,
industry, materials, progress); calculator planet area gains CAP/COL
supply; ui/docs/multi-turn-projection.md.
Depends on: Phases 17, 18, 30. Acceptance: projector output byte- identical to running the engine's per-turn planet update over the same turns (Go parity); inspector + calculator readouts consistent with it.
Phase 36 — Acceptance Pass
Goal: reconcile implementation, documentation, and regression coverage before declaring the client ready for technical beta.
Key artifacts: final cross-platform regression run on a release
candidate; ui/docs/release-checklist.md; visual-regression baselines;
Appium harness (iOS Simulator + Android Emulator) covering login, push,
and a full turn loop, wired into a macOS-runner CI job as a pre-release
gate; a docs/README/ARCHITECTURE/FUNCTIONAL drift sweep.
Depends on: Phases 1–35 (incl. the finalization plan). Acceptance: implementation matches every documented contract; the cross-cutting regression scenarios below pass on web, desktop, and mobile; Appium smoke passes on iOS and Android in CI.
Cross-cutting regression scenarios
- A fresh device generates a keypair, completes email-code login, and signs a follow-up authenticated request on every platform.
- A returning device resumes its session without re-login, preserves queued orders, and keeps receiving push events without gaps.
- Server-side session revocation tears down the push stream and forces re-login on every platform within one second.
- Tampering with
payload_bytes,payload_hash,request_id,message_type, or any signature byte is rejected byui/corewith a stable error code. - Requests outside the freshness window are rejected before network; a clock-skew warning surfaces when the local clock disagrees beyond it.
- The map renderer holds 60 fps with a 1000-primitive fixture on web (Chrome, Edge, Safari, Firefox), desktop (Wails on mac/win/linux), and mobile (latest iPhone, mid-range Android).
- The sidebar preserves tool state across tab switches; the active view preserves state across view switches.
- Order draft survives reloads, view switches, network drops, and history-mode entry/exit.
- Orders queued offline flush in order on reconnect; a turn-cutoff conflict surfaces as a failed-order banner without retrying forever.
- History mode applies to every view; the order tab disappears in history mode and the prior draft is restored on return.
- The calculator's math matches
pkg/calc/byte-for-byte; drift fails CI. - Linux desktop without Secret Service completes login via the
0600file fallback. - The web service worker invalidates on app update and never serves stale code on the first load after deploy.
- Push-event signature verification is mandatory; any failure tears down the stream and reconnects with backoff.
- Locale switch persists across reloads and applies to every visible string on every platform.
Deferred follow-ups (non-blocking)
Explicit deferral decisions, to fold into the phase noted when it lands. (The "build core.wasm in CI / drop the committed artefact" follow-up moved to PLAN-finalize.md F6, since it is timely.)
- Restore
js.CopyBytesToGowhen TinyGo fixes theinstanceof Uint8Arraycheck — the per-element loop inui/wasm/main.go::copyBytesFromJSis a TinyGo 0.41 workaround. Track upstream and revert once a fixed release is pinned. - Migrate TS codegen to the Connect-ES v2 BSR plugin once published —
ui/buf.gen.yamlrunsprotoc-gen-esv2 locally becausebuf.build/connectrpc/esstill emits v1-incompatible imports. - Rename
gateway/internal/grpcapi/→connectapi/— historical name; the package serves Connect/gRPC/gRPC-Web. Pure rename. (Fold into the next gateway change; not a UI task.) - Rename
GATEWAY_AUTHENTICATED_GRPC_*env vars to drop theGRPCinfix — they label the authenticated edge tier, not the wire protocol. (Coordinate with the rename above; not a UI task.) - Add a Docker-stack Connect end-to-end integration test —
integration/connect_call_test.goexercising a unary + a server-streaming Connect call throughtestenv.Bootstrap. - Battle viewer — push event
game.battle.new— emit a backend notification intent (idempotencybattle-new:<game_id>:<turn>:<battle_id>) so the shell shows a toast deep-linking into the Battle Viewer. Needs an engine emit-side change. - Battle viewer — richer ship-class visuals — derive shape/scale
from mass, armament, shields, and ship count instead of the MVP
one-circle-plus-label per
(race, className).