- llm.ts: LlmClient.chatWithTools — bounded OpenAI function-calling loop
(default 6 tool rounds, then one forced no-tools request); shares the
retry/backoff/reasoning-effort machinery with chat(); tools are only
added to the request body when non-empty, so plain chat() bodies and
their tests are unchanged; a tool-call response is a success (never a
truncation); malformed tool arguments become an error tool-message.
- engine.ts: GET_NEIGHBOURS_TOOL + makeNeighboursExecutor (degrades
unknown tools/args/keys without throwing); playerFrontier (sorted,
deduped, other-nation-owned only); buildSystemPrompt gains a tool
variant (contiguity rules, frontier hint) used only when the tool is
actually sent; buildUserPrompt emits a frontier field whenever a
resolver is given; runTurn takes getNeighbours and uses
chatWithTools when both the resolver and the client's capability are
present, falling back to plain chat() otherwise.
- applyTurnResult: snapshot-first sole-ownership algorithm — a region
key transfers only with exactly one claimant and exactly one previous
owner (different from the gainer); double claims and pre-existing
double ownership are left untouched; lone losses are no-ops; every
key has 0 or 2+ -> 1 owner post-turn, never orphaned or duplicated.
- stub.ts: deterministic one-region contiguous conquest per turn from
the prompt's frontier field (bypasses the tool loop; empty frontier
-> no transfer, all existing assertions stay stable).
- regions.ts: optional region-adjacency.json load; hasAdjacency() and
neighboursOf(keys) (union, deduped, queried excluded, sorted).
- index.ts: jump route passes the adjacency resolver when loaded.
- tests: chatWithTools protocol (echo, executor args, bounded rounds,
malformed args, retry), tool-path runTurn wiring + fallback,
playerFrontier, executor, applyTurnResult invariants, neighboursOf/
hasAdjacency (105 unit tests, 14 e2e — all green).
Generate public/region-adjacency.json: regionKey -> neighbourKeys, built
from regions.geo.json (bbox proximity reject + @turf/boolean-intersects
arbitration; symmetric, deduped, no self-refs). Deterministic + idempotent.
This is the data backing the LLM's getNeighbours tool (Stage 2) which lets
it move contiguous frontier regions instead of whole countries.
v0.19.0's region-level ownership did not activate in the production image:
server/regions.ts loadRegions only looked in public/, but the Docker image keeps
only dist/ (Vite copies public/* -> dist/ at build), so the loader found nothing
and expandCountriesToKeys degraded to a no-op — fresh games kept raw ISO_A3 codes
instead of region keys (the client's own /region-keys.json fetch still worked,
since the client is served from dist/).
loadRegions now probes public/ (dev) then dist/ (prod, == clientDir) and uses the
first directory holding both files. A new test covers the prod-only (dist/), dev
(public/), and empty-root (no data) layouts.
Version 0.19.0 + changelog. Breaking: pre-v0.19 saved games are dropped
(territory now holds region keys, not ISO_A3 codes) — deploy wipes the
history-data volume.
- map.test.ts: rewrite the fixture for region keys; the #6 regression guard now
proves the owner resolves via properties.regionKey while the repacked numeric
feature.id and the sub-region regionId do not.
- regions.test.ts (new): the loader sees 1,796 valid keys; FRA expands to its
42 sub-region keys (all valid, all labelled '… (FRA)'); allTerritoryKeys
unions correctly; and with no data it degrades to the raw input.
- engine.test.ts: buildUserPrompt includes a regions table only when provided;
applyTurnResult with a validKeys set applies real keys and ignores bogus ones,
and without one applies all (the legacy/unit path).
- src/lib/regions.ts (new): cached, turf-free same-origin fetches of
/region-keys.json (feature id -> key) and /regions.geo.json (id -> name) so
WorldMap and NationPanel share one fetch each.
- WorldMap.tsx: stamps each feature's properties.regionKey (from the manifest,
by its stable id) and resolves owner / flash / hover / click from that key via
the shared mapFeatureOwner; the per-region flash lights the exact sub-region
whose owner changed; click carries only the sub-region name.
- NationPanel.tsx: territory chips show the region's human name (falling back
to the raw key) with the key in the title / data-attr.
- App.tsx: the selected-region dossier label drops the now-unused parent code.
The single source of truth now keys ownership by region key (one per owned
sub-region) instead of a parent-country ISO_A3:
- shared/types.ts, shared/protocol.ts: TerritoryChange / RegionMap / regionOwner
doc flips; mapFeatureOwner now reads the feature's stamped properties.regionKey
(never the repacked numeric feature.id, nor the sub-region id — the #6 guard).
- server/regions.ts (new, turf-free): loads region-keys.json + regions.geo.json
once at boot; expands a nation's ISO_A3 list to its sub-regions' keys, builds
the LLM 'regions' reference table (key -> "name (parent)"), and exposes the
set of valid keys. Degrades gracefully when the files are missing.
- server/engine.ts: buildSystemPrompt instructs the model to use region keys;
buildUserPrompt gains an optional regionReference table; applyTurnResult gains
an optional validKeys set so hallucinated keys are ignored.
- server/index.ts: loads region data at boot; expands each nation's country list
to region keys at game creation; the jump route passes the reference + valid
keys through runTurn.
The stub is unchanged (it emits an empty, region-key-conformant territoryChanges).
Deterministic <name-slug>-<4char> key per sub-region: name via NFKD strip +
fold table (ə ø ð đ œ ł) + a-z0-9 collapse; 4-char from FNV-1a of the stable
feature id, salted on within-slug collision so all 1,796 keys are unique.
Builtins only (no @turf). Idempotent (byte-identical on re-run).
- DELETE /api/games/:id route + Db.delete() (404 when absent)
- PresetPicker: per-saved-game trash button + confirm dialog (Cancel/Delete)
- App: runDeleteGame; useSse onError -> goHome on 'deleted'/'not found'
(a deleted/missing game now returns the live tab and any deep link home)
- DeleteGameResponse in shared/protocol.ts; api.deleteGame()
- unit test (Db.delete) + 2 e2e tests (home-screen delete; in-flight delete
returns the active tab home and a fresh deep link lands on home)
- version 0.7.0 -> 0.8.0 + changelog
Fixes#20
- Add .gitea/workflows/ci.yml: typecheck -> lint -> unit -> build -> deterministic
Playwright e2e against the dev webServer (LLM_STUB=1). No dind/docker/static IP.
- package-lock.json is committed (reproducible `npm ci`); verified npm install is a no-op.
- Add CHANGELOG.md, set version 0.1.0, add .env (gitignored) + .env.example placeholder.
- Document the mandatory feature-branch -> PR -> CI -> auto-merge workflow in
AGENTS.md + README.md (e2e now runs in CI on every PR).
- Add Playwright html/json reporters (playwright-report/, test-results/e2e.json)
for the CI E2E-results summary step and the failure artifact upload.
Map (basemap was rendering black):
- Inline MapLibre style (dark-navy background + subtle OpenFreeMap ne2sr
raster) instead of the remote "dark" style, whose vector tiles are empty at
world zoom and gated the choropleth on a remote styledata fetch.
- Explicit prod worker URL (maplibre-gl-worker via ?url) so GeoJSON sources
process in the production bundle.
- Hardened the choropleth against geo load-order races (empty sources up
front, ownership stamped once data lands) and bumped fill opacity so the
vivid per-nation colors are the star on the non-black background.
Save/resume (a refresh used to drop users to the home screen):
- Persist the active game id in localStorage (history.activeGameId) on create
and resume; clear it when explicitly starting a new game.
- Auto-resume the last game on a page refresh (GET /api/games/:id + /map),
re-attaching the SSE stream; fall back to home gracefully if it's gone.
- Home screen: prominent "Resume last game" button + a "Saved games" section
(newest first). New GET /api/games list endpoint.
e2e: two new deterministic tests — auto-resume-on-reload and explicit resume
from the home screen (both under LLM_STUB=1).
Quality gates: typecheck, lint, 21 unit tests, 3 e2e tests — all pass.
- MapLibre map is now a fixed, full-viewport background layer (z-0)
- New FloatingWindow: drag by title bar (pointer capture, viewport-clamped,
position persisted to localStorage), collapsible to title bar, bring-to-front
z-index; semi-transparent slate window with amber/cyan accents
- Game screen re-laid out as floating windows over the map: slim top header
(scenario, turn + date, player nation chip, Idle/Simulating status, Focus map
collapse-all), dossier (top-left), event feed (right), Your move deck
(bottom-center), legend (bottom-left), tiny zoom control (bottom-right)
- MapLibre default nav controls removed; custom floating zoom-in/out control
- Panels gain an inWindow mode (no card chrome, same testids)
- server: honor HISTORY_API_PORT as PORT fallback (dev proxy/e2e agreement)
- css: keep maplibre canvas base color and Radix portals above HUD windows
Map (WorldMap):
- Keep the near-black OpenFreeMap `dark` basemap (verified HTTP 200; the
suggested dark-matter / dark-outdoors / liberty-dark variants 404 on that
host) so the choropleth nation colors pop
- Stable per-nation fills at 0.8 opacity, thin subtle borders, and a brighter
thicker amber border for the player nation so "this is me" is always obvious
- Hover brightens a region and shows a color-chipped owner + key stats popup;
click opens a full dossier popup and focuses the left panel on that nation
- Transient, self-cleaning flash (pulsing amber outline) over regions whose
ownership changed this turn
- Bottom-left legend auto-generated from the region map; per-scenario initial
camera (Europe for the WWII scenarios, wider for Cold War / Modern / Aegis)
Command-center UI:
- New header bar: scenario title, prominent turn + date, the player nation
(name + color chip), and a live Idle / "Simulating…" status indicator
- 3-column operations-console grid: nation dossier (left) / dominant map
(center) / command deck + event feed (right); stacks map-first on mobile
- EventFeed: one card per event with a category icon, a colored outcome badge
(success/failure/partial/neutral), a nation-colored accent edge, and the
turn + date; new events slide in and flash at the top
- NationPanel: value-colored animated stat bars (red → green), territory as
color-coded region chips, and diplomacy as a list with a colored indicator
per relation (ally/rival/war/neutral)
- ActionPanel: a "Your move" composer with one prominent Send button and a
spinner + "turns can take a few seconds" hint while a turn resolves
- sonner error toasts: a 502 / model-down surfaces as "AI temporarily
unavailable — try again"
Live LLM rewire (server, previously uncommitted):
- Default endpoint https://ai.kkhost.pl/v1 (vLLM, no API key); reasoning model
qwen3.8-27b; always send a generous max_tokens (4000) so it finishes its
chain-of-thought before emitting content; read only `content`, ignore
`reasoning`. docker-compose + AGENTS.md updated to match.
Tests: the deterministic Playwright e2e gains assertions for the dossier stat
bars, the map legend (all nations), and the event-card type icon + outcome
badge; it stays fully deterministic under LLM_STUB=1.
kamilkrzywda
