Let an operator disable purchases live from the admin — a whole rail/channel or
one account — and show the user a localized reason on their next attempt, so a
provider outage or a misconfig is explained instead of a silent dead button.
- rail kill switch (payments.rail_status, per rail direct:web / direct:android /
vk / telegram): enabled + a per-language message, edited on the catalog page.
Fail-open — a rail with no row stays enabled, so payments are never accidentally
killed. The intake gate (CanPurchase in handleWalletOrder, before the order)
returns payment_unavailable + the localized message, orthogonal to the security
gates.
- per-account override (payments.account_payment_override, a row only for
non-default): allow / deny / default, edited on the user card. "allow" bypasses
ONLY the ops rail switch, never the security gates (trusted platform, the email
anchor, the VK-iOS freeze, the min client version).
- wire: an additive ExecuteResponse.message envelope field (frozen-contract-safe);
the gateway forwards a backend domain-error message; the client shows it on a
payment_unavailable buy attempt.
- admin: rail toggles on the catalog page, the override control on the user card.
- tests: the pure gate (unit, TDD), the store + gate + override end-to-end
(integration, migration 00016), the client (svelte-check / vitest).
- docs: PAYMENTS (+ru), the decisions log (D45/D46). Fiscalization stays
cabinet-side (owner decision) — no itemized-receipt code.
Contour-safe: additive migration (two new tables, no wipe), the wire add is
additive, and fail-open so nothing is disabled until an operator acts.
Split the single Robokassa direct rail into one merchant shop per channel
(web / android; ios later), chosen by the trusted X-Platform subtype, while
every shop still credits the one `direct` wallet — merchant-account separation
for accounting and receipts, not a new wallet.
- config: a channel-keyed shops registry (web seeds from the legacy vars or
_WEB_*, android from _ANDROID_*); an empty shop leaves the rail dormant;
per-shop validation.
- intake: the order picks its shop by subtype (unknown falls back to web); the
per-shop Result callback is verified by that shop's own Password2 at
/pay/robokassa/result/<channel> (the gateway extracts the channel; Caddy's
/pay/* glob already forwards it — no Caddyfile change).
- persistence: an additive `shop` column on the order (migration 00015),
recorded from the payment context, surfaced per entry in the admin report.
- standalone apps sign in by email only, so a direct purchase keeps its email
anchor.
- docs: PAYMENTS (+ru) topology, deploy env vars + compose mapping, the
decisions log (D41 revised for the ИП / 54-ФЗ move; D42-D44), the plan.
Contour-safe: dormant until shops are configured; the migration is additive
(no wipe); no client wire change. Fiscalization (Receipt/Email) and the gateway
`direct/android` subtype follow when the ИП / RuStore are live.
The tg (bot) VPS came up on Europe/Moscow while the main host is UTC, so host-level
timestamps (journald, file mtimes, cron) sat 3h apart across the fleet. Add a
community.general.timezone task to the common role so every provisioned host is UTC
(idempotent — a no-op on the already-UTC main host). Applied live to the tg host in
the same change; the containerised services run in UTC regardless, so no restart.
The backend recorded 172.19.0.9 — the gateway's own docker connection address — as
the client IP for all users: the account's last-login IP shown in the admin console,
and it never reached the backend access log. The gateway forwarded the client IP as
X-Forwarded-For only on chat/feedback calls; every other backend call (including the
profile fetch that stamps last_login_ip) sent none, so the backend fell back to the
peer address.
Carry the client IP on the request context (WithClientIP, mirroring WithPlatform) and
set it once per request in the Connect edge, so the backend client injects
X-Forwarded-For on every downstream REST call. Also add the resolved client IP to the
backend access log.
Test: WithClientIP rides a non-chat call (Profile) as X-Forwarded-For, and is absent
when no IP is set. Docs updated (ARCHITECTURE gateway↔backend + edge).
The "watch for chips" rewarded payout and its per-day / per-hour caps live in the
shared payments.config row and had no admin surface — they could only be changed by
SQL, contrary to D32 (the rewarded rate is meant to be admin-configurable). Add a
"Rewarded ads" form on the /_gm/catalog page: RewardConfig / SetRewardConfig on the
service (non-negative validated), a setRewardConfig store writer (the singleton
config row), the consoleSetReward handler + POST /_gm/catalog/reward route, and the
form pre-filled from the current config. No migration — the columns already exist.
The reward rate is a config value, not a sellable catalog atom (so a "no-ads forever"
style product is still out of scope by D32/D33).
Test: an integration test sets the config, checks the page pre-fill, and refuses a
negative value.
The wallet storefront listed products in creation order — its purchases and its
chip-exchange values were unsorted, unlike the admin console and the public offer.
Extract the ordering (chip packs first by rouble price, then chip-priced values
grouped hints → no-ads → combo → tournament and by chip price) into one comparator,
compareCatalogRank over a small rank key, and apply it at all three sites
(projectCatalog, projectOfferPricing, SortAdminCatalog) — so the subgroup ranking
(valueGroup) and the full order now live in one place. The rewarded "watch for
chips" CTA is a wallet-driven element rendered above the packs, so it is unaffected
and stays on top.
Also add the admin catalog active/all toggle: AdminCatalog takes includeInactive,
the console shows active products by default and ?all=1 lists archived ones too
(the detail and grant forms keep loading all products).
Tests: a storefront canonical-order unit test (reproduced the unsorted bug first);
an integration test for the active/all toggle. Existing offer/admin order tests
unchanged — behaviour preserved.
Two fixes from contour testing of the offline-in-game UX:
- The dictionary word check was unusable offline (Check button disabled) when the
panel was reached while already offline: CheckScreen fetched the variant + pinned
version over the network in onMount, which fails offline, leaving the default
variant so input sanitising stripped every letter. Seed both from the cached game
(present once the board has opened) so the input and the on-device dawg fallback
work without a round-trip; the network refresh still runs when online and on a
cold deep-link.
- The resign and chat controls were only functionally disabled (chat) or not gated
at all (resign) offline, so they still looked active. Hide both while offline,
matching the frozen social controls.
Tests: the in-game offline e2e now asserts the drawer action icons are hidden (not
disabled); a new e2e guards the offline dictionary flow. Docs updated.
CheckScreen's static import of dict/check pulled the dawg loader/reader into the app
entry bundle, pushing it 0.8 KB over the 130 KB budget (CI bundle-size gate). The
dict subsystem is lazy everywhere else, so import localWordCheck dynamically in the
offline branch — it loads only when an offline check actually runs. Main entry back
to 129.0 KB.
When an online game loses the connection the game screen now says so and freezes
instead of silently greying out: a "connection lost" banner appears and the rack,
the move controls, the add-friend/block controls and the chat/dictionary entry all
disable (a started move stays a draft, committed by the player on reconnect). It is
driven off the net-state machine (netState.offline), so it also covers the
Telegram/VK mini-apps, where a lost connection was previously mute in-game.
If the player is already in the dictionary when the drop happens, the word check
falls back to the game's pinned on-device dictionary (exact when that dawg is
cached; "unavailable offline" otherwise) and the network-only complaint + external
look-up hide. Chat, when already open, keeps its existing read-only degrade
(send/nudge disable). New pure helper localWordCheck is unit-tested; the in-game
gating gets an e2e.
An idle live stream whose connection dies without an error (airplane mode
cutting the radio) left the net-state machine stuck "online": the streaming
fetch neither errored nor delivered, no unary call was made, and the OS
`navigator` offline hint is unreliable in the Telegram/VK WebViews — so the
"Connecting…"/offline state only surfaced on the next foreground resync, not
while idle.
The gateway already pushes a keep-alive `heartbeat` every 10s, so add a
client-side watchdog (streamwatchdog.ts) that resets on every delivered event
and, after ~25s of silence, treats the stream as dropped (reportOffline +
reconnect). It is background-aware (paused while suspended) so a throttled
foreground return does not false-trip. The mock client mirrors the heartbeat
so an idle e2e session stays online.
Telegram non-disableably auto-pins each channel post it auto-forwards
into the linked discussion group. The bot now detects that message by
Message.is_automatic_forward in the moderated chat (TELEGRAM_CHAT_ID)
and unpins it by id, so a pin set by a human admin — or by the bot for
another message — is never touched (no unpinAllChatMessages).
Needs the can_pin_messages right in the chat; the startup self-check
now also warns when it is missing. Bot-only; no wire/schema/DB change.
Opening the dedicated /vk/ entry directly in an ordinary browser (no signed VK
launch) fell through to the web flow and silently started a throwaway guest,
which is wrong for a VK-only entry. Mirror the existing /telegram/ launch-error
behaviour: render a compact, shareable, privacy-safe diagnostic screen instead.
- Boot: a new branch `onVKPath() && !insideVK()` sets app.launchError and stops
the fall-through, the VK counterpart of the /telegram/ diagnostic guard.
- Diagnostic (passive, no VK Bridge round-trip): reads the URL launch-parameter
NAMES (never the `sign` value — auth material), whether the URL was signed,
`vk_platform`, the iframe/referrer context, plus the shared client-environment
lines. VK signs the launch URL at load, so — unlike Telegram's initData — the
parameters never arrive late; hence the VK screen offers Share only, no Retry.
- Generalise the screen: TelegramLaunchError.svelte -> LaunchError.svelte, which
renders a neutral pre-formatted report and a per-platform title, with Retry
gated on a `retry` flag (Telegram true, VK false). app.launchError is now a
neutral LaunchDiag { platform, report, retry }.
- New lib/launchdiag.ts holds the shared pieces both platforms reuse: the
LaunchDiag shape, the client-environment lines, and the query field-NAME
reader (moved out of telegram.ts so VK does not duplicate them).
Tests: pure vkDiagLines units, incl. a guard that the `sign` value never leaks
into the report. i18n: launch.errorTitleVk (en/ru). Docs: FUNCTIONAL (+ru) user
story and ARCHITECTURE entry-path note.
A fresh email login that hit the per-IP email rate limit stranded the user in
an unrecoverable "offline" state that survived a full PWA restart, even though
the network was fine.
Root cause: the gateway email class (auth.email.request + auth.email.login,
keyed per IP, 5/10min burst 2) trips on the third event, which in the natural
"request code -> wrong code -> correct code" sequence is the correct-code login.
The gateway returns ResourceExhausted; the client mapped it to 'rate_limited',
which retry.ts classified as retryable + a connection code, so exec() called
reportOffline(). That pushes the net-state machine into connecting, whose
recovery probe is an authenticated profile.get -- but the login screen has no
session, so the probe can never succeed and the machine latches offline. The
transport kill switch (assertOnline) then refuses the very login that would fix
it, and because the IP's email bucket refills only 1 token / 120s, each fresh
attempt after a restart is rate-limited again -> offline again.
Fix (client, narrow -- the trigger):
- A rate-limit is no longer treated as connectivity. retry.ts no longer marks
'rate_limited' retryable or a connection code, so exec() never reports it
offline; it surfaces as the existing error.rate_limited "slow down" message.
Not auto-retrying it also stops the ~20s button freeze and avoids feeding the
gateway's ban tripwire with 6 extra rejections per attempt.
Fix (server):
- Raise the email-code burst 2 -> 4 so the honest request + a mistyped code +
the correct one is not throttled mid-login. Defence-in-depth over the
backend's own per-code guards (5-attempt cap + 15-min TTL + send throttle).
Tests: retry classification units updated to the new semantics; a gateway
regression guard asserts the honest three-event email flow passes under the
default policy. gateway/README.md rate-limit note updated.
The deeper gap (the net-state recovery probe is session-gated, so any real
transport failure on the session-less login/confirm screens still cannot
self-heal) is left as a known residual, deferred by the owner.
The SVG tracer export failed, so the icon was designed collaboratively from
scratch: a wooden «Э» tile with a ✻ score-subscript in Spectral Bold, wood
grain and diagonal light/shadow. Construction is specified in fractions of the
side in docs/ICON_BRANDBOOK.md; the reference generator, pinned Spectral font
and committed masters live in assets/icons/brand/.
One master -> the whole set (assets/icons/brand/build-set.mjs + build-android-res.mjs):
- web (ui/public): favicon.svg (light+dark via prefers-color-scheme), favicon.ico,
apple-touch, PWA any + maskable (+ dark variants), og-image reskinned to
«Эрудит» / Игра в слова
- Capacitor layers: ui/assets/{icon,icon-foreground,icon-background}.png
- Android launcher res, all densities, NO inset, + monochrome themed layer
(anydpi xml: background+foreground+monochrome)
- manual-upload art: assets/icons/brand/{vk,tg,store} (square, no rounding)
Primary variant light; dark where the platform can theme (favicon.svg). The old
LiberationSans generator (assets/icons/build) is removed; docs/ICONS.md retargeted
onto the single master. ANDROID_PLAN icon-rebrand item marked done.
Add a RESUME block to ANDROID_PLAN.md capturing the current state so a fresh session continues without re-asking: the legal-documents track (PR #253, merged into development) and the release-prep hygiene (PR #254, open, this branch) are done; the icon vector, keystore and RuStore account remain (owner-side). Records the locked legal decisions and the exact remaining release chain, and points the icon work at docs/ICONS.md.
Reword every code comment and doc that referenced ANDROID_PLAN.md or its stage anchors (§E, G-step-0, O1) so the living docs no longer depend on the plan file: .claude/CLAUDE.md, deploy/README.md, docs/ARCHITECTURE.md, docs/TESTING.md, ui/README.md, ui/android/.gitignore, ui/e2e/native.spec.ts, ui/src/lib/netstate.ts.
Re-enable the manual signed-APK workflow (android-build.yaml.disabled -> android-build.yaml; the CI host already has the Android SDK; workflow_dispatch-only, so it does not auto-run on this PR). Add docs/ICONS.md — the single-master icon/logo format reference (web, PWA, Android, future iOS, store listings).
The price list is spliced in Russian; the client-side offer transliteration turned its section headings and column headers into transliteration too. Translate the known backend-projected strings (the two section headings and the column headers Наименование/Рубли/Голоса в VK/Stars в Telegram/«Фишки») to English and transliterate only the product names, as requested.
Add English versions of the privacy policy, EULA and offer (ui/legal/*_en.md) and render each legal page as one self-contained bilingual document: both language bodies plus a header language toggle and theme toggle (no back), a small inline ES5 script that switches language client-side (no reload, default Russian + persisted) and applies the theme (system default + persisted override) — mirroring the landing. The offer's English view transliterates the Russian product names spliced from the live catalog.
renderLegalHtml now takes both language sources; renderOffer splices the price list into both; the renderer bakes and reads the _en sources and pre-renders the static pages at boot. Also wrap the /offer/ contour probe in the same retry+timeout as the legal probe (the offer page is now bilingual and larger). Docs + renderer/ui tests updated.
The shared legal-page template inherited the offer's shrink-to-fit product-name column (width:1% + white-space:nowrap), which stopped the privacy data table's long prose first column from wrapping and blew the table out horizontally. Scope that rule to the offer (body.offer) and top-align all legal-table cells for the multi-line privacy rows.
The /eula/ probe fetched the full ~129 KB page. While the monitoring stack boots after the rolling deploy and the CPU-capped (1 CPU / 192 MB) renderer is starved, that transfer exceeded the wget timeout for the entire 60s retry window, while the 3x-smaller /privacy/ passed — pre-rendering the pages did not help because the bottleneck is the transfer, not the render. Fetch only the <head> (head -c 4096) and assert the page's canonical URL, which uniquely identifies the rendered legal doc reaching the edge (the landing's canonical is erudit-game.ru/); the check now transfers ~4 KB regardless of page size.
Serving /eula/ re-parsed the large (~122 KB) EULA markdown on every request. Under the rolling deploy's host contention (the renderer is capped at 1 CPU / 192 MB) that was slow enough that the contour /eula/ probe failed for its whole 60s retry window, while the smaller /privacy/ squeaked through; both serve fine once the host is idle. Render privacy + eula once at boot and serve the cached HTML (now ~1.5 ms, a memcpy). The offer stays per-request for its live price splice; the probe retry stays as a readiness backstop.
The single-shot probe raced the render sidecar during the rolling deploy: across two attempts a different route flaked each time (/privacy/, then /eula/) against an otherwise-healthy renderer, while the contour served all three pages correctly — a deploy-time transient (a slow first render of the large EULA while every container recreates, or a connection blip). Wrap the check in the same 20x3s retry the landing/gateway/backend probe uses, with a 10s per-attempt wget timeout, so it waits the transient out instead of failing the deploy.
Author ui/legal/{privacy,eula}_ru.md, reworked from the source documents: the seller INN is unified (290210610742), contacts unified to the Telegram bot + email + postal address, the EULA governing-law clause leads with Russian law + jurisdiction as residence tiers, the single-binding-language clause is dropped, data collection is scoped to voluntary/support provision, and "Компания" is no longer shout-cased.
Serve both as static pages through the render sidecar, reusing a shared renderLegalHtml generalised from renderOfferHtml: new GET /privacy/ and GET /eula/ (301 from the slashless form), the markdown baked into the image, no backend fetch. Route them at the edge via the @legal caddy matcher with a CI probe asserting each page's canonical URL + the seller INN, so a missing route cannot silently fall through to the landing. Add the two links to the landing footer.
Docs baked in: ARCHITECTURE (renderer legal pages + edge routing), FUNCTIONAL (+_ru) footer, renderer/README routes. Tests: renderer legal.test.mjs, ui renderLegalHtml unit, landing footer e2e.
Intercept /support in the main bot with a dedicated handler (registered
like /start), replying with a fixed support-desk info message: the
operators' working hours and what to include (a description and, when
possible, screenshots). The reply is Russian or English by the sender's
reported Telegram language, and the command is listed in the bot's
command menu (localized).
Being a dedicated handler, it intercepts /support before the support
relay, so the command line itself is not forwarded into an operator
topic while the user's following description still is.
Update the telegram README and FUNCTIONAL.md (+ _ru mirror).
Land the offline-model redesign (ANDROID_PLAN.md O1-O7): replace the explicit offline toggle with a single detected net-state machine, unify the lobby, and add a two-tier client-version gate. Contour-safe: both version vars empty => dormant, the wire change is additive, so web/pwa/vk/tg behaviour is unchanged unless the gate is deliberately configured.
O1 net-state reducer (test-first). O2 store + wiring (connection/offline shims; +@capacitor/network). O3 remove the offline toggle + migrate the pref. O4 two-tier gate: hard update_required degrades to an offline Update/Play-offline notice (not terminal); soft GATEWAY_RECOMMENDED_CLIENT_VERSION -> X-Update-Recommended nudge. O5 unified lobby (device-local + greyed-from-cache server games; self-set identity; closes G-step-0). O6 create flows (with-friends online/offline segment + offline dict guard). O7 docs.
Telegram/VK stay online-only (contour review): the offline model is channel-gated via offlineCapable() (native + plain web; false in the mini-apps). offlineMode.active is hard-gated on it, so the whole model (blue chrome, unified/greyed lobby, transport kill switch, device-local vs_ai/hotseat create) stays inert in Telegram/VK regardless of the detected net state (closing a version-lock path that could have flipped them offline); and New Game's with-friends hides the online/offline segment there, leaving the remote invite alone. ARCHITECTURE already declared "Telegram/VK are exempt" - this makes the code match.
Deploy/CI: wire the version gate through the deploy (GATEWAY_MIN_CLIENT_VERSION + GATEWAY_RECOMMENDED_CLIENT_VERSION as plain unprefixed vars via compose + ci.yaml + prod-deploy.yaml + write-prod-env.sh + .env.example) so the gate is live when set (test-contour commit-hash version fails open; empty => dormant). Disable the manual android-build CI workflow for now (rename .disabled). Bump the app bundle budget 127->130 KB for the added always-loaded wiring. Fix the UI Docker stage (gateway/Dockerfile): install --ignore-scripts so the Alpine/musl SPA build no longer tries to native-build sharp (a local android:assets tool, unused in the image); esbuild's binary is an optional dep so Vite still builds.
Tests: gateway go green (two-tier gate over a real HTTP handler); docker build of the landing + gateway targets green locally; compose --no-interpolate confirms the gateway env; two new telegram.spec.ts e2e (offline signal keeps the chrome online + quick-match opponent choice visible => server enqueue; with-friends shows no pass-and-play), RED-verified; svelte-check 0/0, vitest 617, e2e 248 (chromium + webkit), build + bundle-size 127.8/130.
Append the O1-O7 implementation stages to ANDROID_PLAN.md's "Offline-model
redesign" section (exact files, produced/consumed interfaces, acceptance
criteria, targeted tests; executed via stage-implementation, TDD from the pure
netState reducer in O1). Flag it in Progress as the next actionable work
(gated G-step-0), self-contained for a fresh session to resume from.
Owner-approved brainstormed design in ANDROID_PLAN.md, resolving G-step-0:
remove the explicit offline toggle -> a single net-state machine
(online / connecting / offlineNoNetwork / offlineVersionLocked) with explicit
hysteresis + 12 enumerated edge cases; unify the lobby (server games greyed
from cache offline; identity recognises both ids); two-tier version gate (hard
degrades to forced offline with an "Update / Play offline" notice; new soft
recommended tier via an additive X-Update-Recommended header); keep server
vs_ai (offline->local, online->server, app decides by state); TG/VK
always-online. Exhaustive test matrix. Client + a small additive backend/wire
bit; contour-safe.
Add .gitea/workflows/android-build.yaml (workflow_dispatch + confirm=build,
master-only; mirrors prod-deploy): builds the native SPA, bundles the offline
dicts, assembles a release APK as a run artifact. JDK 21 via setup-java; the
Android SDK is host-provisioned (host-executor runner) with a fail-fast verify
that also checks the runner user's access. Signing degrades gracefully — no
keystore => unsigned release APK, not a failure.
Wire ui/android/app/build.gradle: versionCode/versionName from the release tag
(-P props; '=' assignment, not the command form that binds .toInteger() to the
DSL setter's null return), plus a guarded signingConfigs.release from env.
Rename VITE_STORE_URL -> VITE_RUSTORE_URL (empty until publish).
Bake the E as-built into ANDROID_PLAN.md.
Reconcile the Progress/§D status after this session: D is code-complete,
e2e-verified AND on-device-smoke-verified (Pixel_10 / API 37, airplane mode);
the smoke surfaced + fixed the edge-to-edge safe-area bug (top + bottom).
Mark D.5 done (verified by the e2e + the on-device vs_ai turn). Remaining for
D: the reconcile-online leg on-device (hits prod) + the deferred local-game-
visibility decision.
The earlier safe-area fix reached the bottom bars (they apply
--tg-safe-bottom) but not the header: its top inset lived only in a
Telegram-fullscreen-scoped rule, so on the native build the header .bar had
just 5px top padding and its content sat under the status bar (measured: the
title at y=11px behind the 54px bar; the back button untappable). Give .bar a
top inset from the SystemBars plugin's --safe-area-inset-top (native-only via
the plugin var; unset -> 0 on web/PWA/Telegram/VK, so no regression;
tg-fullscreen still overrides via specificity). Verified on-device (Pixel_10 /
API 37): the title moved from y=11 to y=65, clear of the status bar.
targetSdk 36 forces edge-to-edge, so the WebView draws behind the system
bars. On Android WebView < 140, env(safe-area-inset-*) wrongly reports 0, so
the app chrome (which only read env()) drew under the status bar (top nav
untappable) and the gesture-nav home indicator (the game's centre Hint button
intercepted; side buttons fine). Capacitor 8's SystemBars core plugin (built
into @capacitor/core, insetsHandling:'css' by default) injects the correct
--safe-area-inset-* values on every WebView; consume them ahead of env():
--tg-safe-*: var(--safe-area-inset-*, env(safe-area-inset-*, 0px))
Web / PWA / Telegram / VK are unchanged (--safe-area-inset-* is unset there,
so it falls back to env()). Verified on-device (Pixel_10 / API 37) via the
injected var — the emulator's auto-updated WebView 149 hides the env() bug,
so the visual alone won't reproduce it.
The native (Capacitor) sign-in surface is guest + email only: VK ID
web-login is a full-page redirect to id.vk.com that cannot return into the
WebView, and the Telegram Login Widget is unreliable there. Profile now
gates telegramLinkable/vkLinkable on !nativeShell (clientChannel android/
ios), hiding both LINK buttons on native; email and account management —
including an existing link's redirect-free UNLINK — stay. Native tg/vk
login (native SDKs / deep-link OAuth) is a separate later stage.
Covered by e2e/native.spec.ts (the reconcile test opens Profile and asserts
no Link Telegram / Link VK buttons, email present).
Native (Capacitor) cold boot with no cached session now enters as a
device-local guest in auto-offline mode and lands straight in the lobby
(never /login), so the app opens and plays local vs_ai / hotseat with the
APK's bundled dictionaries and zero network. When the gateway becomes
reachable, reconcileServerGuest silently mints + adopts a server guest and
clears the auto-offline, lighting up online features. Web / PWA / Telegram /
VK are byte-for-byte unchanged.
- transport: exec gains { silent, allowOffline } so background reconciliation
bypasses the offline kill switch (like the reachability probe) and never
raises the terminal update overlay (a too-old client stays a local guest);
new authGuestSilent on the client interface, the real transport and the mock.
- app: native no-session boot branch; reconcileServerGuest fired at boot, by
the recovery poll and the online event; the poll routes the session-less
guest through reconciliation, since checkReachable needs a token it lacks.
- native: initNativeShell tolerates a missing Capacitor bridge.
- e2e: new native.spec.ts (inject window.androidBridge; boot -> offline lobby,
local vs_ai move, reconcile -> online, hotseat start) + playwright.config
bundles the dawgs into dist-e2e/dict for the loader's bundled tier.
Make ANDROID_PLAN.md self-contained so a fresh session resumes Stage D from the repo
alone. Records:
- the done foundations (D.1 bundled dicts + loader tier, D.2 local-guest identity) with
their exact modules and the committed checkpoint (bcd5a1d);
- the session decisions (owner-approved): the blocking Login is bypassed on native only,
soft registration reuses the Profile screen, the Telegram/VK link buttons are hidden on
native (VK's web redirect strands the Capacitor app; native tg/vk is a later stage),
local-guest name = localized common.guest;
- the native-gated loader-tier correction (a web `./dict/` fetch would hit the gateway's
own session-gated /dict/ route, so the bundled tier is only tried on native);
- the detailed remaining path — D.3 boot rewrite with the exact blocking-login site
(app.svelte.ts ~989-991), D.4 reconciliation + the silent seam wiring, D.6 Profile
tg/vk gating — plus the offline-first e2e strategy (simulate native by injecting
window.Capacitor) and its initNativeShell/@capacitor/app gotcha.
The additive groundwork for the native offline-first experience (ANDROID_PLAN.md §D).
Inert until the native cold-boot lands: on the web these paths never fire, so web / VK /
Telegram behaviour is unchanged.
- ui/scripts/bundle-dicts.mjs copies the scrabble-dictionary release DAWGs into
dist/dict/<variant>@<version>.dawg for the native pipeline (run after `pnpm build`,
before `cap sync`).
- The dict loader gains a bundled tier between the IndexedDB and network tiers, attempted
only on a native channel (a packaged app serves ./dict/<key>.dawg from its assets).
Native-gated rather than the plan's "404 on the web" because the relative path would
otherwise hit the gateway's own session-gated /dict/ route.
- __DICT_VERSION__ Vite define (from VITE_DICT_VERSION, default "dev") + its ambient
declaration; the offline vs_ai / hotseat creates in NewGame fall back to it when a
device-local guest has no profile-advertised version.
- New lib/localguest.ts: a persisted device-local guest id (no DB row); the offline vs_ai
human seat uses it (and the localized common.guest name) when there is no server session.
svelte-check clean, vitest green, native + web builds clean. The cold-boot rewrite,
reconciliation, the Profile soft-sign-in gating and the offline-first e2e follow.
Introduce a minimum-supported-client gate so a future incompatible wire change
can turn away installed builds too old to speak it, cleanly, instead of letting
them crash on decode. It rides the outermost stable layer (an HTTP header), never
the FlatBuffers payload.
Gateway:
- New internal/clientver: dependency-free parse + compare of the leading
MAJOR.MINOR.PATCH (a git-describe suffix is tolerated).
- GATEWAY_MIN_CLIENT_VERSION config (empty => gate dormant; validated at load).
- connectsrv checks the X-Client-Version header before decoding the payload:
Execute returns result_code="update_required" (before the registry lookup),
Subscribe returns FailedPrecondition. It fails open on an absent or garbled
header — the header is a client-controlled compatibility signal, not an access
control.
Client:
- Attach X-Client-Version on every call.
- A terminal update.svelte.ts store + a non-dismissable UpdateOverlay (native
opens VITE_STORE_URL, web reloads); retry.ts maps FailedPrecondition to the
update_required sentinel; a mock __update hook drives the e2e.
Wire-additive and contour-safe: no FBS/proto regen, no schema migration; the gate
stays dormant until GATEWAY_MIN_CLIENT_VERSION is deliberately set, so web / VK /
Telegram behaviour is unchanged. The silent reconciliation seam is deferred to the
offline-first work (its only caller). Tests: Go clientver/config/connectsrv gate
tests, retry.test.ts, Playwright update.spec.ts.
The three "plain desktop browser" export tests (the PNG and GCG downloads plus the
legacy-Telegram clipboard copy) assumed navigator.share is absent — true for Chromium
and for WebKit on the Linux CI host, but desktop WebKit on macOS exposes a working Web
Share API. There shareUrlAsFile / pickGcgDelivery take the share branch, so no download
event fires and the "GCG copied to the clipboard" toast never shows, and the tests
failed only on macOS WebKit.
Pin navigator.share/canShare off in those three tests (a withoutWebShare helper that
mirrors the inverse stub the share-sheet test already uses), so the delivery path under
test is deterministic and identical across engines and OSes. Test-only; no production
change. Full e2e suite: 228 passed on Chromium + WebKit.
Make the web SPA behave correctly inside the Capacitor native shell, where the
bundle loads from a local origin:
- New lib/origin.ts gatewayOrigin(): absolute URLs resolve to VITE_GATEWAY_URL on
native (the finished-game export/share URL in Game.svelte and the Wallet
site-root link), falling back to the page origin on web. transport.ts already
resolved via VITE_GATEWAY_URL and is left as-is.
- Skip the PWA service worker on the native channel (the assets are already local;
a worker would risk serving stale content across store updates).
- Hide the money-purchase UI in the MVP: new distribution.purchasesHidden() folds
VITE_PAYMENTS_DISABLED and the Google Play flag. The Wallet "buy" tab shows a
neutral, pointer-free note (wallet.purchasesSoon) for the RuStore MVP and keeps
the RuStore stub Google-Play-only. A ?nopay mock force mirrors ?gp for the e2e.
- Native env types in vite-env.d.ts.
Client-only, contour-safe: no wire/proto/schema change; web/VK/Telegram unchanged.
Tests: origin + purchasesHidden unit tests, a ?nopay wallet e2e. svelte-check
clean, vitest green, web + native vite build clean.
Generate the Android launcher/adaptive icons + splashes with capacitor-assets from ui/assets/icon.png (the existing maskable brand mark upscaled 512→1024) as a placeholder until the mandatory single-vector icon rebrand recorded in ANDROID_PLAN.md. The adaptive icon insets the foreground 16.7% into the safe zone; the AndroidManifest is untouched (its icon refs already point at @mipmap).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add @capacitor/{core,android,app,cli,assets} to ui/, a capacitor.config.ts (appId ru.eruditgame.app, appName Эрудит, webDir dist — bundle model, no server.url), and the generated ui/android/ Gradle project (tracked; build outputs and the machine-specific local.properties gitignored, keystore patterns un-commented so signing material can never be committed). Wire the Android hardware Back button in ui/src/lib/native.ts behind a dynamic @capacitor/app import (web/mock bundles never load it), called from App.svelte onMount and reusing routeDepth for the navigation-root check. Whitelist sharp in pnpm-workspace.yaml for @capacitor/assets icon generation.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Correct ANDROID_PLAN.md: bundled offline dictionaries come from the scrabble-dictionary release (DICT_VERSION), not scrabble-solver; pin the toolchain to Capacitor 8 (compileSdk/targetSdk 36, minSdk 24, JDK 21 — @capacitor/android compiles at Java 21). Add a Progress section marking the scaffolding milestone done, and capture the native-Android build recipe + toolchain gotchas in .claude/CLAUDE.md so a fresh session resumes without re-deriving them.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Refuse a client whose IP is in a curated CIDR feed with 403 in the same
abuseGuard, before the fail2ban ban. Prod-only (keys by real client IP), off by
default.
- ratelimit.Blocklist: a sorted-range IPv4 matcher (binary search) with an
allowlist checked first; ParseDROP reads the feed; ApplyRefresh keeps the
last-good feed on a transient fetch failure and drops it fail-open once stale
(better to under-block than block a legitimate client on a frozen feed). A
separate static CIDR set, not the per-IP fail2ban store.
- gateway: a refresher goroutine re-fetches every few hours (bounded fetch + size
cap); config GATEWAY_BLOCKLIST_{ENABLED,URL,ALLOW,REFRESH,MAX_STALENESS}.
IPv6 is not matched (a v6 client is still covered by fail2ban / the honeypot).
- observability: gateway_blocklist_blocked_total + entries/age gauges; a Grafana
alert warns before the feed is dropped; service-overview panels.
- deploy: compose env + write-prod-env.sh + prod-deploy/rollback wiring (opt-in
via PROD_ vars).
- docs (ARCHITECTURE, deploy/README); unit tests (match, allowlist, IPv6 skip,
parser, fault-tolerance) + a 403 abuse-guard integration test.
Sales (chip packs) first, ascending by rouble price; then the chip-exchange
values, grouped and price-sorted the same way projectOfferPricing lists them, so
the /catalog console mirrors what a buyer sees. Internal cosmetics only — no
product behaviour change. The value-group order moves to a shared helper
(valueGroup) so the offer and the admin list cannot drift.
The bot runs on its own host and exports no telemetry (otelcol is unreachable
from there), so it was a monitoring blind spot. Observe its Bot API health
centrally by wrapping the HTTP client — one place, no per-call-site
instrumentation — and relay it up the existing bot-link as a periodic Health
message the gateway turns into its own metrics.
- proto: add Health (delta connect / api / 429 counters + a last-ok stamp) to
the FromBot oneof (additive, backward-compatible).
- bot: platform/telegram/internal/health wraps the Bot API HTTP client — it
stamps liveness on any 2xx (so the getUpdates long-poll keeps it fresh even
when idle), classifies transport/5xx failures (getUpdates vs other) and 429s,
and honours a 429's Retry-After (bounded) so the bot backs off; a 4xx other
than 429 is a normal per-request outcome and is not counted.
- bot-link client: flush the reporter as a Health message every 30s over a
single-sender loop (a gRPC stream forbids concurrent Send).
- gateway: fold each report into bot_tg_errors_total{kind} and the
bot_tg_last_ok_unix liveness gauge.
- grafana: alerts (bot disconnected, bot not reaching the Bot API, sustained
429s) routed to the operator email — which does not go through the bot — plus
a Telegram-bot dashboard.
- docs (ARCHITECTURE, compose comments); unit tests (observer classification,
Retry-After, snapshot/commit, hub last-ok monotonicity).
- packs sorted by ascending rouble price;
- values grouped hints-only -> no-ads-only -> no-ads+hints -> tournament
(the tournament group is empty until such products become sellable),
ascending chip price within each group;
- price columns right-aligned (GFM "---:" separators);
- tables span the full content width; the name column shrinks to its content
and never wraps;
- muted-but-visible cell borders on the dark theme, where the section rule
colour blends into the background.
The offer price list projects the admin-entered product title into a markdown
table cell that is rendered, unsanitised, into the public /offer/ page. Escape
the title at the projection boundary so it renders as literal text: HTML
metacharacters become entities and the markdown table pipe and link brackets are
escaped, so a title can neither inject markup nor form a javascript: link. The
committed offer prose keeps its trusted-content treatment (code-reviewed, not
runtime input).
Robokassa moderation requires the public offer to list every digital good with
its price. Move /offer/ off the static landing container to the render sidecar:
it splices the live catalog price list (§4.4) into the owner-edited
ui/legal/offer_ru.md and renders it with the shared ui/src/lib/offer.ts — one
renderer, no drift, always matching the current catalog with no redeploy.
- backend: /api/v1/internal/offer/pricing (internal, off the edge allow-list)
projects the active catalog into two markdown tables — chip packs priced per
rail (roubles / VK votes / Telegram Stars) and chip-priced values — through
payments.Money so no float reaches the page. Cached in memory: warmed at boot,
marked stale on every catalog mutation, so a served render issues no query.
- renderer: GET /offer/ fetches the tables and substitutes them at the
<#pricing_template#> marker, then renders; offer_ru.md is baked into the image
and marked is bundled from ui. GET /offer -> 301. Only /offer/ is edge-exposed.
- caddy: route /offer/ to the sidecar; drop the now-dead landing /offer/
handlers and the vite emit-offer plugin.
- offer: fill §4.3 (the chip-payment wording) and drop the in-page back link.
- landing footer: a feedback link (the offer's Telegram contact) beside the
offer link.
- docs (ARCHITECTURE, FUNCTIONAL +_ru, renderer README), CI /offer/ probe,
unit + integration + node tests.
Reworks the Wallet screen to be more compact and to separate the two flows, and fixes a
reported bug.
Bug: buying a value with too few chips showed the generic "Something went wrong" toast — the
backend's `insufficient_chips` (409) code was propagated correctly but had no i18n mapping, so
`errorKey` fell back to `error.generic`. Add `error.insufficient_chips` / `error.product_not_found`.
Also disable a value's Exchange button up front when the spendable balance cannot cover it (the
server stays authoritative).
Redesign:
- a compact **balance** row leads the screen — the context's own chips (🪙 N) then any linked
other-platform chips behind that platform's logo (monospaced digits); inside a VK/TG store only
that store's own segment shows;
- the benefits section is renamed **Active**, moved above the store, shows one inline line
(hints + ad-free) and hides when nothing is active;
- the **store** splits by a two-way toggle into **Buy chips** (money packs + the watch-an-ad row at
the top) and **Spend chips** (values, exchanged for chips);
- a value's action reads **Exchange** and opens a single confirmation dialog (chip spends are
instant, with no provider window to confirm them); the existing cross-platform store-compliance
warning folds into that same dialog when the spend would draw VK/TG chips.
No payments-model or wire change; the money→chips→values model is unchanged (verified: money buys
only chip packs, values are chips-only — "no-ads for money" is structurally impossible).
Tests: wallet e2e updated for the new layout (chromium + webkit green). Docs: FUNCTIONAL (+_ru)
wallet story rewritten. App bundle budget 126 → 127 KB (always-loaded settings hub).
The "WAL archiving stalled" alert fired on prod during a quiet period. An idle Postgres
archives nothing — it never force-switches an empty WAL segment on archive_timeout — so
pg_stat_archiver_last_archive_age grows past 30 min even though archiving is perfectly
healthy (failed_count 0, no .ready segments, pg_wal flat, backups current). A false
positive: no data and no disk at risk (nothing was written, so the frozen recovery point
equals the live state).
Gate the age condition on pg_wal actually growing:
(pg_stat_archiver_last_archive_age > 1800) and on() (delta(pg_wal_size_bytes[35m]) > 16MB)
so it fires only when WAL is being produced but not archived (the real "pg_wal fills the
disk" danger; genuine archive_command failures are already caught by pg_archive_failing).
`and on()` bridges the two metrics' differing label sets (last_archive_age carries a server
label, the pg_wal gauge does not); delta() (not increase) suits the gauge. pg_stat_wal is
not exported by this postgres_exporter, so pg_wal size growth is the available signal.
Validated on prod with promtool: the expression is empty while idle.
Also fix the runbook command in both archive alerts' annotations: `pgbackrest check` needs
`--pg1-user=scrabble`, or it connects as role "root" (which does not exist) and aborts with
"no database found".
Board-highlight bug (reported on the contour): formedGeometry walked cross words
unconditionally, so in a single-word (one-word-per-turn) game a staged tile sitting
next to a committed tile lit up a green "cross word" the engine ignores — and which
need not even be a real word (the reported "БО lights up green in a one-word ПОПА
play"). Gate the cross-word walk on the game's multipleWordsPerTurn flag. The score
(8) was already correct — a premium square under the main word.
Also, from review:
- the turn strip reads the staged play's "WORD+WORD = N" while composing a legal move,
reverting to the turn / result text otherwise;
- the Exchange/Pass dialog shows the bag count ("In the bag: N" / "Bag is empty")
right-aligned in the title row, via a new optional Modal `titleAside`;
- cosmetics: half the turn strip's bottom padding (the plaques below carry their own
top pad); a top gap above the landscape rack (it sat flush under the docked history);
more horizontal padding on tab count badges so a 2-3 digit bag count clears the pill
ends;
- admin console: the game Summary now shows the single-word / multiple-words rule.
Tests: formed single-word case added; full unit (584) + e2e (chromium + webkit, 113
each) green; backend build + adminconsole templates parse. Docs (FUNCTIONAL +_ru,
UI_DESIGN) updated.
Remove the under-board status strip and relocate its signals:
- bag count -> a badge on the exchange/pass control + the foot of the move table
- whose-turn / win-lose -> a thin strip above the score plaques
- the tentative-move caption -> the board itself: staged tiles tint green (legal) or
pink (illegal), the board tiles a formed word runs through go a shade darker, and an
orange score badge sits on the main word (digit sized like a tile value, clamped on-board)
The word geometry (covered cells + badge anchor) is a new pure client-side helper
(ui/src/lib/formed.ts), independent of the move evaluator, so it works on the local or
network preview path alike; the badge's number still comes from the preview score.
Rack: a seven-column grid filling the tray width in both layouts — square, full-width
tiles — with the confirm control in the fixed 7th slot.
Settings: a touch-only "Zoom the board" toggle (default on, device-local) gates the
tile-placement auto-zoom; taking a hint while zoomed in now zooms out so the highlighted
hint word is never left off-screen.
Docs (FUNCTIONAL +_ru, UI_DESIGN, ARCHITECTURE) and e2e/unit tests updated.
The Ansible base-directory loop created /opt/scrabble/certs at mode 0750
(deploy:deploy), like config/dumps/images. The gateway (and backend) run as
the distroless nonroot UID 65532 — not the deploy user — so the container
cannot traverse a 0750 certs dir and fails at startup with
"mtls: load server keypair: ... permission denied", crash-looping.
This is latent: a long-running container holds the keypair in memory and
never re-reads the file, so the misconfig only bites when a container
restarts (a host reboot / redeploy). A hoster maintenance reboot exposed it
on prod — the gateway came back crash-looping while the deploy could not SSH
in mid-reboot.
Split certs out of the 0750 loop and create it 0755 (traversable). The keys
stay 0644 by design (the gateway compose relies on it); the host is
single-tenant + SSH-access-controlled, so a traversable certs dir adds no
meaningful exposure. The live prod host was already chmod-fixed by hand; this
keeps the next provisioning run from re-tightening it.
The Robokassa credentials reached only the test contour (ci.yaml → TEST_
secrets). The prod-deploy / prod-rollback workflows and write-prod-env.sh
never rendered BACKEND_ROBOKASSA_*, so on prod the shop login was empty and
the direct RUB rail stayed disabled — the PROD_BACKEND_ROBOKASSA_* secrets
went nowhere.
Export the shop login + Password1/Password2 (secrets) and a
BACKEND_ROBOKASSA_TEST flag (a variable, so go-live is a flag flip not a
secret rotation) from both prod workflows, and emit the four ROBOKASSA_*
vars from write-prod-env.sh (the shared deploy/rollback env renderer) so a
rollback keeps the rail up. The compose already maps ROBOKASSA_* →
BACKEND_ROBOKASSA_*. Document the new secrets/variable in the deploy README
and .env.example. Password3 (Robokassa's JWT-invoice API) is unused.
Address review of the active-game limit lock:
- Live-event GameViews (opponent_moved, match_found, game_started, ...) did
not carry game_kind, so a lobby patch from an event zeroed a game's kind
and the client's per-kind count under-counted — a capped kind read as
free (a disabled start instead of the lock, and a wrong per-kind result).
Thread Kind through notify.GameSummary → the event GameView.
- New Game screen refreshes the lobby games on mount so the per-kind count
reflects the current set, not a stale cached snapshot.
- The guest funnel's login button routes to the profile screen (the account
controls), not settings.
- Copy: the guest prompt is "sign in to use all the game's features"; the
durable notice is "finish your active games to start a new one".
Regression tests: the notify opponent-moved payload carries kind; the client
lock locks only the reached kind (vs_ai at cap leaves random open).
Carry the caller's per-tier active-game caps and each game's kind on the
wire (Profile.game_limits + GameView.kind, additive FBS + gateway transcode
+ client codec, committed regen). The New Game screen counts the player's
active games per kind from the lobby and locks a capped start: an outline
button with a lock that opens a funnel modal instead of a game -- a sign-in
prompt for a guest, a "finish a current game first" notice for a signed-in
account (native Telegram popup, in-app modal elsewhere). The lock lifts via
the existing profile refetch after a guest->durable upgrade.
Remove the lobby's old at_game_limit New-Game tab disable + notice: the flag
(now the random-kind cap) conflicted with the per-kind lock -- it hid the
screen where the lock lives and wrongly blocked an unfulfilled kind. The New
Game tab is always enabled; the per-kind start lock is the only gate. The
at_game_limit wire field stays (unused by the client) for a later cleanup.
Tests: client lock logic + codec kind/game_limits roundtrip + gateway
transcode encode + native popup builders (unit); a mock e2e for the lock
badge and the modal.
Cap a player's simultaneous unfinished games per kind (vs_ai, random,
friends) with independent guest and durable-account tiers, held in a new
single-row backend.config table (-1 = unlimited) behind an in-memory cache
and editable live in the admin console (/_gm/limits). Each game is tagged
with games.game_kind on creation.
This replaces the earlier flat MaxActiveQuickGames=10 combined cap: the
per-tier/kind config is the single mechanism, enforced at the same handler
gate (ensureUnderGameLimit by kind on lobby/enqueue) plus the durable
friends cap in CreateInvitation. game.Service.AtGameLimit only resolves the
tier and counts; the limit policy stays at the request edge.
Guests are now refused friend requests, friend-code redemption,
befriend-in-game and invitation creation outright (403 guest_forbidden) --
previously only the UI hid these.
Admin: a kind column in both game lists and the config editor.
Defaults: guest 1 vs_ai / 1 random / 0 friends; durable 10 / 10 / 10.
Closes the admin / reports / catalog work. Each fund row on the /_gm finance
panel gains a Refund action (payments.RefundOrderFull): a full-order refund the
operator records after refunding on the rail — a refund ledger row + a floor-0
chip revoke (never negative, D27), idempotent (a second refund reports
already-refunded). A ledger CSV export (/_gm/ledger.csv, payments.LedgerExport)
streams the whole append-only ledger for tax + reconciliation.
Tests: refund an order in full (chips revoked, a refund row), an idempotent
second refund, the CSV export shape; CSRF-guarded.
The in-game hint badge ignored the payments hint wallet: loading a game clobbered
app.profile.hintBalance with the deprecated StateView.wallet_balance (zeroed in the
D31 domain removal but left in the protocol as a dead 0, then synced over the real
balance at game load).
Finish the removal honestly. StateView carries the per-game allowance alone
(hints_remaining); the purchasable wallet lives solely on the profile and the client
adds it (lib/hints). Removed StateView.wallet_balance across every layer — backend
StateView/DTO, notify.PlayerState (live events), gateway StateResp + wire.StateView,
the client model/codec/mock/localgame — and dropped the now-unused wallet arg from
hintsRemaining. The FBS field is tombstoned `(deprecated)` (not deleted) so the vtable
slots after it stay stable across a rolling deploy; no accessor is generated. HintResult
keeps wallet_balance (the real post-spend payments balance the client adopts into the
profile). The StateView type no longer has walletBalance, so the clobber cannot return
without a compile error.
Tests: hintsRemaining (2-arg); hints.hintsLeft (allowance + live wallet, no strip); the
game state/hint integration (allowance-only HintsRemaining); gateway transcode; FBS regen.
The /_gm user card gains a Grant panel: grant raw benefit atoms (hints /
no-ads days / forever) or a defined value product (a reward bundle, including
an archived one), origin-picked. Both write an admin_grant ledger row via
payments.Grant / GrantProduct; the by-product grant records the source
product_id + snapshot. Both refuse a chips atom (never grant currency) or a
tournament atom (no credit target yet); chips/tournament products are also
kept out of the by-product picker.
Tests: the console grant end to end (raw, by-product, refuse a chips pack,
CSRF-guarded).
The /_gm console gains a Catalog editor — the source of truth for products.
Create / edit / archive-unarchive products, their atom composition and per-rail
prices (RUB via direct / VOTE via vk / XTR via telegram / CHIP value), and
hard-delete only a never-transacted product (an order or ledger reference forces
archive-only, backed by the FK RESTRICT). The archived flag reuses the existing
product.active. Activation revalidates the sellable shape — a pack (the chips
atom ⇒ a money price per rail, chips-only) or a value (no chips ⇒ a CHIP price);
a tournament-bearing product is composable but never sellable yet.
Backed by payments AdminCatalog / CreateProduct / UpdateProduct /
SetProductActive / DeleteProduct + a pure validateProduct.
Tests: validateProduct (pack / value / tournament / duplicate / shape); the
console editor end to end (create, edit, archive, delete-if-clean, refuse a
transacted delete).
The /_gm user card gains a Finance panel: an account's chip balances per
funding segment, benefits per origin (hints, no-ads until/forever), the
recorded refund risk (abuse flag + floor-0 loss), and the append-only ledger
history newest-first. Backed by a new payments.AccountStatement read straight
from the materialized tables + the ledger (uncached — an admin, rare view).
Folds the agreed admin / reports / catalog plan into PLAN.md (the PR stack:
this panel, then the catalog editor, admin grant, refund + ledger export) and
moves the tournament-entry storage design to the tournament stage.
Cooldowns were tracked per kind ({move,hint}), so a hint ad and a move ad
did not see each other: after a hinted move's ad the move timer was still
zero, and the next plain move fired an ad immediately — two ads within a
minute, under the cooldown. Track a single shared last-shown time; the kind
only selects the required gap (hint 1m, move 5m, vs_ai 30m) measured from the
last interstitial of any kind. A hint can still fire on its shorter gap
(D30's "independent" hint cooldown), but never stacks a second ad within a
cooldown.
Tests: a hint uses the shorter shared gap; a move does not stack onto a
just-shown hint ad.
Taking a hint fired the post-move interstitial immediately, when the hint's
preview tiles landed — interrupting the turn and reverting the board to the
rack on the ad's close while the hint stayed spent. Move the trigger to the
move confirmation: a hint applied this turn marks it (hintUsedThisTurn), and
the confirmed play fires the hint-kind interstitial (its own cooldown)
instead of the plain move one; the marker clears on any turn boundary
(applyMoveResult) so a hint-then-pass does not leak into the next move.
e2e: taking a hint fires no ad; confirming a move does.
Interstitial video after a confirmed play or a hint, VK-only, offline
banner-only. The gate is client-mirrored: the backend puts the config
cooldowns (global 5m / vs_ai 30m / hint 1m) and a suppressed flag (the
no-ads / no_banner gate, same as the banner) on Profile.ads via adsFor;
the client self-gates on a per-kind last-shown time in localStorage, with
the VITE_ADS_STUB contour "ad fired" toast. Never fires after a pass,
exchange or resign. maybeShowInterstitial + vkShowInterstitial; the codec
and gateway transcode carry the ads block.
Also retires the deprecated accounts.hint_balance / paid_account domain
usage (expand-contract, code only — the columns stay for a later DROP so
image rollback stays DB-safe): drop the Account fields and their scan, the
dead account.SpendHint, account.GrantHints and the admin grant-hints
action; the in-game hint display now comes wholly from the payments hint
benefit. Banner eligibility and account merge no longer read the legacy
flags.
Tests: ads.test.ts (the client-mirrored gate), the codec ads round-trip,
the gateway ads transcode test, and a profile ads-config integration test
(cooldowns + suppressed under no-ads / no_banner). Docs: PAYMENTS §10 (+ru)
interstitial, the decision amends, backend README, the plan.
Testing the rewarded slice surfaced a contradiction: rewarded ads let a VK-iOS
user earn vk chips, but the blanket VK-iOS "spend freeze" then blocked spending
them (the wallet showed "5 (view-only)"). Apple's ToS forbids only BUYING in-app
values on VK-iOS (money -> chips), not spending or earning them — so the freeze is
corrected to purchase-only: vkFrozen() now gates only CreateOrder (the money-in
step), not spendableSources (spending). VK-wallet chips — earned via rewarded ads
or bought on the same account elsewhere (VK Android) — now spend on VK-iOS too.
(Owner's ToS finding.)
Also: the temporary contour diagnostic confirmed VK returns only {result:true} for
a rewarded view (no token/signature) — client-attested is final, no hardening
possible — so the diagnostic (the log and the diag wire field) is removed.
Docs: PAYMENTS(+ru) VK-iOS freeze + D17 amend + PLAN E6. Tests updated (gate /
wallet VK-iOS now spendable).
The first ads slice: a voluntary rewarded video credits chips. VK Mini App ads
(VKWebAppShowNativeAds) expose only a client-side watch result — no
server-to-server verify — so the credit is client-attested, guarded by a server
daily + hourly cap (config reward_daily_cap / reward_hourly_cap, default 50 / 10).
The caps are both anti-abuse (bounding a forger who skips the ad and calls the
endpoint directly) and an economic conversion lever (limiting free chips so a
player who wants more buys). D29 amended to VK's reality.
Backend: CreditReward (VK-only, order-less, idempotent on a client nonce, floored
by the caps; payout from config rewarded_payout_chips, default 0 = off) + the
wallet.reward edge op returning the updated wallet (reward_chips gates the
"watch for chips" CTA). Additive migration (two config columns).
Client: the ads-network abstraction (lib/ads.ts, VK impl) + the VK bridge
(vkRewardedReady / vkShowRewarded) + the Wallet CTA + i18n. A contour test stub
(VITE_ADS_STUB -> a toast instead of a real ad; prod always real) and a temporary
diagnostic that logs the raw VK data, so we confirm on the contour exactly what
VK returns (harden to signature-verify if it carries one).
Tests: backend integration (credit, nonce idempotency, hourly cap, disabled,
non-VK refusal) + codec unit (reward wire). Docs: PAYMENTS(+ru) §10, D29 amend,
PLAN E6. Bundle: shared budget 30->31 (reward i18n strings).
The last money-intake slice: reverse a paid order best-effort, exactly once. All
refunds are admin-triggered (E7) — no rail pushes an unsolicited refund (Robokassa
via its refund API / cabinet, VK via support, Telegram via refundStarPayment), so
this ships the engine they all converge on, not a webhook.
The Refund method matches the paid order, appends a refund ledger row (idempotent on
(provider, provider_refund_id) — distinct from the fund's payment id, so both
coexist), and revokes the funded chips floored at 0 (never negative — D27,
balances_chips_chk). When the chips were already spent, the unrecoverable remainder
is recorded as a per-account loss + abuse flag in the new additive
payments.account_risk table (read by the E7 report). The refund ledger row's chip
delta is what was actually reclaimed (the ledger stays reconcilable); the full
reversal rides in the snapshot; the order stays paid.
Additive migration (a new table only) -> rollback-safe, no contour wipe. Robokassa
refund-status polling is deferred (a worker not worth it at low chargeback volume);
failed events are not wired (no rail signals a hard post-charge server decline).
Tests: integration (full revoke; revoke-after-spend = floor-0 + loss + abuse;
duplicate idempotent; unpaid-order guard). Docs: PAYMENTS(+ru) §9, PLAN (E5 -> DONE).
The wallet order flow plus the Telegram Stars openInvoice / VK order-box launch
and the per-button in-flight state ride the always-loaded Wallet screen, which
crossed the 123 KB gzip ceiling. Bump to 125 (matching the established
feature-driven budget history in the comment) with ~2 KB of headroom.
Tapping one pack's Buy dimmed every buy button at once (the shared busy
flag drove disabled + the :disabled opacity on all of them), so it looked
like both packs reacted to one tap. Track the in-flight product id
(busyId): the concurrency guard stays global, but the pressed/dimmed look
is now scoped to the tapped button. Payments were never affected.
The contour VPN sidecar's netns resolv.conf was pinned to 1.1.1.1 by a DNS=
directive in TEST_AWG_CONF, so internal names (gateway, otelcol) did not resolve
and the bot-link had been down since ~2026-06-29. The directive is removed; this
empty commit re-rolls the contour so the vpn recreates with Docker's 127.0.0.11
resolver and the bot-link (and the Stars rail over it) comes up.
Accept real money via Telegram Stars (XTR) — the third intake rail
alongside Robokassa (direct) and VK Votes.
Only the bot reaches Telegram, so the rail funnels through the reverse
mTLS bot-link:
- the gateway mints the invoice on a CreateInvoice command (the bot
calls createInvoiceLink, XTR; the link goes to WebApp.openInvoice);
- the bot gates each pre_checkout_query via a ValidatePreCheckout unary
(the order must exist, be still creditable and not already paid — the
reusable-invoice double-pay guard; the decline reason is localised to
the order account's language);
- a completed successful_payment is queued in a durable pure-Go SQLite
outbox and forwarded via a ForwardPayment unary, credited once
(idempotent on telegram_payment_charge_id, honours an expired order),
re-driven on restart and every 30s.
The rail is wired by TELEGRAM_STARS_OUTBOX_DIR (default /data) but stays
inert until a chip pack carries an XTR price, so seeding a Stars price in
the admin is the go-live.
Tests: backend integration (order->forward->credit once, duplicate,
pre_checkout gate) + bot outbox unit (idempotent, restart re-drive) +
executor createInvoice. Docs: PAYMENTS(+ru) §9, ARCHITECTURE, the
platform/telegram README, PLAN.
Replace the hardcoded erudit-game.ru host with location.origin, so the iOS
store note's "other version" link points at this deployment's own site root
(prod or the contour) without a hardcoded host or build-arg wiring.
The VK-iOS detection used platformSubtype(), which is server-derived for VK and
so never reported iOS on the client — the pack CTA stayed active and a tap hit a
403 with a generic error. Read the device family from the signed vk_platform
launch param (normalizeSubtype(vkPlatform())) instead. Under the store header on
VK-iOS, add a note that purchases are prohibited by Apple policy, linking to the
web version (opened through VK's external-link route).
Money purchases are not permitted in the VK iOS app (Apple ToS), and the backend
already refuses them (the CreateOrder VK-iOS freeze). Show the pack "Buy" muted
there and explain on tap ("purchases are not available on this platform"),
instead of letting the tap hit a 403 and a generic error toast. The storefront
still lists the packs.
Wire the VK Mini Apps ("голоса") rail end to end, reusing the intake engine. The
wallet.order endpoint branches by rail: a VK context opens a pending order
(provider vk) and returns its id, which the client passes to VKWebAppShowOrderBox.
VK's two-phase payment callback is verified at the gateway with the app protected
key (GATEWAY_VK_APP_SECRET) and proxied to a backend intake handler: get_item
returns the ordered pack's title and vote price; a chargeable order_status_change
credits the vk segment exactly once (the same Fund, idempotent on VK's own order
id) and records a succeeded event, so the dispatcher push refreshes the wallet.
Integration test for the VK order->credit path.
Add the vkpay package: verify a VK Mini Apps payment ("голоса") callback
signature — MD5 (mandated by VK's payment protocol) of the sig-excluded
parameters, sorted by name and concatenated key=value, with the app secret
appended; case-insensitive over the hex digest. Unit-tested (valid / tampered /
wrong-secret / missing). First piece of the VK rail; the two-phase callback
handler, the VK order branch and the client bridge follow.
Deliver payment_events to connected clients as an in-app wallet-refresh push: a
background dispatcher drains undispatched events and publishes a KindNotification
"payment" signal, marking each delivered; the client bumps a wallet-refresh
counter the open Wallet screen watches, re-fetching in place. A return-focus
refetch is the fallback. The Robokassa Success/Fail return now serves a
self-closing page (the payment opens in a separate window) so the customer drops
back into the live app instead of a cold start. Integration test for the event
drain/mark queue.
Record the E5 delivery and resolved decisions in PLAN.md (Shp_order matching,
order-id idempotency, cabinet-side receipt, never-negative → schema-free) and
mark the stage WIP. Add a CI probe asserting /pay/robokassa/result reaches the
gateway rather than the landing catch-all.
Map the Robokassa merchant login + Password1/Password2 into the backend
container env from the deploy secrets (TEST_BACKEND_ROBOKASSA_*), and force
IsTest on the test contour so it can never take real money. An empty login
leaves the direct order + callback endpoints unregistered.
Replace the disabled "Soon" pack action with a real purchase: the Wallet opens a
money order (wallet.order) and sends the player to the provider's hosted-payment
page (window.open via openExternalUrl); the chips are credited later by the
verified server callback. Add a public-offer link under the packs (paying
accepts the offer). Codec order round-trip unit test + a mock-e2e purchase test;
the Google Play stub and the chip-spend paths are unchanged.
Add the public /pay/robokassa/result callback proxy (rate-limited; forwards the
provider's form parameters to the backend intake, the single writer, and echoes
its "OK<InvId>" back to Robokassa) and the /pay/robokassa/{success,fail}
browser-return redirects into the app. Route /pay/* to the gateway in the
contour Caddyfile so the callback reaches the edge, not the landing catch-all.
The backend intake now returns the echo body as JSON for the gateway to relay.
Add the WalletOrderRequest / WalletOrderResponse FlatBuffers messages and the
wallet.order Connect op: the gateway decodes the order request, forwards it to
the backend POST /wallet/order and returns the created order id plus the
provider launch URL the client opens. Regenerate the committed Go and TS
FlatBuffers code.
Cover the order→callback→credit path over Postgres: a funded order credits its
segment exactly once; a replayed callback for the same order credits nothing
(the ledger idempotency index holds); a mismatched paid amount is refused with
no write; an expired pending order is still honoured by a late valid callback.
Wire the Robokassa direct rail into the backend transport. POST
/api/v1/user/wallet/order (walletGate + a D36 confirmed-email gate for the
direct rail) opens a pending order and returns the signed Robokassa payment
URL. The internal, gateway-only /payments/robokassa/result endpoint verifies
the Result signature, credits the matched order exactly once via Fund (honoured
even if expired), records a succeeded payment event, and answers Robokassa's
"OK<InvId>". Add the Robokassa env config, an account HasConfirmedEmail check
(D36), the payment_events writer, and a periodic pending-order reaper. The
routes register only when a Robokassa merchant login is configured.
Add the payment-intake write path (provider-agnostic) and the Robokassa
direct-rail glue, both unit-tested; transport, wire and UI follow.
- payments: extend the ledger insert to thread order_id/provider/
provider_payment_id (spend/grant pass nil); add the order store
(create/read/expire + a pack-price loader) and the fund credit — a
fund ledger row + a guarded balance upsert + mark-paid in one tx,
idempotent on the (provider, provider_payment_id) unique index, cache
invalidated after commit. A valid callback is honoured even on an
expired order. Service CreateOrder/Fund/ExpireOrders; Money.Major for
the provider amount field.
- robokassa: build the signed hosted-payment URL (SHA-256, order id via
Shp_order, InvId unused) and verify the Result callback signature
(Password2), extracting the order and amount. Receipt/fiscalisation is
configured shop-side, so no Receipt parameter is sent.
The offer is the legal document a purchase accepts, needed before real-money
intake goes live. Convert the source PDF to an editable ui/legal/offer_ru.md and
render it to a standalone static dist/offer/index.html at build (a vite
emit-offer plugin using marked); the landing container serves it at /offer/,
with a bare /offer redirecting in. Add a small centered "Публичная оферта"
footer link on the landing (ru/en) and a CI probe asserting /offer/ serves the
page rather than silently falling through to the landing shell.
docker exec defaults to root, so the systemd base-backup timer connected to the
database as role "root" (then "postgres") — neither exists; the superuser role is
POSTGRES_USER (scrabble). Run the timer's pgBackRest as the postgres OS user (-u postgres,
for lock-dir/PGDATA consistency with archive-push) and connect with --pg1-user=scrabble.
archive_command (run by the postgres server process) was already correct.
Also record the point-in-time-recovery arming + restore drill in deploy/README.md
(correct the runbook commands to the same invocation, fill the drill log) and mark the
durability work done in PLAN.md.
The S3 endpoint is a host only; some providers publish a separate, non-443
port. Add an optional PGBACKREST_S3_PORT (default 443) alongside the endpoint
host, wired through the prod overlay, write-prod-env.sh and both prod workflows,
and clarify in the docs that the endpoint variable takes the host alone.
Add pgBackRest-based PITR for the production database, shipped disarmed
(archive_mode and the base-backup timer gated off) so it stays inert until the
operator arms it before real payments — an un-armed deploy cannot pile WAL onto
the disk.
- Postgres now builds from a thin image carrying pgBackRest (archive_command runs
in-process); the prod overlay wires an encrypted (AES-256-CBC), path-style S3
repository with 30-day retention and a daily full base-backup systemd timer.
- Repository config/secrets flow through the PROD_PGBACKREST_* Gitea set and
write-prod-env.sh; prod-rollback re-renders the same env so a rollback cannot
disarm archiving.
- Grafana alerts watch pg_stat_archiver (failing / stalled), absent-safe on the
test contour, which never archives.
- Fix the pre-migration pg_dump to snapshot the whole database (was backend-only,
silently excluding payments).
- Document the PITR runbook, the arming sequence and the restore drill in
deploy/README.md; record the measured cost/perf assessment.
Add the "Кошелёк" section to the settings hub: context-visible chip
balances, active benefits (no-ads term/forever, hints) and a storefront of
chip-priced values and money-priced chip packs. Guests have no wallet; the
Google Play build hides the money purchases behind a RuStore stub; a web
purchase that would draw VK/Telegram chips warns first.
Add the catalog read path the storefront needs — a context-projected
GET /api/v1/user/wallet/catalog (payments service + store, gateway op
wallet.catalog, FBS Catalog/CatalogProduct/CatalogAtom, client decode) —
plus the client leg for the existing wallet.get/buy ops. Value spends reuse
the existing spend path; the chip-pack purchase (money order flow) arrives
with payment intake, so its action is a disabled placeholder for now.
Covered by Go unit (catalog projection) + integration (/wallet/catalog over
Postgres), vitest (formatting, spendable selection, web-spend warning, GP
flag, codec + gateway encode round-trips) and Playwright mock e2e (render,
guest-hidden, GP stub, warning) on Chromium + WebKit.
Stand up the internal chip/benefit mechanic behind the narrow payments interface:
context-aware balances and benefits, an atomic chip spend, admin grants as
zero-price value sales, the one-directional store-compliance gate (VK/TG same-
origin only, web draws direct→vk→tg, VK-iOS frozen, untrusted fail-closed), and
per-origin hint and no-ads application with term stacking. Reads are served from
an in-process, account-keyed write-through cache (mirroring the suspension gate),
so hot paths issue no query to the payments schema.
Flip the online-game hint wallet and the ad-banner suppression from the deprecated
accounts.hint_balance / paid_account columns to the payments benefit (a hint
balance no longer suppresses the banner — only a no-ads benefit does), and fold
chip segments and benefits by origin on account merge, inside the merge tx. Add
the GET/POST /api/v1/user/wallet edge chain (REST → Connect → FlatBuffers) plus
its codec unit test; no wallet UI yet.
Bring the frozen owner decisions log into the repo at
docs/PAYMENTS_DECISIONS_ru.md (it was untracked under .vscode) and reference it
from PLAN.md; record the read-cache design and the present-sources interface in
PLAN.md and docs/PAYMENTS.md (+ RU mirror).
Record the execution platform (kind vk|telegram|direct + device subtype
ios|android|web) on each session, captured at creation and carried
gateway->backend as a trusted X-Platform header, so the upcoming
store-compliance gate has an unforgeable execution context.
- backend.sessions gains nullable platform_kind/platform_subtype columns
(migration 00011, CHECK-constrained, jet regenerated); session.Platform
captures them at mint, resolve returns them, middleware exposes platform(c).
kind is derived from the establish endpoint, never a client field; the
account-merge session mint inherits the caller's platform.
- gateway derives the platform (VK subtype from the signed vk_platform via
vkauth, Telegram/direct best-effort from the client) and injects X-Platform
on every authenticated backend call through the request context.
- ui submits a best-effort device subtype on the telegram/guest/email login
requests (new FBS subtype field); VK is server-derived from the signed params.
- an unattributed session is untrusted (view-only); VK/TG self-heal on the next
cold-start re-mint, direct/email on re-login.
Signal plumbing only, no user-visible change; X-Platform is inert until the
gate consumes it.
The committed go-jet code had drifted from the schema: robot_blocks and
robot_friend_requests (created in the baseline) had no generated tables, the
feedback_messages model was missing app_version/browser_tz (added in 00003 and
00004), and UseSchema omitted account_best_move and the two robot tables.
Regenerate so the jet layer mirrors the migrations again.
Additive only — two nullable columns appended to feedback_messages plus two new
tables; the full build, vet, unit and integration suites stay green.
Stand up the payments data foundation: a self-contained `payments` Postgres
schema for the in-game currency, wallets, benefits, catalog, orders and the
append-only operations ledger, behind a domain package — nothing wired to real
money yet.
- Migration 00010: the `payments` schema and a NOLOGIN confinement role (ALL on
payments.*, nothing on backend); the ledger with a BEFORE UPDATE/DELETE
append-only trigger and a partial idempotency index; the materialised
balances/benefits; the catalog (atoms seeded) + products + per-method prices;
the typed single-row config; orders and payment_events. There is no
cross-schema foreign key — account_id is a plain uuid kept consistent in code,
which keeps the domain extractable. Expand-contract and reversible.
- Money is a bigint in the currency's minor units carried by a `Money` value
type (exact, math/big): no float ever touches an amount, and a whole-unit
currency cannot hold a fraction.
- Extend jetgen to generate the payments schema; construct the service in the
composition root behind a narrow interface with a boot reachability check.
- Tests: integration (role confinement via SET ROLE, the append-only trigger,
CHECK constraints, the idempotency index, and a forward+backward migration),
Money unit tests, and an import-boundary test keeping the payments jet code
private to the domain.
- Docs: PLAN.md, docs/PAYMENTS.md (+ _ru mirror) updated to the built model.
The deploy job gated only on the event/ref, so a docs-only PR into
development (where unit/integration/ui/conformance path-skip) still
redeployed the test contour for nothing. Gate it on the `changes` job
too: deploy only when the Go or UI side changed. `changes` still
defaults both true when the diff is uncomputable, and a workflow/deploy
edit forces both true, so ambiguous or infra changes still deploy.
The 2-4 hotseat players now start in a random seating order (so who goes
first, and the turn order, is random) — matching the online games. The
engine starts at seat 0, so shuffling the seats at creation picks the
starter; the shuffle is seed-driven (a distinct derivation from the tile
bag), so replay reproduces it. Scoped to hotseat — vs_ai stays human-first.
- lib/roster.ts: shuffleSeeded (unit-tested); NewGame shuffles buildSeats
with the game seed before create.
- e2e: pin a seed that keeps the roster order so the lock assertions stay
deterministic.
Bring the online lobby badge and the in-game 'you won/lost' text to the
same competition ranking the hotseat seat medals use: in a 3-4 player game
a TIE for the lead is a SHARED victory for the top scorers and a placed
finish for those below — not a full draw for everyone (winner() is -1 on
any tie, so isWinner alone conflated them). An aborted game and a genuine
all-level finish stay a draw; a win by resignation (winner at a not-higher
score) is unchanged.
- result.ts: resultBadge no-winner branch ranks by final score; placeBadge
helper; resigned seats excluded from the ranking.
- Game.svelte: resultText aligned (shared lead = won, below = lost).
Reported: a hotseat game ended by 6 scoreless passes deducts each rack
(correct rule, same as online: -8/-14/-8 here), which tied two seats for
the lead. The engine's winner() returns -1 on a tie, so seatMedal read it
as a full draw and gave EVERY seat the last-place medal.
- result.ts: seatMedal now ranks by the FINAL score (competition ranking —
a tie for the lead shares the trophy), not the single-winner flag. A
resigned / host-excluded seat places last with no medal and does not push
the others down.
- model.ts: Seat.resigned (offline-only); engine.resignedOf getter; the
local source surfaces it on the game view.
- The scoreless rack deduction is unchanged (standard rules, owner-confirmed).
Per owner: a vs_ai game (one human) keeps its 'you won/lost' status text
AND its lobby medal — the hidden-status + per-seat-medal treatment applies
only to hotseat, where 2-4 local players make a single 'you' meaningless.
isLocalGameId -> game.hotseat at the three call sites (statusBlock, seat
plaque, lobby card).
For a local (offline) game the outcome is not always about a single
'you' (hotseat is 2-4 players), so:
- Game.svelte: a finished local game no longer shows the viewer-centric
'you won/lost/draw' status; instead each seat plaque shows a per-seat
place medal, left of the name (result.seatMedal — trophy for the winner,
then places by score; a draw medals everyone).
- Lobby.svelte: a local game shows no lobby medal (resultBadge is
viewer-centric and no seat matches the account) — its result lives on
the in-game plaques.
- result.ts: seatMedal(game, seat), unit-tested.
- History drawer in a local (offline) game: the seat plaques no longer
show the add-friend/block controls — canAddFriend/canBlock now exclude a
local game (hotseat seats have synthetic account ids that slipped past
the vs_ai-only guard) — and the Dictionary entry is restored: an active
hotseat game keeps the comms button, and CommsHub is Dictionary-only for
a chatless vs_ai OR hotseat game (ChatScreen never mounts offline).
- Enter on any single-line <input> now dismisses the soft keyboard (blur);
a <textarea> (feedback) keeps Enter for newlines. One global handler.
- Login email code: the friend-code spread-digit style (.codein), a
6-char cap, auto-submit on the 6th digit, and Enter to submit.
- e2e: the local-game history has no social controls and keeps the
dictionary entry.
iOS Safari/WKWebView does not shrink the layout viewport for the soft
keyboard (Android Chrome does): the visual viewport shrinks AND offsets
down toward the focused field, and the values do not cleanly revert. The
pinned shell tracked only the height (--vvh), not the offset, so its
top-anchored content misaligned on iOS — empty space below the form,
worst on a repeat focus. Fix (one place; covers NewGame, Chat, Feedback):
- app.svelte.ts syncViewport: also mirror visualViewport.offsetTop into
--vv-top, and scroll the focused field into view on the keyboard-open
transition (iOS does not reliably scroll a pinned document to it).
- app.css: the pinned app-shell body follows top=--vv-top + height=--vvh
(was inset:0), so it stays on the visible area.
- e2e/viewport.spec.ts emulates the visual-viewport resize+offset (a fake
window.visualViewport) to verify the shell follows, on Chromium+WebKit.
Hotseat: returning to the lobby now re-locks the current seat (its PIN is
re-prompted) — LocalSource.relock, cleared in Game.svelte onDestroy.
Root-caused a crash it exposed: a $props() value (id) reads back undefined
during Svelte teardown, so isLocalGameId(id) threw and aborted onDestroy
(breaking navigation) — read the id from the loaded view instead.
- NewGame: all three forms (quick / friends / hotseat) now use the
Profile sub-view's natural-flow .page (no forced height, no pinned-CTA
.grow/.fg scroll), so the screen's single .content scroll + --vvh handle
overflow and a focused name input scrolls into view cleanly — fixes the
keyboard blank-space / deep-scroll regression. One layout rule, one place.
- PinPad: pause 250ms after the 4th digit before the verdict, so the fill
(and the shake on a wrong PIN) reads as a deliberate response.
- e2e: typePin waits for empty dots before typing (robust to the pause).
The nested scroll region regressed the keyboard behaviour (a double
scroll: the whole screen scrolled with a huge blank area under the Start
button, and a focused bottom row hid behind the keyboard). Drop it: the
hotseat form now flows naturally and scrolls with the screen's own scroll
(.page.scrollform: no forced full height, no pinned button), and a name
input scrolls itself into view (centred above the keyboard) on focus.
Owner review of the pass-and-play PR:
- Roster delete (bug): a correct host PIN now ARMS a delete cross next to
the row (mirroring the lobby delete) instead of removing it silently;
tapping the kebab again clears the host authorisation.
- Variant picker (UX): replace the dropdown with the Quick-Match variant
plaques + the multiple-words toggle, one-to-one.
- Keyboard layout (UX): the roster sits in a scroll region with the Start
button pinned (friends-form pattern), so a name input's soft keyboard
shrinks the region instead of leaving a full-height blank spacer painted
behind the keyboard.
- e2e: variant via plaque + a row-delete (arm then cross) regression step.
PWA stale-version (pre-existing, same PR):
- sw.ts: register the network-first NavigationRoute BEFORE precacheAndRoute.
Workbox matches in registration order and the precache route (directoryIndex
index.html) shadowed the shell navigation, serving it cache-first — the old
build's __APP_VERSION__ until a second load. Fixes both the maintenance
self-reload and a cold open after a deploy. Doc updated (ARCHITECTURE).
- source: a finished hotseat game is never locked (its final rack is
shown for review, not an Unlock button); guard locked on !isOver.
- bundle-size: raise the app-entry budget 115 -> 120 KB for the hotseat
UI (PIN pad, roster, host menu), which lives in the always-loaded New
Game / Game / Lobby screens; the offline engine + PIN hashing stay lazy.
Add the Playwright mock e2e for offline pass-and-play (Chromium+WebKit):
create a 2-seat hotseat with a PIN-locked seat, verify the rack is
withheld, unlock (wrong then right PIN), host-skip the current turn, and
terminate from the lobby behind the master PIN.
The e2e surfaced a persistence bug: the roster PINs are $state proxies,
which fail structured-clone on the IndexedDB put (silently — best-effort
store), so a created hotseat game vanished on reload. Snapshot the seats
+ host PIN to plain objects at the create() call site.
Lobby.svelte: hotseat games show the kebab / swipe delete in BOTH the
active and finished groups (not finished-only); deleting one opens a
master-PIN pad first — active = terminate (no result), finished = remove
— so the last mover cannot instantly wipe a game. Local games stay
deletable offline. Non-hotseat games are unchanged (finished-only, free).
Game.svelte drives a hotseat game:
- the seat-to-move's rack is replaced by an Unlock button when the seat
is PIN-locked (board stays visible); canMove gates the move controls
on the unlock; PinPad(verify) -> source.unlockSeat reveals it.
- the hint slot becomes a host button (hints off in hotseat), always
enabled while the game runs (acts on a locked seat too). Master-PIN ->
action sheet: skip current / exclude a player / end the game, each with
a confirm + a fading check; terminate deletes and returns to the lobby.
- per-turn advance re-fetches the next seat's (locked) state; self-resign
and chat are hidden (hotseat resign is a host action, no comms).
- gamesource: expose unlockSeat / verifyHostPin / hostAction on the local
proxy. i18n hotseat.* (en + ru).
NewGame offline 'with friends' now builds a local pass-and-play game:
- lib/roster.ts: keep-last-valid name + roster->seats (unit-tested).
- NewGame.svelte: master host-PIN gate (rows disabled until set),
'are you playing too?' prompt seating the host at row 0, 2-4 player
rows with inline name validation + optional per-seat PIN, add/remove
(remove behind the master PIN), variant + multiple-words, Start.
- PinPad: verification via a verify(pin) callback (UI-held lock at
creation, source-held in-game) instead of a lock prop.
- i18n: hotseat.* (en + ru); the offline mode selector is now shown offline.
Extend the local game to offline pass-and-play (hotseat):
- serialize.ts: Seat.pin, record hotseat + hostPin (all optional; old
vs_ai records read hotseat as false).
- source.ts: create() takes hotseat/hostPin + per-seat pins; stateView
reveals the seat-to-move's rack and withholds it (locked) when that
seat is PIN-locked, until unlockSeat; ephemeral per-turn unlock
re-locks on advance; new unlockSeat / verifyHostPin / hostAction
(skip/resign/terminate); gameView sets vsAi=!hotseat + the hotseat flag.
- model.ts: GameView.hotseat, StateView.locked (offline-only, optional).
Tests: source.hotseat (lock/unlock/re-lock, host skip/resign/terminate,
master-PIN gate, natural-end persistence) + serialize hotseat shape.
Groundwork for offline pass-and-play (hotseat) seat/host PIN locks:
- lib/pin.ts: salted SHA-256 PinLock (newSalt/hashPin/newLock/verifyPin);
a social lock for a shared device (documented, not cryptography).
- components/PinPad.svelte: 4-digit keypad (set/verify/change), auto-submit
on the 4th digit, shake on mismatch, hardware-keyboard support.
- i18n: pin.* keys (en + ru).
Engine/source/UI wiring follows.
navigate() wrote location.hash and left router.route to the asynchronous
hashchange event. bootstrap flips app.ready in the same tick right after
navigate('/login') on an unauthenticated cold start, so for one frame the app
rendered with app.ready=true and the stale route ('lobby', from the empty hash)
— briefly mounting the lobby shell (tab bar + a doomed games.list) under the new
login screen before hashchange settled it.
Update router.route synchronously inside navigate(); the later hashchange
re-parses to the same value. This removes the boot lobby-flash (and its spurious
games.list 401) and, as the root cause, the offline e2e flake: enterLobby could
latch that transient lobby tab-bar instead of clicking through login, then the
Settings-tab click at line 44 fought the tab detaching into the login slide
(~7% of runs, both engines).
Also:
- offline.spec enterLobby: wait for the guest button and click it, instead of a
point-in-time count() that a pre-login splash frame sampled as 0 (skipping the
click and hanging / latching the transient).
- New e2e router.spec pins the synchronous-route property (RED on the old code,
which returned the previous route; GREEN now) via a mock-only __router seam.
Verify: check 0 / unit 490 / build / bundle 114.3/115 / full e2e 200 / the
offline+router stress at 40x on both engines 160/160 (was 6/80 failing).
Entering a vs_ai game from the lobby armed the idle-hint countdown only in load() (after
the gameState round-trip), so the lock popped in after a visible delay. Arm it on the
instant warm-cache render in onMount too — the preloadGames-warmed StateView carries
hint_unlock_left_seconds; load() then refreshes the snapshot. A first move (0 seconds
left) stays open.
Online vs_ai hints were broken: #207 made the vs_ai hint button always-enabled and
wallet-free (assuming all vs_ai = the offline idle-gate), but the backend still served
online vs_ai from the allowance/wallet, so over-clicking hit ErrNoHintsLeft -> a generic
error toast. Now online vs_ai uses the SAME idle-gate model as offline.
Backend: Hint() for a vs_ai game skips the allowance/wallet and increments no hints_used
(owner: vs_ai counts toward no hint statistic), and is idle-gated from the SERVER clock --
it returns ErrHintLocked (code hint_locked) until the robot's last move + 30 min, else
serves the top move. GameState/StateView expose hint_unlock_left_seconds (server-computed
seconds remaining; 0 for a human game / first move / not-your-turn). Pure helper
hintUnlockLeftSeconds unit-tested.
Wire: StateView gains hint_unlock_left_seconds (FlatBuffers, additive); pkg/wire + gateway
transcode carry it (round-trip test).
Client: the gate counts down from a MONOTONIC clock (performance.now()) anchored to the
source's seconds-left when it lands (on load from the view; to the full window when the
robot moves), so a client clock change cannot skew it and a relaunch re-reads a fresh
value. The vs_ai hint button (online + offline) shows the lock + toast; doHint handles the
hint_locked backstop by re-syncing. Replaces #207's absolute hintUnlockAtMs on the
wire/model/delta (the offline record keeps the absolute for persistence; the view exposes
seconds-left).
Docs FUNCTIONAL(+_ru)/ARCHITECTURE updated. Verified: go build/vet + game/server/transcode
tests (+ new hintUnlockLeftSeconds); ui check 0 / unit 490 / e2e 198 (one pre-existing
webkit offline flake) / app entry 114.2/115.
Two pre-existing offline bugs, both exposed once offline mode became usable (cold boot +
toggling), owner-reported on the contour.
Bug 1 -- offline tiles render weight 0. The rack/board read tile values from the
lib/alphabet cache, populated only by the online wire codec (server-sent alphabet) or the
mock (so the e2e masked it). A local game never goes through the codec, so on a cold
offline boot with no prior online session the cache was empty and every value read 0
(scores stayed correct -- they come from the offline ruleset). Fix: the local source
seeds lib/alphabet from the static offline ruleset (letters + values) when it builds a
game view.
Bug 2 -- the lobby showed (and could open) the other mode's game after a toggle. Two
causes: (a) the lobby cache was not mode-tagged, so getLobby() instant-rendered the last
snapshot regardless of mode; (b) load() wrote the module list after an await with no
generation guard, so a slow online gamesList() finishing after a fast offline list() (or
vice versa) left the wrong mode's games on screen. Fix: tag the snapshot with offline and
gate getLobby() on it; add a load-sequence guard so only the latest load() writes.
- localgame/source.ts: ensureAlphabet from the ruleset in gameView (+ unit test).
- lobbycache.ts: LobbySnapshot.offline + getLobby(offline) mode check (+ tests).
- Lobby.svelte: setLobby tags the mode; load() carries a generation guard.
check 0 / unit 485 / e2e 198 / app entry 113.8/114.
Per the owner's call, the idle-hint gate now PERSISTS across leaving and reopening the
app, instead of the session-scoped monotonic clock: the unlock is a wall-clock instant
(hintUnlockAtMs) stamped from the robot's reply, stored on the local record, carried on
the game view + the opponent_moved move delta, and read through a sanitiser that caps it
at now + the window. So:
- the wait survives a relaunch (a stuck turn is not forgotten);
- a device clock set BACK cannot freeze the gate (the cap bounds the remaining to the
window and self-heals on the next read);
- a clock set FORWARD just opens the hint early -- accepted as harmless for a solo game.
- lib/hints.ts hintGateRemainingMs now takes the unlock instant + clamps to the window.
- localgame: re-add hintUnlockAtMs to the record/meta; stamp it off the robot's reply;
sanitise on read (a shared hintUnlock helper feeds stateView + the opponent_moved event).
- gamedelta + PushEvent: the move delta carries hintUnlockAtMs so the view stays fresh.
- Game.svelte: hintUnlockAt derives from the view; the tick + lock + toast unchanged.
- offline.spec.ts: also assert the lock survives a reload (the wait persisted).
- Docs FUNCTIONAL(+_ru)/ARCHITECTURE updated (persist + sanitiser, not monotonic-resets).
A vs_ai hint is now unlimited and wallet-free, but idle-gated as an anti-frustration
aid: it unlocks only after the player has been stuck ~30 min on a turn (timed from the
robot's last move; the human's first move, before the robot has played, is exempt).
While gated the hint button carries a small lock badge and a tap shows the remaining
minutes ('Available in N min.'); the lock lifts live at the mark.
The gate is enforced CLIENT-SIDE against a MONOTONIC clock (performance.now()), never a
wall-clock timestamp: a device clock the player controls (or an auto-sync) must not be
able to open or freeze it. The wait is therefore session-scoped -- a reload restarts it
(there is no tamper-proof way to carry idle time across a relaunch without a wall clock).
An online vs_ai game will gate the same way but from the server's clock (a follow-up).
- lib/hints.ts: HINT_GATE_MS + pure hintGateRemainingMs (monotonic) + hintLockMinutes;
removed the wall-clock hintRemainingMs. Unit-tested red->green.
- Game.svelte: the monotonic gate (hintGateStart/monoNow, armed on the turn change), a
vs_ai hint button (plain: no confirm, no count; lock badge + gated tap -> toast), a
live 10s tick.
- localgame: removed the wall-clock hint gate and the now-dead robotLastMoveAtUnix field
from source.ts/serialize.ts (the client is authoritative); hint() just serves the top
move.
- i18n game.hintLockedIn.
- offline.spec.ts: assert the first move is un-gated and the lock arms after the robot moves.
- Docs FUNCTIONAL(+_ru) + ARCHITECTURE; bundle budget 114->115 (game-screen feature).
A Playwright spec drives the whole offline flow in the mock build: force the
installed-PWA display mode, enter offline via the Settings toggle (its readiness check
fetches the dawgs), assert the blue chrome + online-games-hidden + Stats-disabled, then
create and play a device-local English vs_ai game with a pinned bag seed (deterministic
rack NEWYMAO) -- play the opening WAY across the centre, watch the robot reply with a
real move, and reload to confirm the IndexedDB replay.
Enabling infra (all e2e-only; nothing enters the production build):
- mock/client.ts fetchDict serves the per-variant dawgs from the preview build's
/e2edict/ (was: threw 'unsupported in mock').
- scripts/e2e-dict.mjs copies the real dawgs into dist-e2e from E2E_DICT_DIR (the ui CI
job fetches the scrabble-dictionary release like the Go jobs; local default: the
sibling scrabble-solver/dawg); playwright.config runs it between build and preview.
- localgame/id.ts setForcedSeed + gateway.ts window.__mock.setLocalSeed: a mock-only
seam to pin a local game's bag seed (tree-shaken from prod).
- ci.yaml ui job: fetch the dawgs + pass E2E_DICT_DIR to the e2e step.
- docs/TESTING.md: the offline e2e + the mock-dawg wiring.
Verified: check 0 / unit 482 / e2e 198 (both engines) / app entry 113.8/114.
Flipping the Settings toggle to offline now checks that every enabled variant's
dictionary is on the device before entering offline mode: it fetches missing ones
cache-first and waits up to ~5 s (raceOfflineReady + the lazy dict/offlineready),
greying the toggle meanwhile. If they cannot be readied in time it stays online and
shows a 'needs internet' note, while the fetch keeps warming the cache in the
background so a later flip is instant. Leaving offline is never gated.
Prevents entering a half-baked offline mode (no dawg -> cannot create/play a local
game) when the background preload has not finished (poor connection, or an immediate
flip right after install).
- offline.ts: raceOfflineReady (pure, injected sleep; unit-tested red->green)
- dict/offlineready.ts: ensureOfflineDicts (cache-first preloadDicts, lazy chunk)
- offline.svelte.ts: requestOffline + TOGGLE_READY_BUDGET_MS
- Settings.svelte: checking/needsData state, disabled toggle, inline note
- i18n: settings.offlineChecking / settings.offlineNeedsData (en+ru)
- docs: FUNCTIONAL(+_ru) offline story + ARCHITECTURE offline paragraph
The offline auto-detect shipped in #202 (cold-start no-connection dialog / auto
offline) and #203 (mid-session flight-mode reactivity) but neither PR baked the
docs. Add the user story to FUNCTIONAL.md (+ _ru mirror) and the mechanism to the
ARCHITECTURE offline section: the auto vs deliberate distinction, the cold-start
navigator.onLine / bounded reachability-probe decision, the mid-session window
online/offline events backed by a navigator.onLine poll (the events are unreliable
on some platforms, notably iOS PWAs), and the reachability probe being the one call
exempt from the offline kill switch.
The auto-offline -> online return was dead because setOfflineMode never set the
'auto' flag: I added the state + getter but forgot the assignment, so it stayed
false. So scheduleRecovery bailed immediately (no poll ran) and the online event's
'if (active && auto)' was false. The contour diagnostic confirmed offline.auto
stayed false after an auto-offline. Add 'auto = on && !persist'.
Also remove the temporary network diagnostic panel (netdiag + the lobby strip) --
it did its job of pinpointing this.
A fixed strip in the lobby showing live navigator.onLine / offline.active /
offline.auto / connection.online plus a timestamped event log (offline/online
events, poll ticks, checkReachable results, mode changes) - to pinpoint where the
auto-offline -> online transition breaks on the contour. A 1s heartbeat logs
navigator.onLine flips even if the events never fire. Skipped in the mock; to be
reverted with the fix.
The auto-offline -> online return still did not fire: the retry hung on the
'online' event, which an installed PWA often does not deliver. Replace it with a
lightweight poll while in auto-offline that reads navigator.onLine (a reliable
live flag, unlike the event) and only hits the network for a reachability check
when the interface is actually up - so flight mode ON costs no radio, and flight
mode OFF is detected within ~4s and returns online. The 'online' event, when it
does fire, just kicks an immediate check. Runs only in auto-offline (deliberate
offline is the player's choice); wired for both the mid-session and cold-start
auto-offline paths.
Two mid-session issues found on the contour:
- Auto-offline did not return to online after the network came back: a single
reachability check right after the 'online' event failed (the interface is up
before the gateway is actually reachable again). Retry a few times with backoff
(tryReturnOnline) — that is what actually gets the app back online.
- An online game viewed while offline (flight mode on) still enabled its network
actions, so they hit the kill switch and raised 'something went wrong' toasts.
Gate them: netReady = isLocalGame || (connection.online && !offlineMode.active)
— a local game stays fully usable; an online game's make/exchange/hint/resign
disable while offline and re-enable when back. Also suppress the 'offline' code
in handleError (a blocked call in offline mode is expected, not a toast).
React to the network changing while the app is open (e.g. the player toggling
flight mode), via passive online/offline events - no polling, no battery cost:
- interface lost -> enter offline mode for the session (auto);
- interface back -> if the offline was auto, verify the gateway is really
reachable (an interface being up does not guarantee it) and return online; a
deliberate offline (the toggle or the cold-start dialog) is left as-is.
- offline.svelte: track `auto` (auto-detected vs the player's deliberate choice).
- connection.svelte: checkReachable is now a pure one-shot (the caller decides);
the reachability watcher never probes in offline mode (events drive recovery).
- transport.ts: the reachability probe is exempt from the kill switch - it IS
the mechanism that decides whether to return online, fired only deliberately.
- app.svelte.ts: initNetworkReactivity wires the events (web-only, skipped in
the mock); called from bootstrap.
Online unaffected (skipped in the mock e2e): e2e 196. Mid-session reactivity is
contour-verified.
A sticky-online cold start with no network hung the splash on adoptSession's
retrying profile fetch. Now, for an offline-capable web install with a cached
profile:
- No network interface (navigator.onLine === false) -> enter offline mode for
the session (no dialog; the next launch re-evaluates).
- Interface up but the gateway is unreachable within 3s (a single-attempt
reachability probe, not the 6-retry loop) -> a 'No connection. Enable offline
mode?' dialog: Enable -> sticky offline; Keep trying -> the normal online
adopt (retries, 'Connecting...').
- connection.svelte: checkReachable(timeout) - a bounded single probe.
- offline.svelte: setOfflineMode(on, persist) - auto-offline is session-only, a
deliberate choice (dialog/toggle) is sticky.
- app.svelte.ts: the cold-start auto-detect in bootstrap + the dialog resolver;
App.svelte renders the boot dialog. i18n en/ru.
- App-entry bundle budget 113->114 (the boot path cannot be lazy-loaded).
Online cold-start unaffected (auto-detect gated to isStandalone, off in the mock
e2e): e2e 196. The offline paths are contour-verified.
Next: PR2 - mid-session flight-mode reactivity (online/offline events).
The lobby's finished-game delete (hide) always called gateway.hideGame; for a
local (offline) game the transport kill switch refused it, so it toasted and the
game stayed. Route by id: a local game is removed from the device store + the
source cache (LocalSource.delete), an online game still hides on the backend.
The C1 service worker served the precached shell for every navigation, even
online, so a new deploy only reached a client after the worker updated its
precache in the background (a one-launch lag): a cold online start showed the
old version until then, and only a reinstall forced it fresh.
- sw.ts: navigations are now network-first — fetch the fresh (no-cache) shell
from the server on each online launch (it references the new hashed assets,
fetched fresh), with a 3s timeout falling back to the precached shell when the
network is unreachable or too slow. Only the tiny HTML is re-fetched; the
immutable hashed assets stay cache-first and re-download only when their hash
(the version) changes. Offline cold-launch still works via the fallback.
- webui.go: serve sw.js with Cache-Control: no-cache so the browser reliably
detects a new deploy's worker (regression-tested).
Online launch is contour-verified (the mock e2e disables the SW); e2e 196.
No new deps — the network-first handler is hand-written (~12 lines).
NewGame's onMount fetched the friend list (gateway.friendsList) for a non-guest;
offline the transport kill switch refuses it, so it toasted on entering the New
Game screen (the local game still created fine). The friends section is hidden
offline anyway — skip the fetch when offline.
Offline mode was 'fiction': the UI only disabled the Stats tab, but Profile was
reachable and a profile save actually hit the server and reported 'saved'.
Two layers now:
- Transport kill switch (transport.ts): in offline mode every network op — each
unary RPC (exec), the live stream (subscribe), the dict/metrics fetches and
the reachability probe — is refused before it leaves the device. Offline
truly means offline regardless of any UI that slips through. The local
(device-only) game path never uses this transport, so it is unaffected.
- UI gating: the Profile and Friends tabs (SettingsHub) and the Feedback entry
(About) are disabled offline, and the hub falls back to Settings (which holds
the online/offline toggle); the lobby Stats tab was already disabled. In-game
social/export are already hidden for a vs_ai game.
Online unaffected (offlineMode is off): e2e 196. Offline gating is
contour-verified (the mock e2e cannot enter offline).
In a local game the human seat's account id was a synthetic 'local:human:0',
so seatName's `accountId === session.userId` check never matched: the game
header rendered BOTH seats as 🤖 (the vs_ai fallback), and the lobby's
groupGames could not find the viewer's seat, so a human's turn read as
'Their turn' with the hourglass. Carry the real account id on the human seat
(create -> record -> GameView); the robot keeps its synthetic id.
Local games created before this fix keep the old display (no migration).
The offline cold-boot never engaged: it persisted app.profile — a Svelte
$state proxy — which is not structured-cloneable, so the IndexedDB write threw
and fell back to a localStorage entry that loadProfile (IDB-only on a
successful-empty read) never reads. loadProfile returned null, the boot
short-circuit missed, and it even cleared the sticky offline flag — so offline
mode stopped persisting across relaunches (owner-observed on the contour).
- Persist $state.snapshot(app.profile) (a plain object) at both persist sites,
so the IndexedDB write succeeds and loadProfile round-trips.
- Drop the setOfflineMode(false) fallback: keep the deliberate offline flag on a
cache miss (the mode is the player's choice; an online boot re-persists the
profile so the next launch goes offline). A truly offline launch with no
cached profile is unreachable (enabling offline needs a prior online session).
An installed PWA relaunched with the network off hung on the splash: C1's
precache served the shell, but bootstrap() adopted the session and fetched the
profile over the network, which hung. Persist the profile (on every online
adopt + refresh, cleared on logout) and short-circuit bootstrap when the
deliberate offline flag is sticky-on: with a cached session + profile, skip the
network and land straight in the offline lobby. Without a cached profile (never
online), drop the sticky flag and boot online — the first launch must be online.
- session.ts: saveProfile/loadProfile/clearProfile (mirror saveSession).
- offline.ts: shouldBootOffline decision (unit-tested).
- app.svelte.ts: persist in adoptSession + refreshProfile, clear on logout,
the offline boot short-circuit in bootstrap.
Docs: ARCHITECTURE. Online cold-start unaffected (e2e 196); the offline boot is
contour-verified (the mock e2e cannot enter sticky-offline).
For an honest-AI (vs_ai) game the comms hub relied on a post-mount $effect to
switch from the Chat tab to the Dictionary, so ChatScreen mounted for a beat
and fired its chat/state fetch over the network. Online that was a wasted
call; offline it threw and raised a 'something went wrong' toast on entering
the word-check form (the check itself already worked). Start the comms hub on
the Dictionary tab immediately for a vs_ai game, so ChatScreen never mounts.
Two offline bugs found on the contour:
- The word-check screen (CheckScreen) called gateway.gameState/checkWord
directly, so in an offline (local) game it hit the network and errored
("something went wrong"). Route both through gameSource(id) — the local
source answers from the device dawg. The complaint control (online-only,
no offline backend) is hidden for a local game.
- The lobby did not react to the offline-mode toggle (which lives in Settings,
so the lobby can stay mounted): an online game lingered until the next
reload. Reload on an offlineMode flip so entering offline immediately shows
only device-local games.
Cold offline launch hanging on the splash (boot still fetches the profile over
the network) is the separate C2 offline-boot follow-up (needs a persisted
profile + a boot short-circuit).
In offline mode the lobby now shows only the device-local games and its
New-vs-AI entry creates one through the in-browser engine — the visible
payoff of the offline mode.
- LocalSource.list() reconstructs a lobby GameView per stored local game by
replay, exposed through the lazy gamesource proxy; unit-tested via an
in-memory store.
- Lobby.load() branches on offlineMode: lists local games and skips every
gateway call (no online games/invitations/incoming); the Stats tab is
disabled offline.
- NewGame offline: find() creates a device-local vs_ai game via
LocalSource.create using the profile's advertised dict version + a local
seed; the friends flow and the random-opponent option are hidden, and the
variant picker / Start are enabled offline (were gated on connection).
- id.ts: newLocalGameId + randomSeed (tested).
Docs: ARCHITECTURE + FUNCTIONAL(+ru) offline-mode section.
Deferred to fast-follow: the Settings Friends/Profile/Feedback affordance
gating, the flip-to-offline readiness wait, the offline mock e2e (needs
mock-dawg support), and the local-hint UI. The offline flow is verified on
the test contour — the mock e2e cannot enter offline mode (the toggle is
gated to an installed PWA).
Migrate the former install-only public/sw.js to a custom (injectManifest)
service worker built from ui/src/sw.ts by vite-plugin-pwa. It precaches the
app shell + hashed assets (Workbox) so an installed web PWA cold-launches with
no network, and falls in-scope navigations back to the precached shell (the
hash router resolves the route client-side). This is the C1 prerequisite for a
usable offline mode: without it a cold offline launch cannot load the bundle.
- ui/src/sw.ts: skipWaiting + clientsClaim + cleanupOutdatedCaches +
precacheAndRoute(__WB_MANIFEST) + a NavigationRoute fallback to index.html,
deny-listing the RPC path and /_gm. The Connect stream and API POSTs are
never precached nor intercepted.
- vite.config.ts: VitePWA(injectManifest); manifest:false (we ship our own),
injectRegister:false (registration stays manual + web-only in pwa.svelte.ts),
disabled in the mock build (Playwright unperturbed); landing + polyfills
excluded from the precache.
- Removed public/sw.js; the registration path (dist/sw.js at /app/) is unchanged.
- New dev deps: vite-plugin-pwa + workbox-{core,precaching,routing}.
Docs: ARCHITECTURE + gateway/README. Offline cold-launch is contour-verified
(the mock e2e harness disables the SW).
An installed PWA (standalone web + confirmed email) warms the dictionaries
for the player's enabled variants while online — on lobby entry and on a
variant-preference change — using the per-variant version the profile now
advertises, so a later switch to offline mode already has the data.
- dict/preload.ts: pure preloadDicts (retry + linear backoff, honours the
session dictionary miss-breaker); node-tested.
- dict/preloadrun.ts: lazy browser orchestration (real getDawg), imported
dynamically so the loader and move generator stay out of the main bundle.
- offline.svelte.ts: kickDictPreload, gated by the pure, tested
offlinePreloadEligible, plus the reactive first-lobby preload warning.
- Header shows a poor-connection notice in the ad-banner slot on a
first-lobby preload failure (offline.preloadWarning, en/ru).
- App-entry bundle budget 112->113 for the irreducible main-side wiring
(documented in bundle-size.mjs); the heavy parts remain lazy chunks.
Docs: ARCHITECTURE offline-mode + dict-preload mechanism.
The Profile now carries dict_versions (game variant -> current dictionary
version), populated from the dictionary registry at the profileResponse
choke point, so an installed PWA can preload the matching dawg per enabled
variant off the existing cold-start profile request instead of adding a
round-trip for a rare feature.
Wire path: FBS DictVersion table + Profile.dict_versions (additive,
backward-compatible trailing field) -> backend dto/registry -> gateway
ProfileResp + FBS encoder -> client codec decode into a per-variant map on
model.Profile. Empty in a degenerate no-dictionary deployment; the mock
serves v1.3.0 for all three variants. Codec decode covered by a
bite-tested round-trip unit test.
The deliberate offline mode is now visible: an installed web-PWA user with a confirmed
email can switch to offline in Settings, and the header turns blue with an "Offline"
chip. (Network gating, the dict preload, the offline lobby + local-game creation, and the
service-worker precache follow in later Phase C steps.)
- offline.ts / offline.svelte.ts: a sticky, device-scoped reactive offline flag
(offlineMode), distinct from connection.svelte's transient reachability signal, with the
pure persistence + readiness helpers (offlineReady / missingDicts) unit-tested.
- Settings.svelte: an Online / Offline toggle, shown only for an installed web PWA with a
confirmed email (the SW-launch + durable-account preconditions).
- Header.svelte: a blue-tinted nav (color-mix from the accent, so it tracks light/dark and
a Telegram theme override) + an "Offline" chip while offline; the deliberate mode
suppresses the transient "Connecting…" indicator.
Behaviour-preserving online (the toggle is hidden and offlineMode is false there; e2e 196
green). App entry stays within its size budget (111.6 / 112 KB gzip).
The game screen now drives a local vs_ai game through the offline engine, dispatched by
game id — completing the playable local game (on top of the source, #193). Online play
is unchanged.
- gamesource.ts: gameSource(id) returns the local source for a `local:` id, else the
gateway (the same game-loop interface). The offline engine stays OUT of the app entry
bundle — it is dynamically imported on first use (a separate chunk), so online-only
users never pay for it (the app entry stays within its size budget).
- localgame/id.ts: the tiny id helper (no engine imports) the dispatcher branches on.
- Game.svelte: the game-loop calls (state/history/submit/pass/exchange/resign/hint/
evaluate/draft) go through gameSource(id) instead of the gateway directly; a local
game's robot-reply events route through the same app event hub the network stream
feeds, so the screen reacts to opponent_moved / game_over identically.
Behaviour-preserving for network games (gameSource returns the gateway for them). Local
verify green: check + test:unit + build + bundle-size gate + e2e (196 passed).
A GatewayClient-shaped facade over the offline engine, so the same game screen can drive
a local vs_ai game with no backend (the wiring into Game.svelte is B3.2).
- source.ts: LocalSource implements the game-loop subset (GameLoopSource) for a local game
id — gameState/gameHistory via replay, submitPlay/pass/exchange/resign apply the human
move then run the robot (decide(generateMoves)) synchronously, persisting both and
delivering the robot's move through a per-game event emitter (no live stream). hint is
gated to >30 min since the robot's last move; evaluate/checkWord are local. It translates
the UI's glyph space to the engine's index space with the static letters table.
- ruleset.ts: add the static per-variant letters (glyphs), pinned to the Go alphabet —
offline is now fully self-contained (no reliance on a warm server alphabet cache).
- engine.ts: submitPlay (infers the direction like the server SubmitPlay), evaluatePlay +
dictionaryHas for the move preview / word check, and record the main-word coordinate +
the words on a play (for the history MoveRecord).
- source.test.ts: create -> human pass -> synchronous robot reply via the event, the hint
gate, decoded history, and a whole game driven to completion.
Pure additive library code; no runtime behavior change (bundle unchanged).
Store an offline game durably and reconstruct it — the offline counterpart of the
server's replay rehydration. Builds on the engine (#191); not wired into the UI yet
(Phase B3).
- serialize.ts: a LocalGameRecord (seed + rules + seat metadata + the alphabet-index
move journal) and replayGame() — reconstruct a live LocalGame by seeding a fresh
engine identically and replaying the journal. The board/bag/racks are not stored; they
are deterministic from the seed and the replayed operations, so the record stays small.
The journal is dictionary-independent (alphabet-index space, stable per variant).
- store.ts: an IndexedDB store for local games (save/get/list/delete), mirroring
lib/dict/store.ts — its own database, best-effort, guarded when IndexedDB is absent.
- engine.ts: record the swapped tiles on an exchange (needed for exact replay) and expose
the game's rule config.
- serialize.test.ts: a round-trip — reconstruct a mid-game and a finished game by replay
and assert the state (board/racks/bag/scores/turn/log) is identical.
Pure additive library code; no runtime behavior change (bundle unchanged).
The offline vs_ai game engine — a faithful TS port of backend/internal/engine that
drives a whole local game with no backend. Composes with the move generator (#188) and
robot strategy (#189) from Phase A; not yet wired into the UI (Phase B2/B3).
- ui/src/lib/localgame/ruleset.ts: static per-variant tile values / bag counts / blank
count, mirrored from rules.go (offline scoring is self-contained; online uses the
server alphabet). Pinned by ruleset.parity.test.ts against a Go fixture.
- bag.ts: the tile bag (fill from counts/blanks, draw-from-end, return+reshuffle) on a
deterministic in-house PRNG — a game replays from its seed, not bit-identical to a
server game (per plan).
- board.ts: the mutable board, satisfying the validator/generator read view + set().
- engine.ts: LocalGame — deal / play (reusing validate.ts) / pass / exchange / resign,
scoreless(6) & out-of-tiles end detection, end-of-game rack penalties, winner; mirrors
game.go. The end-game math is exported as pure functions, pinned against the Go engine
(engine.parity.test.ts, 9 constructed positions).
- engine.test.ts: a full-loop smoke — two robots play a whole vs_ai game to completion
via generateMoves + decide, and it is reproducible from the seed.
- backend: movegen now dumps the per-variant rulesets; a new in-package engine emitter
(endfixture_test.go, env-gated) produces the end-game golden.
Pure additive library code; no runtime behavior change (unused at runtime, bundle unchanged).
Phase A (A4): prove the ported move generator (#188) against the FULL shipped
dictionaries, not just the tiny samples — the deep graphs and complete 26/33-letter
alphabets the samples cannot reach.
- backend/cmd/movegen: add a -dawg-dir mode that emits per-variant golden move-gen
vectors from the real dawgs (a bounded first-move + a blank case + a deep 7-tile
mid-game position). Regenerated in CI to /tmp, never committed (like the
dictgen/validategen vectors), so no dictionary version is pinned into the repo.
- ui/src/lib/dict/generate.realparity.test.ts: env-gated (DICT_DAWG_DIR +
DICT_MOVEGEN_DIR) parity against that golden — 9 positions across scrabble_en /
scrabble_ru / erudit_ru match the Go solver exactly. Skips cleanly when unset.
- .gitea/workflows/ci.yaml: the conformance job now generates the movegen golden
and points the gated vitest at it (DICT_MOVEGEN_DIR).
Phase A (2/2) of PWA offline mode: the offline robot picks its move exactly as the
server does, so a local vs_ai game plays the same. Builds on the move generator
from #188; not wired into a game loop yet (Phase B).
- ui/src/lib/robot/strategy.ts: port of backend/internal/robot/strategy.go's
move-choice slice — mix (FNV-1a, via BigInt for bit-exact uint64), playToWin
(~40% play-to-win), deviates (the fading off-strategy wobble) and selectMove
(pick the candidate whose resulting margin lands closest to the +/-[1,30] band,
conservative tie-break), composed by decide(). The generator's ranked moves feed
straight in. Think-time/sleep/nudge scheduling is server-only and not ported.
- backend/internal/robot/strategyfixture_test.go: an in-package, env-gated emitter
(EMIT_STRATEGY_FIXTURES=1) writing golden fixtures from the real Go strategy — it
reaches the unexported mix/playToWin/deviates/selectMove.
- strategy.parity.test.ts: 21 mix + 56 decision cases match Go exactly (play/
exchange/pass, the deviate flip, tie-break, band overshoot).
Pure additive library code; no runtime behavior change (unused at runtime, so the
bundle is unchanged).
First engine-first step of PWA offline mode (Phase A): the client-side move
generator — the "robot brain" a local vs_ai game will run on-device — with no
runtime wiring yet (Phase B).
- dawg.ts: add the step-by-step cursor (root/final/next/arcs), a faithful port
of dafsa traverse.go over the reader's existing bitstream.
- generate.ts: the Appel-Jacobson generator (leftPart/extendRight + cross-sets +
counts-rack + board transpose + moveKey ranking), reusing the cursor and
validate.ts evaluate/connected. A cross-set LetterSet is a Uint8Array, so the
33-letter Russian alphabet (index 32) is exact under JS bit ops.
- validate.ts: export connected for the generator's connectivity filter.
- backend/cmd/movegen: dev tool building small sample dictionaries and emitting
golden move-generation fixtures from the real Go solver (EN + RU).
- tests: dawg.cursor.test.ts (enumeration bijection vs indexOf) and
generate.parity.test.ts (7/7 vs the Go solver: empty board, mid-game, blank,
single-word rule, Russian index-32 cross-set). The committed EN sample also
unblocks the existing skipped dawg.parity.test.ts once wired with DICT_* in CI.
Pure additive library code; no runtime behavior change.
The edge request-latency histogram records seconds but used the OTel SDK's
default millisecond-calibrated bucket boundaries (first boundary 5), so every
sub-5s request fell into one bucket and histogram_quantile(0.99) interpolated
to ~4.95s for every message type — regardless of the real (millisecond)
latency. That tripped the >1s "Gateway request latency p99 high" alert on
essentially every request, flapping fire/resolve on each app open (seen on the
test and prod contours).
Set explicit second-scale bucket boundaries (0.005 … 10s) straddling the 1s SLO
so the p99 reflects real latency and the alert fires only on genuine slowness.
Regression test asserts the histogram carries a sub-second boundary.
Two email changes, per the owner:
1. Code-only login from an installed PWA. A login requested while running as a
standalone PWA now omits the one-tap confirm link from the email — the link
would open in a separate browser whose minted session cannot reach the PWA,
stranding the login. The code is typed in the same window instead. The client
sends a `pwa` flag (isStandalone) on the email-code request (a new FBS field,
threaded through the gateway); the backend omits the deeplink when it is set,
reusing the existing deletion-code link-omission path. Non-PWA browser logins
keep the one-tap link. No polling, no migration.
2. Security: the operator alert digest no longer embeds the admin-console (/_gm)
URL. An admin link must never travel in an email, where a mail provider could
cache or index it; the operator opens the console directly.
Tests: an inttest asserting the PWA login email omits the link (and a browser one
keeps it); a codec round-trip of the new pwa field; the alert-digest test flipped
to guard the admin link is absent. Docs: ARCHITECTURE + FUNCTIONAL (+ru).
Make the web SPA an installable PWA and surface it to users:
- manifest.webmanifest + an install-only service worker + 192/512/maskable
icons (ui/public); PWA head tags in index.html; the .webmanifest MIME type
registered in the gateway (the distroless image has no /etc/mime.types).
- Platform-adaptive install CTA (components/InstallApp.svelte + lib/pwa):
one-tap on Chromium, manual Add-to-Home-Screen instructions on iOS Safari,
hidden elsewhere / once installed / inside a Mini App. Shown under the
logged-out login card and at the bottom of Settings.
- The landing gains a third entry linking /app/ (the brand tile), with a
caption under all three (Telegram / VK / Веб-версия).
The service worker is navigation-only (network-first, cached-shell fallback);
hashed assets and the Connect stream are untouched. It exists to satisfy
Chromium's installability requirement and is the single growth point for a
future opt-in offline mode.
Tests: pwa.ts unit tests; a webui probe (manifest/sw.js content-type + the
SPA-fallback-to-HTML trap); e2e for the one-tap CTA, the iOS instructions
modal and the landing web entry. Docs: ARCHITECTURE §13, FUNCTIONAL (+ru),
gateway README.
Non-default campaigns gain an optional colour override (background / text /
link) in two sets — one for every theme, one for the dark theme only — and an
"urgent" flag.
- Colours ride profile.get as six trailing FlatBuffers strings on
BannerCampaign (backward-compatible). The client resolves the cascade
(dark <- dark ?? all, light <- all) per rendered theme and derives the strip
border from the background in JS (no CSS color-mix, for the old Android
WebView floor); AdBanner applies them as inline vars scoped to the strip.
- Urgent is resolved entirely server-side: while any enabled, in-window urgent
campaign exists, computeActiveSet returns only the urgent campaigns and
bannerFor skips the eligibility gate — so a system notice reaches every viewer
(paid / hint-holding / no_banner included) and preempts the ordinary feed. No
wire field; it appears on each viewer's next profile.get.
- Admin console (/_gm/banners): native colour pickers + a live light/dark
preview of the strip, and an urgent toggle. The default campaign stays plain,
enforced by the service and a DB CHECK.
Migration 00009 is additive (nullable colour columns + a bool default +
all-or-nothing / hex / default-plain CHECKs) — expand-contract, rollback-safe.
Docs: ARCHITECTURE §10, UI_DESIGN, FUNCTIONAL (+ru). Tests: ads unit (urgent
preempt + colour validation), codec + resolver unit, gateway transcode, and
integration (colour round-trip + urgent bypass against real Postgres).
The info toast began drifting up and fading the instant it finished
appearing (the CSS keyframes went straight from the 12% appeared-stop to
the 100% risen-and-faded stop), so a glanced message was already leaving.
Insert a ~1s hold at full opacity/rest before the rise-and-fade: extend the
animation 2s → 3s with keyframe stops at 8% (appeared, ~240ms) and 41% (end
of the ~1s hold), keeping the original rise-and-fade pace for the tail. The
reduced-motion variant gets the same appear/hold/fade timing (fade only, no
travel). The showToast dismissal timer is bumped 2000 → 3000ms to stay in
lockstep with the animation (the error toast's 4s dwell is unchanged).
An email link/code can be confirmed out of band: the recipient taps the
one-tap link in the email, which confirms in another browser/session. The
backend already publishes a `notify` `profile` re-fetch signal for this
(handlers_auth.go handleEmailConfirmLink), and the client re-fetches on it.
But the live stream is single-shot with no replay: a Mini App backgrounded
while the user is in their mail app tapping the link drops the stream and
misses the event (the gateway hub has no subscriber to deliver to), and the
reconnect on foreground does not re-sync — so the open code form stayed
until a manual reload. On Telegram Desktop the app is never backgrounded,
so the push works and there was no bug.
Add a client-side fallback: while an add-email confirmation is pending
(a code was sent, no email yet), poll `profile.get` on a 4s interval and on
foreground regain until the address lands; the effect stops as soon as the
email appears. The live push still updates instantly when foregrounded —
this only covers the backgrounded-miss gap.
Tests: a mock e2e attaches the email WITHOUT emitting a live event (new
window.__mock.clearEmail / confirmEmailOutOfBand seams), so it exercises the
poll, not the push, and asserts the code form collapses into the email row.
Docs: ARCHITECTURE.md §10 notes the single-shot gap + the poll fallback.
Inside a Telegram/VK Mini App the host provider is auto-linked. Once the
player also linked an email, `canUnlink` turned true and the "Unlink"
control appeared on the host platform's own row — letting them unlink the
very platform they are signed in through, which is meaningless.
Gate the Telegram row on `!insideTelegram()` and the VK row on
`!insideVK()`, reusing the runtime host detectors that already gate the
"link" buttons. Symmetric: inside TG only the TG row is hidden (the VK row
still shows, since VK is not the current host), and vice versa. The web and
native builds are unchanged (both detectors are false there); the backend
is untouched — this is a UI display gate, and `linkUnlink` still refuses to
remove the last identity.
Docs: FUNCTIONAL.md (+_ru mirror).
The index.html boot guard now fires a fire-and-forget beacon (POST /telemetry/unsupported)
when it turns a client away on the unsupported-engine screen, so the owner can see — on the
Users dashboard beside "app opens" — how many real clients hit it and on which engines.
- Client: navigator.sendBeacon (fetch fallback) with a localStorage dedup keyed by app version
+ reason + Chromium, so a user reopening the app is one report, not ten.
- Gateway: a new unauthenticated POST /telemetry/unsupported handler (the client never booted,
so it carries no session), per-IP public-limited and body-capped, mirroring the export-download
route. It folds the beacon into the OTel counter unsupported_engine_total {reason =
no_bigint/no_proxy/boot_error/other, chromium}, with reason allow-listed and the Chromium major
reduced to a bounded range (normalizeUnsupported) so a spoofed beacon cannot inflate the metric
cardinality; the full user agent is logged, not labelled.
- Caddy: /telemetry/* added to the @gateway matcher (else it falls to the landing catch-all).
- Grafana: two panels on the Scrabble — Users dashboard (by reason, by Chromium major).
- Docs: ARCHITECTURE.md §11.
Tests: recordUnsupportedEngine (metric split), normalizeUnsupported (cardinality bounding), and
the handler route end to end (204 / 405 / counter). go build+vet+test and gofmt clean; ui
check/build/e2e green; the client beacon + dedup verified against a forced hard-gate (BigInt
removed) — one POST, deduped on reopen, correct reason/chromium/version labels.
Old Android System WebViews (the Telegram/VK in-app browser; a real device with Google
Play auto-updates the WebView, but emulators / no-Play / restricted devices stay frozen)
showed a white screen. Two causes, fixed in order:
- build.target es2022 shipped ?./?? verbatim, which the old engine cannot parse. Lowered
to es2019 so esbuild down-levels the syntax.
- The es2020+ runtime globals the bundle and its deps call (globalThis, structuredClone,
Array.at, ...) are missing on those engines. A conditional core-js loader (emitPolyfills
writes dist/polyfills.js, document.write'd by an index.html gate only when needed) covers
them; modern engines download nothing, so the bundle-size budget is untouched.
BigInt (the 64-bit FlatBuffers timestamp decode) and Proxy (Svelte 5 runes) cannot be
polyfilled, so the effective floor is Chrome 67. Below it, a permanent ES5 boot guard in
index.html shows a friendly "this device's OS or browser can't run the app" screen (with
the web-version link inside a Mini App) and a "Diagnostic information" view with a Copy
button, instead of a white screen. A reactive net raises the same screen on an uncaught
boot error that never signals window.__booted.
Board tile glyphs used container-query units (cqw, Chrome 105+) under .cell font-size:0,
so they collapsed to 0px on Chrome 74 (invisible letters); added vmin fallbacks.
The unsupported-engine telemetry beacon is a separate follow-up PR.
The first v1.8.0 prod-deploy build job failed: `docker compose build` interpolates the
WHOLE compose file, and the backend's :?-guarded EXPORT_SIGN_KEY — added with the
finished-game export after the last prod release (v1.7.0), so the prod build never
exercised it — was absent from the build job's env, tripping the guard before any image
built. Prod was untouched (build-only failure: no image pushed, no deploy, no migration).
Add EXPORT_SIGN_KEY (audited every :?-guarded compose var against the build job env — it
was the only genuinely-missing one; TELEGRAM_MINIAPP_URL is derived in the run step).
Verified by reproducing the build-job env + `docker compose config`.
The maintenance overlay never showed on the test contour: only prod-deploy.sh raised
the caddy maintenance flag, so a test redeploy was a bare gateway/backend recreate — the
SPA saw a transient 502 ("Reconnecting…"), never the 503 + X-Scrabble-Maintenance marker
the overlay keys on. So the feature (PR #171) was unverifiable off prod.
Raise the flag around the recreate in ci.yaml's deploy job too, mirroring prod. Ordering
matters because of the config reseed: ci does `rm -rf $conf` + recreate, so the running
caddy sits on the stale pre-reseed bind mount and cannot see a flag written to the new
dir until it is recreated. So: raise the flag, force-recreate caddy FIRST (onto the fresh
mount, where it carries the flag and 503s), then recreate gateway/backend (the window),
then the other config services, then lower it. A trap clears the flag if the step fails,
and the next deploy's reseed wipes a stale one — so the contour can't stick in maintenance.
The caddy-routed probes run in the next step, after the flag is lowered.
Prod was already correct (prod-deploy.sh); this only makes the test contour faithful so the
overlay + reload can be seen there. An open SPA must already be on the PR-#171 bundle (which
carries the overlay code) — reload once to bootstrap onto it.
The deploy that ends a maintenance window may ship a client-incompatible change
(wire/schema bump), and the in-session bundle is the old one. On recovery, reload to
pick up the fresh client instead of just hiding the overlay. Ordering is safe: the edge
maintenance flag (deploy/prod-deploy.sh) spans the WHOLE roll and clears only at script
exit — after the gateway (which serves the embedded SPA) has rolled — so the edge 503s
everything until the entire deploy is live; recovery therefore always serves the new SPA.
A window.__maint.recover() hook + e2e cover the reload.
The edge caddy 503 carries X-Scrabble-Maintenance during a deploy window, but a user
already in the running SPA only sees their calls start failing — the static caddy page
catches only a fresh load. Detect that marker (strictly — not any transient
'unavailable', which the Connecting indicator already covers) on the raw ConnectError at
the two transport catch sites (unary exec + the live subscribe stream), before
toGatewayError discards the response headers, and raise a non-dismissable dimmed overlay
that mirrors the static caddy page. It self-clears: a capped-backoff poll (a cheap read,
mirroring connection.svelte.ts) lifts it on the first success, and a manual "retry"
button forces an immediate re-check — so it can never get stuck. On detection the read
retry loop fails fast, so a window doesn't burn the retry budget on every call.
- pure detector maintenance.ts (maintenanceRetryMs / parseRetryAfterMs) + unit tests;
the store + self-clearing poll in maintenance.svelte.ts (mirrors connection.svelte.ts)
- MaintenanceOverlay.svelte (clones Splash's fixed/inset/dimmed shell, non-dismissable),
mounted app-global in App.svelte after Coachmark
- transport.ts detects + reports at both catch sites, clears on any successful read
- i18n RU "Технические работы" / EN "Under maintenance"; a window.__maint mock hook
(the mock can't emit a real 503) + a Playwright spec
Prod is same-origin (VITE_GATEWAY_URL empty) so the marker header is readable without a
CORS expose-header. Verified: pnpm check (0), unit (402), build, e2e (186, Chromium+WebKit).
Two prod-deploy hardening measures (owner-requested):
- Swap file (Ansible common role, swap_size=1G, vm.swappiness=10). The per-container
memory caps enforce that one service can't eat all RAM, but they overcommit the
1.9 GiB main host (~2.8 GiB of caps), so a simultaneous spike could hit the kernel
OOM-killer (and it might pick postgres). A small swap absorbs the overshoot.
Idempotent, builtin-only (no ansible.posix).
- Edge maintenance page. prod-deploy.sh raises a flag around the rolling swap /
migration window that the caddy edge serves a static 503 "технические работы" page
from (deploy/caddy/maintenance.html), for the user-facing routes only — /_gm
(Grafana) stays reachable. Cleared on any exit (success, health failure + rollback,
or error) by a shell trap so it can never stick on. The 503 carries Retry-After +
an X-Scrabble-Maintenance marker so a follow-up SPA overlay can tell a planned
window apart from a transient error (the static page only catches a fresh load; an
in-session user needs the app-side overlay). Not zero-downtime — the single
stateful backend still blips — but the window is graceful instead of raw 502s.
Verified: caddy validate; the gate 503s every non-/_gm path incl. the Connect/gRPC
edge and serves the page + markers; /_gm bypasses; toggling needs no reload
(per-request stat). Ansible --syntax-check + compose config (base+prod) pass.
The planted honeytoken bearer trap already existed in the gateway + compose, but
no workflow fed GATEWAY_HONEYTOKEN, so it was always empty (inert). Wire the
per-contour TEST_/PROD_GATEWAY_HONEYTOKEN secret into the ci deploy, prod-deploy
and prod-rollback env (the prod path via the shared deploy/write-prod-env.sh),
document it (deploy/README + .env.example), and fix the stale compose comment.
Empty secret = trap off (no ":?" guard), so a deploy is safe before the operator
sets the value + plants the bait. On prod (IP ban on) presenting it earns a 24h
ban + alarm; on test (ban off) it logs + a ban metric.
GF_SMTP_ENABLED is enabled separately via the TEST_/PROD_GF_SMTP_ENABLED Gitea
variables (=true) — no code change.
Collapse identical TEST_/PROD_ pairs to single unprefixed Gitea entries, derive
the public URLs from PUBLIC_BASE_URL at deploy time, and share the prod env.sh
renderer between deploy and rollback.
- Collapse to one unprefixed variable: DICT_VERSION, SMTP_RELAY_HOST/PORT/TLS,
GRAFANA_SMTP_PORT, VITE_VK_APP_LINK, VITE_VK_APP_ID (one Selectel relay + one
pair of VK apps serve every contour). Secrets collapsed by the owner:
SMTP_RELAY_USER/PASS, GATEWAY_VK_APP_SECRET, GATEWAY_VK_ID_CLIENT_SECRET.
- Derive at deploy from PUBLIC_BASE_URL (no longer stored): TELEGRAM_MINIAPP_URL,
GRAFANA_ROOT_URL, VITE_VK_ID_REDIRECT_URL. Removes the prod/test asymmetry and
fills the missing test VITE_VK_APP_ID (VK web login was half-configured on test).
- Extract deploy/write-prod-env.sh + write-prod-bot-env.sh, shared by prod-deploy
and prod-rollback so the two cannot drift: a rollback now re-renders the FULL
runtime env (email / VK login / Grafana alerts previously went dark after a
rollback) and passes TELEGRAM_SUPPORT_CHAT_ID.
- Single-source DICT_VERSION (CI env + both deploys), fix the v1.3.0/v1.3.1 drift in
.env.example/README, correct the misleading honeytoken/abuse-ban compose comment,
and rewrite the deploy/README variable list (+ the previously undocumented
GATEWAY_VK_APP_SECRET and TELEGRAM_SUPPORT_CHAT_ID).
The Gitea variables are reworked via the API; the stale TEST_/PROD_ entries are
deleted after the test contour goes green.
An account merge blanket-reassigned all of the secondary's identities to the primary,
so merging two accounts that each had a confirmed email (or telegram/vk) left the
survivor with two identities of one kind — which the profile and the retention dossier
both treat as singular (the profile showed an arbitrary one; a later change-email
journaled only one). The merge now keeps the primary's identity and journals the
secondary's colliding one to retained_identities (reason=merge) before dropping it, so
the survivor has one identity per kind and the absorbed credential still lands in the
legal dossier.
- migration 00008: widen retained_identities.reason CHECK to admit 'merge'
(expand-contract — Up only widens the set, so an image rollback stays DB-safe).
- accountmerge: dedupeIdentities + retainMergedIdentity before the identity reassign.
- inttest TestAccountMergeDedupesEmail; docs ARCHITECTURE §4 + retention.go reason note.
A browser has no signed VK Mini App launch params, so linking VK on the web uses
VK ID's raw OAuth 2.1 flow (PKCE, no @vkid/sdk): the SPA redirects to VK's hosted
login and returns with an authorization code, which the gateway exchanges
server-side (confidential, under the VK "Web" app's protected key) for the trusted
vk user id — then the existing link/merge machinery attaches or merges it.
- fbs LinkVKRequest{code, device_id, code_verifier}; codec + TS bindings.
- backend link.Service ConfirmVK/MergeVK/attachVK (KindVK, mirror Telegram),
handleLinkVK[Merge], routes /user/link/vk[/merge], backendclient LinkVK[Merge].
- gateway internal/vkid confidential code exchange (id.vk.com/oauth2/auth);
transcode link.vk.confirm/merge (registered only when configured) + config
GATEWAY_VK_ID_{APP_ID,CLIENT_SECRET,REDIRECT_URL} + main wiring.
- UI lib/vkid (PKCE authorize redirect + callback), Profile "Link VK" control,
boot callback handling; a merge re-authorizes for a fresh code (VK codes are
single-use). Web-only (a redirect strands a Mini App webview).
- Deploy: VITE_VK_APP_ID + VITE_VK_ID_REDIRECT_URL build args + gateway env,
ci.yaml/prod-deploy TEST_/PROD_ vars, compose/Dockerfile/.env.example/README.
- Tests: vkid exchange unit (string/number user_id, id_token fallback, errors),
transcode link.vk, backend ConfirmVK/MergeVK inttest, codec encodeLinkVK.
- Docs: ARCHITECTURE §4, FUNCTIONAL(+ru), gateway README.
Two contour-deploy failures from PR4: (1) the deploy did not stage deploy/blackbox/, so
blackbox_exporter crash-looped on a missing config mount — add blackbox to the ci.yaml cp
and the prod-deploy tar; (2) Grafana rejects the 'Name <addr>' From form the backend go-mail
accepts and validates it even when SMTP is disabled, crash-looping Grafana — the deploy now
splits SMTP_RELAY_SERVICE_FROM into a bare GF_SMTP_FROM_ADDRESS + GRAFANA_SMTP_FROM_NAME
(handles both display and bare forms). Verified the sed split + compose render.
Drop the separate GRAFANA_SMTP_HOST (it duplicated SMTP_RELAY_HOST): Grafana differs from
the backend only in needing the STARTTLS port, so GF_SMTP_HOST is composed from
SMTP_RELAY_HOST + GRAFANA_SMTP_PORT (an explicit per-contour var, no magic default), reusing
SMTP_RELAY_USER/PASS. Wired through ci/prod/.env.example/README.
§11 gains the alerting layer: Grafana infra rules (scrape-down, edge error-rate/p99, host
mem/disk/cpu, postgres connections, TLS cert < 20d via blackbox), noDataState=OK, and the
backend admin-alert worker (coalesced email on new feedback / complaints).
Grafana: GF_SMTP from the shared relay (STARTTLS host:port) + alerting provisioning
(contact point → SERVICE_EMAIL, route-all policy, and rules for scrape-target down,
gateway internal-error rate + p99 latency, host mem/disk/cpu, postgres connections, and
TLS cert < 20 days). All rules noDataState=OK so an absent metric never false-alerts.
blackbox_exporter probes the edge caddy's TLS (probe_ssl_earliest_cert_expiry) — effective
on prod (caddy terminates TLS; contour caddy is HTTP-only so the metric is absent).
Wires the new env through compose (backend admin From/To, Grafana SMTP), ci.yaml (TEST_),
prod-deploy (PROD_ + env.sh), .env.example and the README var table.
New adminalert worker polls for feedback + word complaints arriving since the last check
and coalesces a burst into one digest email per interval (5 min), inert unless a distinct
admin sender (BACKEND_SMTP_ADMIN_FROM) and recipient (BACKEND_ADMIN_EMAIL, comma-separated
allowed) are configured. The mailer gains a per-message From override and splits a
comma-separated To into separate recipients (go-mail needs them as a list). Feedback/game
stores gain CountSince/CountComplaintsSince. Unit tests cover the digest, the skip-when-
empty, and the recipient split.
The email/name/external-id search was scoped to one tab and only looked in the live
identities table, so a deleted account (whose credentials moved to retained_identities on
deletion) could not be found by the email/id it held. Now a people search spans live and
deleted accounts in one query (robots only on the Robots tab) and also matches the
retention journal and the retained real name; results carry a 'deleted' badge. Integration
test: a deleted account is found by its held email and external id.
- Admin /users gains a Deleted scope (between People and Robots); every other scope now
hides tombstoned accounts (deleted_at filter in UserFilter).
- Admin delete-user is offered for any non-deleted account (drop the not-guest gate, so a
stale is_guest account can still be tombstoned).
- Harden EmailService.ConfirmCode to ClearGuest — defence-in-depth so no confirmed-email
path leaves is_guest set (the currently-live paths already do).
- Space the delete/change dialog's action row from its input field.
Integration tests: the Deleted filter scoping + ConfirmCode guest promotion.
FUNCTIONAL (+ru): the deletion user story — anonymised live surfaces ([Deleted]),
freed sign-in methods, a two-year-retained dossier, the code/phrase step-up, game
forfeit + all-robot drop, fresh account on reopen. ARCHITECTURE: the retention model
(retained_identities journal on every detach, tombstone + [Deleted] sentinel, drop
all-robot games, purpose=delete step-up, cold-load last-login, two-year TTL reaper).
The user-detail console gains a Deletion & retention panel: last login (time + IP),
the tombstone (deleted-at + retained real name), and the retention journal (the legal
dossier of detached credentials). A Delete-user action runs the same deletion
orchestration as the in-app flow (mirrors the email-erase pattern). Store readers
RetainedIdentities + DeletionInfo back the view; integration test covers them.
Profile gains a Delete-account control (durable accounts) opening a step-up dialog: a
mailed code for an email account, or the typed DELETE phrase for a platform-only one.
On success the app swaps to a terminal AccountDeleted screen ('Учётная запись удалена')
with a Close that closes the host Mini App (telegramClose / vkClose; web = no close).
Wires deleteRequest/deleteConfirm through client/transport/mock/codec; ru/en i18n;
codec wire test + Chromium/WebKit e2e.
Step-up: email accounts confirm with a mailed code (purpose=delete, no deeplink —
ConfirmByToken refuses a delete token so a stray click can't delete); platform-only
accounts type a fixed phrase (anti-impulse). Endpoints /user/delete/{request,confirm};
the confirm orchestration resigns active games, drops all-robot games, tombstones +
anonymizes the account (freeing its creds), and revokes its sessions — the tombstone is
the point of no return, the rest best-effort. Gateway account.delete.{request,confirm}
ops + fbs AccountDeleteConfirm/AccountDeleteRequestResult + branded ru/en delete email.
Integration tests cover the step-up (code + no-email) and the orchestration pieces.
Q2=B: DropAllRobotGames deletes the deletee's games with no human opponent (vs-AI or
auto-match-robot; children cascade), keeping games with any human seat (anonymized
instead). Q1=A: the anon label is the sentinel [Deleted] — the editable-name rule forbids
brackets, so no live player can impersonate a deleted account. Integration test covers
drop-vs-keep.
Daily background reaper purges the deletion dossier past its two-year TTL: every
retained_identities row by detached_at (covering unlink/change on live accounts too),
and — for accounts tombstoned before the cutoff — the retained feedback thread plus the
dossier PII (deleted_display_name, last_login_ip). Chat is kept (a shared game artifact)
and the tombstone row stays. Started from main next to the guest reaper. Integration
test covers the cutoff boundary and the deleted-account feedback/PII purge.
New accountdelete package: AnonymizeAndTombstone journals every credential to the
retention log (reason=delete) then removes them (freeing email/vk/tg for reuse),
snapshots the real display name into deleted_display_name and scrubs the live one to
'Удалённый игрок', sets deleted_at, anonymizes the account's game-seat snapshots, and
drops its friendships/blocks/invitations/friend-codes/drafts/pending-codes — all in one
transaction. Chat, feedback and complaints are kept (the tombstone keeps their
no-cascade FKs valid). Session revocation + game forfeit are orchestrated a layer up.
Integration test covers journalling, tombstone/scrub and credential reuse.
The profile GET (fetched once per cold app-load by the SPA) stamps accounts
.last_login_at/.last_login_ip, throttled to at most once per hour per account
(best-effort, never blocks the read). IP from the gateway-forwarded X-Forwarded-For.
Feeds the account-deletion dossier. Integration test covers the throttle.
On unlink (RemoveIdentity, reason=unlink) and email change (replaceEmailIdentity,
reason=change) write the outgoing credential to retained_identities before removing
the live identities row — so the legal dossier survives while the (kind, external_id)
frees for reuse. Same transaction, so the dossier and live state cannot diverge.
Integration tests cover both reasons.
Additive/expand-contract: new append-only retained_identities table (the legal
dossier of every detached credential, keyed by account_id, TTL'd by detached_at)
plus nullable accounts columns last_login_at/last_login_ip (cold-load stamp) and
deleted_at/deleted_display_name (tombstone + retained real name). Regenerates the
go-jet models for accounts + retained_identities only.
Replace the stale skipped linking specs: change-email refuses a taken address with
the non-disclosing message (never revealing the other account) and replaces a free
one; the web Telegram link control links then unlinks through the confirm dialog.
Chromium + WebKit.
FUNCTIONAL (+ru): drop the stale 'linking UI hidden' note; document the profile
sign-in-methods matrix, the unlink last-identity guard, and the non-disclosing
atomic email change. ARCHITECTURE: unlink + change-email behaviour and the
purpose=change deeplink branch.
Integration: unlinking a provider keeps the other identities and refuses removing
the last one; change-email replaces the address (freeing the old), refuses a taken
address without merging, and works through the one-tap deeplink token. Unit: the
change-email template renders localised ru/en copy.
Profile shows the account's sign-in methods for guests and durable accounts alike:
add or change email, link Telegram on the web (login widget), and unlink a linked
provider. Email is never unlinked (it is changed); unlink is offered only when
another identity remains, and a change to an address owned by another account shows
the non-disclosing 'check the address or contact support'. Add-VK-on-web stays
deferred (no VK OAuth).
Wires linkUnlink / changeEmailRequest / changeEmailConfirm through the client,
transport, mock and codec (encodeLinkUnlink + LinkResult 'unlinked'/'changed'
statuses); codec wire tests; ru/en i18n.
Unlink: POST /user/link/unlink (telegram|vk) via account.RemoveIdentity, refusing
the last identity; email is never unlinked. New fbs LinkUnlinkRequest + gateway
link.unlink op, returning the refreshed profile.
Change-email: purposeChange confirm-codes (RequestChangeCode/ConfirmChange) that
atomically replace the account's confirmed email (account.replaceEmailIdentity);
a new address owned by another account is refused without disclosure, never merged.
The one-tap deeplink handles purposeChange too. Reuses the LinkEmail* fbs tables;
gateway link.email.change.{request,confirm} ops + backendclient methods; branded
ru/en change-email copy.
Generalize the email-erase store op to any identity kind (with the same
last-identity guard and, for email, the pending-confirmation cleanup) so the
profile Unlink control can reuse it for Telegram/VK. RemoveEmailIdentity stays as a
thin wrapper for the admin console.
Add email / telegram_linked / vk_linked to the Profile (fbs table + regenerated
Go/TS bindings, gateway ProfileResp + encodeProfile, backend DTO, UI model +
decode). They are filled outside the pure projection — Server.profileResponse now
reads the account's identities (like the banner seam) — and will drive the profile's
Add / Unlink / change-email controls.
The session-less /confirm page defaulted to English, so it showed the English
app title. Carry the recipient's language on the deeplink (?lang) and setLocale on
load so the page matches the email. Show a localised brand wordmark (Эрудит / Erudit,
matching the email) on the success state only; the invalid/expired state now shows
just the message, no header.
Add an exact (strict) email filter to the /users list (UserFilter.EmailExact →
a kind='email' identity match) with a search input, and an 'Erase email' action on
the user card that deletes the bound email identity and its pending confirmations,
freeing the address. It refuses to remove the account's only identity
(ErrLastIdentity), which would leave it unreachable. Integration tests for both.
Drop the confirm button: the /confirm screen confirms the token as soon as it
loads. The token rides the URL fragment (never sent to the server), so a plain link
prefetch cannot reach it, and the manual six-digit code is the fallback if an
aggressive scanner runs the page. Update the ARCHITECTURE/FUNCTIONAL wording
accordingly and swap the confirm.prompt/action strings for confirm.busy (en+ru).
ConfirmByToken attached the confirmed email to the account but, on the link path,
skipped ClearGuest — so a guest who bound an email via the one-tap deeplink stayed a
guest (the code-based flow clears it in the link service). Clear the flag on a free
link too, promoting the guest to a durable account; the profile live event then
refreshes the open session. Integration test added.
An email account was provisioned with no display name (unlike Telegram/VK, which
seed one), so an email login showed an empty name. Seed it from the email's local
part (before '@', trimmed and capped to the column width) on first contact; the
user can rename it later. Only new accounts are seeded — an existing account's name
is never overwritten.
Document the one-tap confirm deeplink in ARCHITECTURE (§4 the login magic-link /
link confirm+profile-refresh, prefetch-safe token) and the new notify 'profile'
sub-kind (§10), and add the one-tap link to the FUNCTIONAL email story (+ru). Add
codec round-trip assertions for the EmailRequest language field and
encodeEmailConfirmLink (the mock e2e bypasses the codec).
Add the /confirm/<token> SPA route and Confirm screen: a prefetch-safe button
POSTs the token via a new confirmEmailLink RPC (client/transport/mock/codec +
ConfirmLinkResult model). A login adopts the minted session and enters the app; a
link shows confirmed / merge-in-the-app; an invalid or expired token asks for a new
code. Exempt /confirm from the no-session /login redirect. Forward the client
locale on the email request (authEmailRequest gains language → app.locale) so a
fresh web login email is localised. Handle the new 'profile' live-event sub-kind by
re-fetching the profile, so a link confirmed in another browser reflects in-app at
once. i18n en+ru.
Add the confirm-link edge method (auth.email.confirm_link): a new
EmailConfirmLinkRequest/Result fbs table, the transcode const + handler + encoder,
and the backend-client call to the existing /sessions/email/confirm-link endpoint —
it rides Execute under the existing service prefix, so no proto/Caddy change. Add a
language field to EmailRequestRequest and forward it (the backend already seeds it).
Add the NotifyProfile sub-kind + notify.ProfileChanged, published by the confirm-link
handler on a successful link so an in-app session re-fetches its profile when the
email was confirmed in another browser. Regenerated fbs bindings (Go + TS).
Add POST /internal/sessions/email/confirm-link: it verifies a deeplink token via
ConfirmByToken and, for a login, mints a session (the deeplink page signs in with
it); for a link, attaches the confirmed email and reports "confirmed" or
"merge_required" (the app drives the interactive merge). The token, not a request
session, is the authorization. Add LinkConfirmation.IsLogin() and integration tests
for the login, link and merge branches (the token is read from the mailed link).
The gateway RPC, live event and SPA route follow.
Re-apply the deeplink backend deferred out of PR1a: migration 00006 adds
email_confirmations.purpose + link_token_hash (hand-edited jet), each issued code
now carries an opaque 256-bit token (only its SHA-256 stored), and ConfirmByToken
resolves a token to a login (confirm + clear guest) or a link (attach when free,
signal merge when owned elsewhere). issueCode now embeds the /app/#/confirm/<token>
link in the email. The confirm endpoint, gateway RPC and SPA route follow.
The 465-only implicit-TLS heuristic mis-classified Selectel's SSL port (1127),
which the mailer would have dialled with STARTTLS and failed. Add BACKEND_SMTP_TLS
(ssl|starttls) — empty still derives the mode from the port (implicit on 465, else
STARTTLS) — and dial implicit TLS with WithSSL()+WithPort so any port works, not
just 465. Wire SMTP_RELAY_TLS through compose/ci/prod/.env.example and document it
(Selectel: 1127 = SSL, 1126 = STARTTLS). Unit-tested.
Un-hide the Profile email box for guest accounts (hidden={!p.isGuest}): a guest
binds an email to register / sign in, and a returning address opens the existing
merge dialog. Provider linking stays hidden — the Telegram control keeps its
wiring behind a hidden attribute — until the non-guest linking matrix (PR2). The
two linking e2e specs remain skipped (they assume a non-guest login and the
visible Telegram control); update their stale comments.
Add an integration test asserting the guest-until-confirmed squat fix (an
email-login account is a reapable guest until the code is confirmed, then
durable). Document the branded relay pipeline in ARCHITECTURE (go-mail, TLS by
port, no client cert, PUBLIC_BASE_URL anti-injection, per-recipient send throttle,
guest-until-confirmed), FUNCTIONAL (+ru), TESTING and the backend README config
table (BACKEND_SMTP_* + BACKEND_PUBLIC_BASE_URL).
Add the backend confirm-code relay env to compose (BACKEND_SMTP_HOST/PORT/
USERNAME/PASSWORD/FROM + BACKEND_PUBLIC_BASE_URL), sourced from SMTP_RELAY_* /
SMTP_RELAY_FROM / PUBLIC_BASE_URL. An empty host keeps the backend on the log
mailer so a contour without relay credentials still boots. Port defaults to 465
(implicit TLS). Map the TEST_ set in ci.yaml and the PROD_ set in prod-deploy.yaml
(both the deploy-main env and the env.sh heredoc). Document the six variables in
.env.example and the deploy README (secrets: user/pass; variables: host/port/from
+ the per-contour PUBLIC_BASE_URL, required whenever the relay host is set).
Add an in-memory SendLimiter enforcing a one-per-minute cooldown and a
five-per-rolling-hour cap per recipient address, checked before provisioning or
sending in RequestCode, RequestLoginCode and RequestLinkCode. It guards against
email bombing and protects the relay quota. The limiter is injected in main
(nil in tests, so the domain suite is unaffected); ErrTooManyRequests maps to
HTTP 429.
Swap the net/smtp mailer for go-mail behind the existing Mailer seam: the
Message struct now carries a text + HTML body, TLS mode is chosen from the port
(implicit TLS on 465, else mandatory STARTTLS), a dial timeout bounds the
synchronous send, and no client certificate is needed. Add a branded, image-free,
mobile-friendly ru/en HTML template (with a plain-text alternative) rendering a
large readable code and an ignore-notice footer with a landing link.
Add BACKEND_PUBLIC_BASE_URL config (the canonical origin for the email footer
link, never the request Host — anti-injection), required when a relay is
configured.
Fix the email-login address squat: ProvisionEmail creates the account flagged
is_guest until the code is confirmed, so an abandoned login is reaped like any
guest and its address freed; confirming (login or link) clears the flag. Seed the
new account's language from the client, plumbed through the email-login request.
The confirm deeplink, its transport surface and the send rate-limit land in
follow-up work.
The finished-game export (GCG + a new PNG of the final position) is one
signed, short-lived relative URL (game.export_url; HMAC-SHA256, 10-min
TTL, BACKEND_EXPORT_SIGN_KEY) resolved against the client's own origin
and delivered by the best affordance each platform has (five on-device
review rounds):
- TG Android/desktop: native showPopup chooser -> native downloadFile
dialog (bridge-only chain, activation-safe).
- TG iOS: app-modal chooser -> OS share sheet with the fetched file
(a popup callback cannot supply the activation the sheet needs).
- VK iOS: VKWebAppDownloadFile for both formats.
- VK Android: the PNG opens in VK's native image viewer, the GCG copies
to the clipboard (the VK Android downloader hangs on any download,
Content-Length/Range notwithstanding).
- VK desktop iframe / desktop browsers: plain anchor downloads.
- Mobile browsers: the OS share sheet (fetch-then-share).
- Legacy TG (< Bot API 8.0): app modal + GCG clipboard, no image option.
The PNG is rasterized on demand by the new internal `renderer` sidecar
(node:22-slim + skia-canvas + baked Liberation/Noto Color Emoji fonts)
executing the SAME ui/src/lib/gameimage.ts the ui project unit-tests;
the backend rebuilds the render payload from the journal +
engine.AlphabetTable, and the device date locale, IANA time zone and
localized non-play labels ride the signed URL. Nothing is stored — the
artifact re-derives from the immutable journal on each GET. The gateway
forwards /dl/* (caddy @gateway matcher extended) behind the per-IP
public rate limiter and serves bytes via http.ServeContent.
Deploy: renderer service in compose + prod overlay + rolling order +
prod push list; TEST_/PROD_EXPORT_SIGN_KEY secrets; the sidecar smoke
runs in the ui CI job. Docs: ARCHITECTURE, FUNCTIONAL(+_ru), UI_DESIGN,
TESTING, deploy/README, renderer/README.
Telegram's native showPopup delivers its callback with no user
activation, so navigator.share (TG iOS) and clipboard writes (TG
Android GCG copy) silently fail from it — the chooser is now always the
app's own modal, keeping the button click's gesture alive for the
delivery APIs.
The data:URL preview modal is dropped: the Android TG/VK long-press
menu mangles data: URLs (dead download, black-screen open, base64
clipboard garbage), so a binary PNG has no working client-side route in
those webviews at all. The image option is withheld there until the
server-rendered signed-URL delivery (Telegram downloadFile /
VKWebAppDownloadFile) lands; the plain web and mobile browsers keep it.
On-device findings by the owner on the test contour (TG iOS, TG
Android, VK Android).
The history header's export button now opens a chooser — Telegram's
native popup inside Telegram, the app's own modal elsewhere — offering
the GCG file and a new client-rendered PNG of the final position
(lib/gameimage, Canvas 2D, lazy dynamic import, zero dependencies):
light theme, classic A..O/1..15 axes, label-free premium fills, and a
fixed-typography per-seat scoresheet with GCG-style move coordinates,
multi-word sub-lines, endgame rack-settlement row, winner trophy and a
hostname + device-locale finish date footer; a long game stretches the
board, never the typography.
Delivery mirrors the GCG rules (Web Share with no blob fallback, else
download) except on Android Telegram/VK WebViews and the desktop VK
iframe, where a binary PNG has no clipboard-text fallback: those get a
preview modal with a long-press/right-click save hint and a copy-image
button where ClipboardItem exists.
Landscape fits the whole board, so the coarse-pointer auto-zoom on tile
placement, drag hover-hold and hint only hid the rest of the position.
Gate all three on portrait; manual double-tap/pinch zoom is unchanged.
New e2e lock both sides: landscape placement stays unzoomed, portrait
placement still auto-zooms (touch-emulated, both engines).
The landing now always opens in Russian (saved 🌐 choice still wins) —
browser-language detection made the indexed content nondeterministic
(Googlebot renders with en-US). landing.html gains the static Russian
SEO head: title/description, canonical pinned to the production origin,
Open Graph card (Telegram/VK link previews), twitter:card, JSON-LD,
theme-color and the favicon set; the SPA shell turns noindex and its
tab title becomes «Эрудит (Скрэббл)». New assets/icons generator
(same tile design as the VK loader) produces favicon.svg/ico,
apple-touch-icon.png and og-image.png into ui/public/, plus robots.txt.
The small 1️⃣ in the status bar's score-preview slot read as noise; the
single-word rule keeps its spelled-out label in the history header and
the lobby invitation card.
Add the VK Mini App logo next to the Telegram one on the landing hero,
linked via the new VITE_VK_APP_LINK build-arg (full URL, wired through
compose, CI and prod-deploy from TEST_/PROD_VITE_VK_APP_LINK).
Also restore the landing's Telegram link itself: commit 57c778f
collapsed the per-language vars to VITE_TELEGRAM_GAME_CHANNEL_NAME in
compose/CI but left gateway/Dockerfile with the stale _EN/_RU ARGs and
without the plain one, so the built bundle saw the var as undefined and
dead-code-eliminated the whole channel-link branch — deployed landings
(prod included) have shown no Telegram logo since.
The Android VK client's WebView ignores target=_blank and navigates the
Mini App's own window to the target, stranding the player outside the
game with no way back (the dictionary lookup, About/Feedback links, the
ad banner and the bot-link modal fallbacks). vk-bridge 3.x has no method
to open an external URL, so external links are routed through VK's own
leave-VK redirect (vk.com/away.php), which the client intercepts
natively and hands to the system browser.
onExternalLinkClick moves from lib/telegram to a new lib/links that
composes the Telegram and VK routers; iOS and desktop VK open _blank
correctly and are left alone.
Measure uptake of the client-side local move-preview accelerator (§5) so
adoption can be watched before defaulting it on: app cold starts, dictionary
loads by result (fetched / cache_hit / miss) and move previews by path
(local / network — the backend load shed).
A small best-effort client beacon (POST /metrics/local-eval, session-gated)
batches counter deltas and posts them on a 60s timer and when the app is
backgrounded — never on the gameplay path: the in-app counters are plain
in-memory increments, only the periodic flush touches the network and it is
fire-and-forget. The gateway folds each batch into three OTel counters
(local_eval_cold_start_total, local_eval_dict_load_total,
local_eval_preview_total), clamped against a spoofed inflation.
- gateway: counters + recordLocalEval + session-gated /metrics/local-eval handler
- ui: localeval-metrics accumulator/beacon; hooks in the dict loader, in
Game.recompute and in bootstrap (skipped under the mock harness)
- caddy: route /metrics/* to the gateway
- docs: ARCHITECTURE §11; Grafana "Scrabble — Users" dashboard panels
Score and validate a tentative move on-device instead of a per-arrangement network
round trip. The dawg reader and the validate/score/direction slice of the
scrabble-solver engine are ported to TypeScript (ui/src/lib/dict), pinned
byte-for-byte to the Go engine by a `conformance` CI job (full-dictionary reader
parity plus a battery of plays across every variant and both cross-word rules,
including the inferred orientation). The server stays authoritative — submit_play
re-validates — so the local result is an advisory accelerator only.
- backend: Registry.DictBytes + an authed GET /api/v1/user/dict/{variant}/{version}
(immutable) streaming the pinned per-game dawg; caddy routes /dict to the gateway.
- gateway: a session-gated /dict edge route proxying it; fetchDict on the transport.
- client: the dictionary loads on game open (low priority so it never starves the
game on a slow link; aborted at a 5s cap or when leaving the game), is cached in
IndexedDB (best-effort, self-healing on a rejected blob) and reused across
sessions; a warm-up overlay covers a cold load, then the network preview is the
fallback; a bad-connection breaker stops warming after repeated misses; the move
preview cancels its in-flight request when the tiles change.
- parity generators backend/cmd/{dictgen,validategen} + gated Vitest suites, run in
CI against the release dictionaries. A hidden debug readout lists the cached
dictionaries + breaker state, and its reset clears the cache.
- docs: ARCHITECTURE §5, TESTING, UI_DESIGN, FUNCTIONAL (+ru).
The production UI build shipped `.map` files with full `sourcesContent`, and
the gateway/landing images serve `dist/` verbatim, so anyone could fetch
`/app/assets/main-*.js.map` (same assets under `/vk/`, `/telegram/`) and
reconstruct the entire TypeScript/Svelte source at the edge.
Gate `build.sourcemap` off for `mode === 'production'` (the Docker image build).
Dev and the `mock` e2e build (`vite build --mode mock`) keep maps for debugging.
Document the posture in ARCHITECTURE.md §12.
Parity with the Telegram chrome painting: on a VK Mini App launch and on
every theme change, set VK's status-bar appearance (and, on Android, the
action/navigation bar colours) from the app's live theme tokens, so the VK
chrome matches the UI instead of clashing.
syncVKChrome mirrors syncTelegramChrome; the status-bar appearance is
derived from the --bg token's luminance (appearanceForBg, unit-tested).
Wired into the VK onScheme handler (fires on launch + on theme change) and
setTheme. Docs: UI_DESIGN VK integration.
On-device diagnostics from Android Telegram and VK confirmed both expose
no navigator.share AND no navigator.canShare, so the export fell to a Blob
<a download> that those WebViews silently ignore — nothing happened.
pickGcgDelivery is now a 3-way decision: Web Share where available (iOS),
a clipboard copy in an Android in-app WebView (Telegram/VK: no share, dead
download), else a desktop Blob download. shareOrDownloadGcg reports the
outcome so the game shows a "GCG copied" toast; the copy is VKWebAppCopyText
inside VK (which also covers the desktop VK iframe, where navigator.clipboard
is blocked) and navigator.clipboard otherwise.
Unit tests cover the 3-way choice and the copy/failed outcomes; new i18n key
game.gcgCopied (en+ru); docs ARCHITECTURE/FUNCTIONAL(+ru)/UI_DESIGN/TESTING.
The global -webkit-tap-highlight-color on #app was not enough on Android
in-app WebViews (Telegram, VK): they only suppress the selection-like tap
flash when -webkit-user-select / -webkit-tap-highlight-color /
-webkit-touch-callout sit DIRECTLY on the tapped element, not inherited.
Set them on the two tappable header controls — the title (a 10-tap debug
target) and the back chevron.
- Tap-highlight: add -webkit-tap-highlight-color: transparent on #app.
Android in-app WebViews (Telegram, VK) flashed a momentary selection-like
box on tappable nodes on tap — seen on the header title and the back
chevron; user-select (already none) does not govern it. Inherited, so
this clears it app-wide.
- VK haptics: mirror the Telegram haptic set on VK via VK Bridge taptic
(impact / notification / selection), routed through a new shared
lib/haptics.ts dispatcher. VK users previously got no haptics; the game
and error call sites now fire haptic() instead of telegramHaptic().
- VK swipe-back: disable VK's horizontal swipe-back at launch
(VKWebAppSetSwipeSettings history:false) so it does not fight the app's
own edge-swipe-back and on-board tile drag — parity with Telegram's
disabled vertical swipes; the app owns navigation via its back chevron.
Docs: UI_DESIGN (no-select / tap-highlight, VK integration).
Nudge badges lingered in the lobby on games that ended by turn-timeout,
resignation or forfeit: those paths commit the finish directly, bypassing
the move path's per-mover NudgeClearer, so the awaited seat's nudge was
never marked read. A finished game's nudges are stale, so clear them all.
Add a wired NudgeExpirer (social.ExpireNudges) called from the shared
commit finish block — covering every completion path — and from the
voidGame recovery path. It clears every seat's nudge bits for the game
and leaves chat messages unread; unlike ClearNudges it records no
publish-to-read latency, since a completion is an expiry, not a read.
An integration test reproduces the timeout case (nudge cleared, chat
kept). Docs: ARCHITECTURE §9.1, FUNCTIONAL (+_ru), backend/README.
A one-time coachmark overlay walks a new player through the lobby and their
first game board: a light dimmed layer draws one tail-pointed hint bubble at a
time, advancing on a tap anywhere and removing itself for good after the last
hint. Two independent series (lobby: settings/stats/new game; game:
header/pass-exchange/hints/shuffle/rack), gated by a per-device persisted flag
and marked done only after the last hint, so an interrupted run replays from
the start. A deep-link into Settings -> Friends still triggers the lobby series
on the first trip back to the lobby.
Targets carry a data-coach attribute, so one positioning engine anchors the
bubble in both portrait and landscape, re-measuring each frame until the
geometry settles (route slide, hidden-banner reflow, fonts). The promo banner
hides while the overlay is up (app.coachActive); a hidden DebugPanel "Reset
visited" control replays the walk-through. Off by default in the mock build so
the Playwright smoke is unaffected; ?coach forces it on for the dedicated e2e.
Pure geometry (step lists, nextVisibleStep, placeBubble) in lib/coachmark.ts
(unit-tested); Coachmark.svelte renders. Docs: FUNCTIONAL(+ru) onboarding
story, UI_DESIGN coachmark section.
A 96x96 transparent Lottie for the VK app loading screen: the «Э» tile
(score 8) drops in under gravity, lands with a soft cushion squash
(left/right edges bulge convex, corners follow via a chordal spline),
springs back and loops (~1.1s). A warm glint is caught on the bounce.
Pure 2D shapes (ddd:0) for player compatibility; ~11KB (<24KB limit).
Glyphs are real LiberationSans outlines (Arial-metric, matching the
game's font stack). Includes a reproducible, dependency-light build
pipeline (build/generate.js + extract.js + glyphs.json) and a README.
The preview GIF's green-baize background is preview-only; the asset
itself is transparent.
The #142 polish shipped three VK behaviours the docs never picked up (they were
written for the #140 auth MVP):
- the "auto" theme follows the VK client light/dark (VKWebAppUpdateConfig), not
the VK webview's prefers-color-scheme;
- the layout clears the VK mobile home bar via CSS env() max'd with the VK bridge
insets (VKWebAppUpdateInsets) — the bridge value is needed on Android, where the
VK webview exposes no env() inset (.claude said env() handled it alone);
- share/copy route through the bridge (VKWebAppShare/VKWebAppCopyText).
FUNCTIONAL (+_ru) VK paragraph gains theme+home-bar parity with the Telegram one;
UI_DESIGN gets a VK-integration note beside the Telegram one.
Two doc edits that did not land alongside their code:
- FUNCTIONAL (+_ru) claimed a friend invitation is shareable as a Telegram
*or VK* deep link "that opens it directly". On VK that is false: vkShareLink
emits a plain vk.com/app<id> link (VK forwards no payload to the Mini App),
so the recipient enters the copied code by hand (#142, commit d76f1f4 pivoted
off the earlier #f<code> plan; the doc kept the old wording).
- UI_DESIGN omitted the desktop / landscape-iframe mouse drag-to-pan of the
zoomed board added in #141 (touch scrolls the viewport natively; a mouse
cannot, so an explicit drag-to-pan handler moves it).
The scrabble-dictionary release v1.3.1 drops the obscene lexicon (mat).
Bump the centralised CI seed so unit/integration jobs validate against the
same dictionary the prod deploy now pins (vars.PROD_DICT_VERSION=v1.3.1).
The contour diagnostic confirmed VK forwards only the signed vk_* params to the iframe — a
custom payload on the app link is dropped (the '#' eaten by the vk.com SPA, a '?' query
stripped), so the friend-code cannot ride the link. The VK share link is now just
vk.com/app<id>; the recipient enters the copied code by hand (VKWebAppCopyText works). The
vkStartParam reader + bootVK routing stay as a no-op, ready for a post-moderation channel.
Removes the temporary deep-link diagnostic.
Contour review of the VK Bridge group:
- #6 invite link: VK's documented '#' direct-link payload is eaten by the vk.com SPA before it
reaches the app, so the friend-code link now carries the payload as a query param
(vk.com/app<id>?hash=f<code>); the recipient already reads the `hash` query param (vkStartParam).
(Whether VK forwards the '?' through to the iframe is being confirmed on the contour.)
- #8 Android: the VK mobile webview does not surface the home-bar inset via CSS env() (config
insets are iOS-only), so subscribe to the bridge insets (VKWebAppUpdateConfig + VKWebAppUpdateInsets)
and set --tg-safe-* to max(env(), the VK value).
- #8 landscape colour: the home-indicator strip was the (grey) page background because the two-pane
landscape game has no bottom bar. The left-panel controls bar now paints its own chrome into the
inset (Screen gains a selfInset flag that drops the shell's detached padding strip), and the
game-land runs flush to the edge.
Verified: svelte-check, 347 unit, build, bundle-gate; the landscape safe-area painting reproduced in
the mock (controls bar + board reach the edge, strip takes the bar colour). The VK-Bridge / VK launch
behaviours (Android insets, the ?hash forward) need the live contour.
Group B of the VK integration — the contour-verified follow-up to the launch+auth MVP:
- Share/copy inside the VK iframe go through VK Bridge: the friend-code invite shares via
VKWebAppShare and copies via VKWebAppCopyText, since navigator.share is absent in the desktop
iframe and navigator.clipboard is blocked there.
- The invite link is a VK Mini App direct link (vk.com/app<id>#f<code>) on VK instead of the
Telegram link; the app id comes from vk_app_id in the launch params (no build arg needed). The
recipient's launch routes the deep link from VK's `hash` launch query parameter.
- The app's "auto" theme follows the VK client's light/dark appearance (VKWebAppUpdateConfig),
which the VK mobile webview's prefers-color-scheme does not track.
- The safe-area CSS vars default to env(safe-area-inset-*), so the VK mobile layout clears the
home bar (and Capacitor/PWA too); Telegram still overrides them from its SDK.
vk.ts adds vkAppId/vkStartParam/vkShare/vkCopyText/vkOnScheme. Verified: svelte-check, 347 unit
(+ vkAppId/vkStartParam/vkShareLink), build, bundle-gate. The VK-Bridge behaviours need the live
contour (not reproducible headless).
Follow-up to the contour review: with the rack now exactly seven fixed-size slots in the
narrow landscape panel, the 56px confirm button (sized for the roomy portrait rack) was
wider than the single slot a staged tile frees, so it overlapped the now-rightmost tile.
Match the button width to one landscape tile slot, so it sits inside the freed slot with
its right edge level with the rack's right edge — the mirror of the first tile's left edge.
Follow-up to the contour review of the landscape fixes:
- Zoom still positioned "in two steps": a wide viewport overflows vertically as soon as
the square board grows past its height, but horizontally only once it grows past the
(much wider) width — so the browser pins scrollLeft to 0 while the vertical axis
already pans (no time/ease tween fixes it; the horizontal scroll range physically is
not there yet). Found by instrumenting the scroll trajectory. Now drive both axes by
one progress = how far the board has grown past the viewport width (when a horizontal
pan first becomes possible): until then the board just zooms centred, past it both
axes pan together in one diagonal motion. Also disable scroll-anchoring so the browser
stops fighting the programmatic scroll mid-transition. Re-verified: both axes now
start and move together.
- Rack tiles resized when a tile was placed: landscape used flex-grow, so removing a
tile regrew the rest. Give them a fixed size (1/7 of the fixed-width rack), like the
portrait rack, so placing a tile leaves the rest put.
Bugs surfaced while testing in VK but present on every platform (verified across
browsers, not VK-specific):
- Rack tiles: the letter used a fixed rem size, so in landscape — where tiles shrink
below 46px in the narrow left panel — it overflowed and dropped into the bottom-left
corner. Size it relative to the rack (cqw, like the board's labels; the tile's own
container-query size resolves unreliably under flex + aspect-ratio).
- Landscape board zoom positioned "in two steps": the per-frame clamp-to-max reached
the wide axis before the tall one. Interpolate both scroll axes together by time over
the grow/shrink transition instead (settles on zoom-out too).
- The zoomed board could not be panned with a mouse (touch scrolls the overflow:auto
viewport natively; a mouse cannot drag-scroll a div). Add a drag-to-pan handler, active
only while zoomed, off pending tiles, past a small movement threshold, swallowing the
trailing click so it does not also act on a cell.
- Shorten the in-game block-confirm label ru "Блокируем?" -> "В бан?" (the long form
overflowed the seat score chip on mobile).
Mirror the Telegram Mini App wrapper for VK: the SPA loads at a new /vk/
entry, authenticates from VK's signed launch parameters, and provisions a
'vk' platform identity — the minimum to run the game in VK test mode.
- Gateway verifies the launch signature in-process (internal/vkauth:
HMAC-SHA256 over the sorted vk_* params under GATEWAY_VK_APP_SECRET,
base64url) — a pure offline check, no side-service. New auth.vk op
(gated on the secret), backendclient.VKAuth, /vk/ SPA mount.
- Backend: KindVK + ProvisionVK/vkSeed, /sessions/vk handler, identity
kind widened to include 'vk' (migration 00005, expand-contract).
- UI: src/lib/vk.ts (VK Bridge, lazy-imported), bootVK + the /vk/ boot
dispatch, encodeVKLogin + authVK across transport/client/mock. VK omits
the name from the signed params, so the client reads it via
VKWebAppGetUserInfo as an unsigned display seed.
- Deploy: /vk in the edge Caddyfile, GATEWAY_VK_APP_SECRET wired through
compose + .env.example + CI (TEST_) + prod-deploy (PROD_).
- Admin console: surface the VK user id (link to the VK profile) next to
the Telegram id on the user card.
- Docs: ARCHITECTURE §12/§13, FUNCTIONAL (+ _ru), gateway README; VK
integration reference under .claude/.
Signature algorithm verified against dev.vk.com plus independent Node/Python
references and a %2C edge-case vector.
The deploy-check skill is repo-specific: it encodes this project's hard-won
pre-deploy runtime constraints (distroless nonroot, edge Alt-Svc/HTTP3, caddy
recreate, DICT_VERSION boot, expand-contract migrations, the Telegram permission
model) and points at deploy/README.md, docs/ARCHITECTURE.md, docs/EDGE_HTTP3.md and
the agent memory files. It belongs with the deploy logic it guards, not in the global
config, so it travels with the repo and stays versioned alongside it.
The stale-invite / welcome-on-redeem notices are raised during boot, so the native
popup fired over the loading splash. Gate both the native popup and the in-app Modal
on the current route's loading cover being gone — the tile splash on the lobby
(splashDone), the plain loading screen elsewhere (app.ready) — so the notice appears
with the settled screen on every build (Telegram and native/web alike).
The deep-link info modals (stale invite, welcome-on-redeem) captured
insideTelegram() && telegramDialogsAvailable() in a const at component init, but they
mount in App before bootstrap loads the SDK, so the value was always false and they
fell back to the in-app Modal even inside Telegram. Evaluate it at fire time (in the
effect and the {#if} guard), as the destructive confirms already do at click time.
Inside the Mini App, route the destructive confirmations (resign, block, unfriend)
through Telegram's native showConfirm, and present the deep-link info modals (stale
invite, welcome-on-redeem) as a native showPopup whose button opens the bot chat.
Outside Telegram, or on a client predating the dialogs, the existing in-app Modal is
used unchanged; offline also keeps the modal so the action retains its disabled state.
Adds showConfirm/showPopup wrappers to telegram.ts and a pure popup-params builder
(nativedialogs.ts) with unit tests; the deep-link modal components choose native vs
in-app via an effect gated on insideTelegram + dialog availability.
Record the current state in ARCHITECTURE and FUNCTIONAL (+ _ru mirror): inside the
Mini App the client tracks Telegram's live theme switch, fits the full device
safe-area, exposes a native Settings button into the in-app settings, and syncs the
device-independent display prefs (theme, reduce-motion, board labels — not the
interface language) across the user's Telegram devices via CloudStorage; the
validator denies a bot user (is_bot) before provisioning an account.
The validator parsed only id/username/first_name/language_code from the signed
Telegram user, so a WebAppUser flagged is_bot would have been provisioned a normal
account. The HMAC already proves Telegram signed the payload, so is_bot==true is
Telegram itself attesting the launching principal is a bot.
Parse is_bot and reject it (ErrInvalidInitData -> gateway 4xx -> launch error). A
real user opening the Mini App never carries it, so this is a defensive deny. Tests
cover both the denied (is_bot true) and allowed (is_bot false) paths.
Theme, reduce-motion and board labels lived only in local IndexedDB/localStorage,
so they did not follow the user across devices and could be lost when the Telegram
WebView cleared storage.
Mirror these three device-independent prefs to Telegram CloudStorage (Bot API 6.9)
from the single local-persist point (persistPrefs), and reconcile them from
CloudStorage in the background on launch (reconcileCloudPrefs) so a change made on
another device follows the user here. The local store stays the instant-render
cache; the interface language is intentionally excluded (it syncs via the durable
account). Pure encode/decode extracted to cloudprefs.ts with unit tests; the
CloudStorage transport wrappers are added to telegram.ts. No-op outside Telegram or
on a client predating CloudStorage.
Inside the Mini App, reveal Telegram's native Settings button (Bot API 7.0)
and route its taps to the Settings screen — the standard Mini App affordance.
The in-app gear entry stays the primary path (two entry points by design).
No-op outside Telegram or on a client predating the button.
The safe-area bottom inset was reserved on the Screen wrapper, so the strip
that holds space for the home indicator showed the content background and read
as detached from the coloured bottom bar / header above it.
Move the bottom inset onto the bottom bar: the Screen .tabbar wrapper now paints
--bg-elev and pads itself by --tg-safe-bottom, so the strip continues the
TabBar's chrome; a screen with no tab bar pads its bottom-most content
(.content:last-child) instead, so the strip takes that content's own colour.
Only the device safe-area TOP inset was mirrored, so on phones with a home
indicator the rack / bottom bar sat under it, and in landscape the notch
clipped the screen edges.
Mirror the full device safe-area inset (bottom / left / right) into new
--tg-safe-bottom / --tg-safe-left / --tg-safe-right CSS vars (0 outside
Telegram) and pad the shared Screen wrapper by them, so every screen clears the
home indicator and the landscape notch; the top inset stays owned by the header.
Replace telegramSafeAreaTop with telegramSafeAreaInset (the full inset object),
with a unit test.
The Mini App read Telegram's themeParams/colorScheme only once at launch, so
switching Telegram's light/dark theme (or its auto day/night) while the app was
open left the SPA on stale colours until a relaunch.
Subscribe to the themeChanged WebApp event and re-apply the theme live. Extract
the launch-time token + colour-scheme + chrome application into syncTelegramTheme
(reading live themeParams when no launch snapshot is passed) and call it from both
applyTelegramChrome (launch) and the new event handler. Add a telegramThemeParams
live getter (+ unit test). Drop the stale 'immersive fullscreen' note from
applyTelegramChrome — the app deliberately does not request fullscreen.
The promo message body now renders "@<bot>" as an HTML text_link to the same ?startapp deep link the button uses (ParseMode HTML), so tapping the mention opens the seeded Mini App instead of the bot profile. Same payload (campaign start-param, else the forwarded /start payload) backs both the button and the mention.
The promo bot button carries a configurable variant-seed start-param (default verudit_ru-scrabble_en). The gateway parses start_param from the validated initData and forwards it; the backend, on first contact only, seeds the new account variant_preferences from it (English Scrabble alongside the default Erudit).
No schema change (the scrabble_en CHECK is already in the baseline) and the gateway<->backend REST field is additive, so the rolling deploy is safe in either order. TELEGRAM_PROMO_START_PARAM configures the payload (empty forwards the user own /start payload). Covered by account unit tests, a gateway transcode test, and an integration test asserting new-only seeding.
The custom title bar was too tall. Halve the standard bar padding (10->5px) and, in the Telegram path, the notch gap (16->8px) and bottom (6->3px); the notch safe-area inset is unchanged. Title and back chevron stay vertically centred.
The share button uses the OS Web Share sheet outside Telegram (the TG share picker only inside it), so "via Telegram" was misleading. Rename to a neutral "Share"/"Поделиться" and drop the now-unused tgshare class.
Telegram silently routes a copyMessage aimed at a deleted forum topic
into the chat's General topic with NO error, so the bot's error-based
recreate never fired — the user's next message landed in General with no
card. Confirmed against the prod bot logs: a relay after a topic deletion
logged neither "relay to topic failed" nor "topic gone, reopening".
Probe topic liveness before reusing it: re-applying the info card's reply
markup is a no-op that errors only when the card (hence the topic) is
gone, so the bot detects the deletion and reopens the topic + card rather
than relying on the (absent) copy error. The post-copy recreate stays as
a race backstop.
The topic info card built the profile link as `tg://user?id=`, which
clients render as dead text for a user they don't already know (the
test-vs-prod difference an operator saw), and it put the raw display
name through HTML — so a name starting with "/" was auto-detected as a
tappable bot command that fired when tapped.
Render the card with message entities instead: the display name is
covered by a `text_mention` entity (the reliable way to mention a
username-less user the bot has already seen — it messaged the bot),
which links the profile AND, because the name sits inside an entity,
suppresses the "/command" and "@mention" auto-detection on it. Drop the
HTML parse mode and the tg:// link. Entity offsets are UTF-16.
Users who DM the bot anything but /start are relayed into a private
forum supergroup, one topic per user. Operators (the chat's admins)
reply in the topic and the bot copies it back to the user; an info card
opening each topic carries a Block/Unblock toggle and a Clear button.
State is a small JSON file on a new /data volume — the bot host has no
database. Off by default (TELEGRAM_SUPPORT_CHAT_ID=0): the prod bot is
unchanged until the operator sets the chat id and adds the bot as a
forum admin.
- internal/support: concurrency-safe JSON store (topic map, block list,
relayed message ids) with field-targeted mutators and atomic save
- bot/support.go: relay both ways via copyMessage, short-TTL admin
cache, callback buttons, per-user topic-create lock, loop guard
(skip the bot's own posts), reopen a deleted topic on the next message
- config + compose + CI/prod-deploy: TELEGRAM_SUPPORT_CHAT_ID per
contour + TELEGRAM_SUPPORT_STATE_DIR; bot-state named volume; /data
pre-owned by UID 65532 so a fresh volume is writable under distroless
- docs: ARCHITECTURE §15 + decision record, FUNCTIONAL (+ru), README
Finalises the Telegram Mini App navigation work after on-device testing (Pixel
10 / Android 17 + iOS, fresh beta clients):
- Remove requestFullscreen entirely. Immersive fullscreen hid Telegram's native
header (and its BackButton) and the Android system swipe-back minimised the
app; the owner prefers the windowed full-size (expand) presentation, so the
app never requests fullscreen on any platform now.
- The app's own back chevron (Header, showBack = !!back) drives back-navigation
on every platform; the native Telegram BackButton is dropped — it does not
render in the windowed Mini App (backVisible=false on iOS and Android), so
relying on it lost back navigation (notably none on iOS).
- Replace the temporary always-on diagnostic overlay with a hidden debug panel
(components/DebugPanel): ten quick taps on the header title open it; it shows a
privacy-safe client diagnostic snapshot (app version, locale, online, userId,
Telegram chrome / viewport / SDK state — no secrets, no IP) and shares it via
the OS share sheet / clipboard; a tap anywhere except Share dismisses it.
- Drop the now-dead telegramRequestFullscreen / telegramBackButton /
isTelegramAndroid helpers and the iOS-fullscreen unit test.
Telegram has no native non-modal notification API (only modal showPopup /
showAlert), so in-app toasts stay ours. Docs: UI_DESIGN.md.
The Telegram native BackButton does not render in the windowed Mini App (the
owner's emulator + a fresh beta TG report backVisible=false on both iOS and
Android), so relying on it lost back navigation — iOS had no back affordance at
all. Show the app's own back chevron whenever there is a back target, on every
platform (Header showBack = !!back), and drop the now-dead native BackButton
effect (App.svelte). The native close control stays — a windowed Mini App
cannot hide it (no Telegram API).
WIP: the temp lobby diagnostic overlay and the requestFullscreen no-op remain
for the owner's emulator test; finalize after confirmation.
WIP for the Android nav investigation (TEMP bits reverted before merge):
- telegramRequestFullscreen: temporarily a no-op (incl. iOS) so the owner can
confirm the Android "fullscreen look" is Telegram's own Mini App presentation,
not our requestFullscreen (isFullscreen is already false on Android).
- Header: show the app's own back chevron in Telegram on Android, where the
native BackButton does not render — a reliable tap-back. [keep]
- Diagnostic overlay moved app-wide (pointer-events:none) and now reports
BackButton state (req/present/visible) + viewport geometry, to see whether the
native BackButton can capture the Android system swipe-back.
Renders Telegram viewport/fullscreen state (isFullscreen, isExpanded, viewport
heights, innerH vs screenH, safe-area insets) in the lobby, inside Telegram
only, to diagnose why the app still opens fullscreen on Android with
requestFullscreen now iOS-only and no persisted state (fresh TG + test account).
REVERT before merge.
Three Android Mini App issues, all in the Telegram chrome:
- Entering a game showed no native back button, and the Android system
swipe-back minimised the app instead of navigating. Root cause: immersive
fullscreen (requestFullscreen). On Android, fullscreen replaces the native
header — and its BackButton, which also captures the system back — with a bare
close/menu pill, so back navigation has no control to land on. Request
fullscreen on iOS only; Android stays windowed, keeping the native header +
BackButton (and the system swipe-back that routes to it). iOS is unchanged.
- Closing the game board always prompted "changes that you made may not be
saved", even on a board just opened and untouched. The close-confirmation was
armed unconditionally on game mount. Remove it entirely: move drafts auto-save
(debounced during play + flushed on destroy), so nothing is lost on close.
Drops the now-unused telegramClosingConfirmation + isMobilePlatform helpers and
their SDK interface fields. Docs: UI_DESIGN.md.
The diagnostic showed only sdk: yes/no (window.Telegram presence), not why the
SDK was absent. Capture how the dynamic telegram-web-app.js load resolved —
present / loaded / no-webapp / error / timeout — and surface it as
"sdk-load: <outcome>". error/timeout pinpoint a blocked or hanging telegram.org
(the prime suspect for an empty launch); no-webapp a loaded-but-broken script.
loadTelegramSDK records the outcome (telegramSdkOutcome); collectTelegramDiag
carries it into the screen. Unit tests cover each outcome; the blocked-script
e2e now asserts sdk-load: error.
The Telegram SDK was a render-blocking <script src="telegram.org/..."> in the
shared index.html shell, so it ran on every entry (/telegram/, /app/, native).
On a network that blocks telegram.org — common where Telegram itself reaches
users only over a proxy — the script hangs forever, stranding the whole page,
including the launch-diagnostic screen meant to surface exactly this failure.
This is the likely root cause of the Android "won't open" reports (all launch
methods fail identically; iOS on a different network works).
Remove the head <script> and load the SDK dynamically (lib/telegram.ts
loadTelegramSDK) with a 10s timeout, only on a Telegram entry (the /telegram/
path or a tgWebApp launch fragment). The SPA — served from our own reachable
origin — boots first and controls the load: on a block/error it falls through
to the diagnostic screen (reporting sdk: no) instead of hanging, and Retry
re-attempts. /app/ and the native build no longer touch telegram.org.
Pin the SDK to the version the official page recommends (?62) for the newer
client features the app already uses (fullscreen, safe areas, swipe guard).
Tests: loadTelegramSDK unit tests (present / error / timeout); an e2e that
aborts the script fetch and asserts the diagnostic still renders.
A Mini App launch on /telegram/ without sign-in data (empty initData) used to
location.replace('/') — bouncing the visitor to the marketing landing. That
destroyed all diagnosability and, on some Android clients that reach the entry
with no initData, simply looked like the app refusing to open.
Replace the bounce with a compact, privacy-safe launch-error screen
(screens/TelegramLaunchError.svelte) that renders a one-screenshot diagnostic
snapshot captured at the moment of failure: SDK/WebApp presence, Telegram
platform/version, whether initData is empty, whether the URL fragment carried
tgWebAppData, the initData field NAMES present/missing (never the signed
values, never an IP), and OS/mobile/browser via User-Agent Client Hints plus
the full User-Agent. A Share button delivers it through the OS share sheet
(clipboard copy on desktop, reusing the GCG no-webview-strand guard); Retry
re-checks in place to recover a late initData without a reload (which would
discard the launch fragment).
This is the instrument to root-cause the Android empty-initData failure, which
is not reproducible in Playwright.
Tests: collectTelegramDiag + shareText/pickTextShare unit tests; the /telegram/
e2e now asserts the diagnostic screen, not a redirect. Docs: ARCHITECTURE.md +
UI_DESIGN.md updated.
Cover the reworked Settings -> Friends interactions in the mock e2e
(Chromium + WebKit): the row kebab slides open the block/remove icons and an
outside tap collapses it; blocking confirms (naming the friend) and moves them
to Blocked; removing confirms and drops the friendship.
A slid-open friend row now collapses when the user taps anywhere outside its
action buttons (taps on a kebab are skipped so its own toggle still drives the
open/close). Uses the same capture-phase window pointerdown idiom as Screen,
active only while a row is revealed.
Settings -> Friends previously rendered each friend as a bordered card with
two always-visible text buttons (Remove / Block) that fired immediately. Rework
the whole screen to the lobby's visual language: one-line rows split by
hairline separators across all three sections (friends, incoming requests,
blocked).
Each friend row gains a right-hand kebab that slides the row open to reveal two
icon actions split by a vertical divider -- block (no-entry) and remove (cross)
-- mirroring the lobby's slide-to-reveal. Both actions now require a
confirmation modal; since the slide moves a short name off-screen, the modal
keeps a generic title and shows the friend's name in the body, above the
buttons, so a long name cannot stretch the sheet. Incoming keeps its
accept/decline buttons and blocked keeps unblock, inline on their rows.
Add the friends.actions / friends.blockConfirm / friends.unfriendConfirm keys
to both i18n catalogs and document the flow in FUNCTIONAL (+_ru).
The prod rolling deploy rolls each service with `compose up -d --no-deps <svc>`.
For caddy that is a no-op on a config-only release: its image is pinned
(caddy:2-alpine, no $TAG), so the compose definition is unchanged between
releases, compose treats the container as current and does not recreate it, and
admin is off so there is no hot reload. The new bind-mounted Caddyfile is seeded
to the host but never loaded -- the v1.2.2 `Alt-Svc: clear` edge fix deployed
green yet did not take effect until caddy was restarted by hand.
Force a recreate for caddy on its roll (every other service already recreates on
its new $TAG image), so a bind-mounted Caddyfile change always applies. Costs a
~1-2s caddy blip per deploy, acceptable for the infrequent manual prod rollout.
The "Database size" stat had no thresholds, so Grafana applied its stat
default (green base, red at >=80). The query is pg_database_size_bytes, so a
healthy ~9 MiB database (9.4M >> 80) rendered permanently RED on the
Scrabble - Resources dashboard (test + prod), reading as a false alert; the
neighbouring percentunit cache-hit stat stayed green only because its 0..1
values fall under 80.
Add absolute byte thresholds sized to the 40 GiB prod disk (4.6 GiB used,
observability bounded -- Tempo <=1 GiB, Prometheus 7d -- so the DB is the
only unbounded grower): green up to 8 GiB, yellow at 8 GiB (~20% of disk),
red at 16 GiB (~40%), an early warning with ample runway before the disk
tightens, not a panic line. Cosmetic panel coloring only; there are no
Grafana alert rules provisioned.
Caddy enables HTTP/3 by default on any TLS listener and emits
Alt-Svc: h3=":443"; ma=2592000, but UDP/443 is never reachable: the prod
compose maps only "443:443" (TCP) and ufw opens 443/tcp (test contour: the
host caddy publishes only :443/tcp). A client that cached the 30-day advert
tries QUIC first on later opens, gets no response, and waits for the QUIC
attempt to time out before falling back to h2 -- which surfaced as the
Telegram Mini App intermittently hanging on load (a barely-noticeable pause
up to a blank window). The h2/TCP serving path itself is healthy (~10ms TTFB).
Emit Alt-Svc: clear site-wide at the contour caddy so clients actively drop
any cached alternative and stay on h2/h1. This caddy terminates TLS in prod
(the fix target); in the test contour it serves plain :80 and the host caddy
re-stamps its own Alt-Svc, so the live test fix lives in the host caddy. Add
docs/EDGE_HTTP3.md (symptom, diagnosis method, verify, and option B -- serving
h3 for real -- if it recurs) and link it from ARCHITECTURE.md.
2026-06-22 21:20:31 +02:00
787 changed files with 82798 additions and 2730 deletions
description: "Use before any deploy-touching change to this repo — phrases like '/deploy-check', 'is this prod-safe', 'before we deploy', 'deploy safety review', 'проверь перед деплоем', 'это безопасно для прода'. Runs a pre-deploy checklist of this project's hard-won runtime constraints against the current diff, so the crash classes that have bitten live environments get caught before shipping instead of after."
---
# Pre-deploy runtime-constraint check
Triggered before shipping anything that touches the deploy contour (Dockerfiles,
`deploy/`, Caddyfile, compose, migrations, boot guards, the Telegram side-service,
edge config). The worst frictions in this repo were never logic bugs — they were
**environment mismatches that crashed a live env and forced a redesign**. Run this
list against the diff first; turn crash-and-redesign into a single pass.
This checklist is a prompt, **not** the source of truth. The canonical detail
lives in `deploy/README.md`, `docs/ARCHITECTURE.md`, `docs/EDGE_HTTP3.md`, and the
agent memory files referenced below — read them when an item is in play, and add a
new class here when a new incident teaches one.
## How to run it
1.`git diff <base>...HEAD --stat` to see what the change actually touches.
2. For every risk class below that the diff touches, perform the **Check** and
report PASS / FAIL with the exact file:line to fix. Skip classes the diff does
not touch — say which you skipped and why.
3. Remember: **deploy-job green ≠ healthy**. CI's deploy probe has historically
passed with a dead backend (it only checked static landing+gateway). Verify the
real feature live (`/readyz`, the actual flow) after deploy, not just the green.
## Risk classes
### 1. Container user — distroless nonroot UID 65532
- **Bit us:** TLS keys `chmod 600` for the host owner crash-looped gateway + bot at
boot with "permission denied" — service images run UID 65532.
- **Check:** any new/changed mounted secret, key, or config file must be readable by
UID 65532 (`0644`, not `0600`). Scan the diff for file modes, `chmod`, and new
echo "::error::sdkmanager not found/executable at $sm — set the ANDROID_SDK_DIR variable if the SDK lives elsewhere, or grant the runner user read+exec: sudo chmod -R a+rX \"$ANDROID_HOME\""; exit 1
fi
for pkg in "platforms/android-36" "build-tools"; do
tar xzf /tmp/dawg.tar.gz -C "${GITHUB_WORKSPACE}/dawg"
- name:Set up Go
uses:actions/setup-go@v5
with:
go-version-file:go.work
cache:true
- name:Generate golden parity vectors
run:|
go run ./backend/cmd/dictgen -dawg-dir "${GITHUB_WORKSPACE}/dawg" -out /tmp/dictgold
go run ./backend/cmd/validategen -dawg-dir "${GITHUB_WORKSPACE}/dawg" -out /tmp/validgold
go run ./backend/cmd/movegen -dawg-dir "${GITHUB_WORKSPACE}/dawg" -out "${GITHUB_WORKSPACE}/movegold"
- name:Set up Node
uses:actions/setup-node@v4
with:
node-version:22
- name:Install pnpm
run:npm install -g pnpm@11.0.9
- name:Install deps
working-directory:ui
run:pnpm install --frozen-lockfile
- name:Local-eval conformance (reader + validator vs the Go engine)
working-directory:ui
env:
DICT_DAWG_DIR:${{ github.workspace }}/dawg
DICT_GOLD_DIR:/tmp/dictgold
DICT_VALID_DIR:/tmp/validgold
DICT_MOVEGEN_DIR:${{ github.workspace }}/movegold
run:pnpm exec vitest run src/lib/dict/
# gate is the single branch-protection required check. It always runs and passes
# only when each upstream job succeeded or was skipped (a path-filtered no-op),
# failing the merge if any actually failed or was cancelled.
gate:
needs:[unit, integration, ui]
needs:[unit, integration, ui, conformance]
if:always()
runs-on:ubuntu-latest
defaults:
@@ -221,7 +301,7 @@ jobs:
- name:Aggregate required checks
run:|
fail=
for r in "unit:${{ needs.unit.result }}" "integration:${{ needs.integration.result }}" "ui:${{ needs.ui.result }}"; do
for r in "unit:${{ needs.unit.result }}" "integration:${{ needs.integration.result }}" "ui:${{ needs.ui.result }}" "conformance:${{ needs.conformance.result }}"; do
name="${r%%:*}"; res="${r#*:}"
echo "$name = $res"
case "$res" in
@@ -236,9 +316,13 @@ jobs:
# Auto test-deploy on a PR into development and on the push that merges it.
# A PR into master is test-only (this job is skipped); prod deploy is manual.
# Gates on `gate` (so a real test failure blocks the deploy) but runs even when
g+=`<text x="${0.055*S}" y="${S-64}" font-size="18" font-weight="600" fill="#fff">Wood grain: 6 equal columns; a vertical stripe on each column's right edge,</text>`;
g+=`<text x="${0.055*S}" y="${S-40}" font-size="18" font-weight="600" fill="#fff">width 1/32 of side; the whole set shifted left 1/16; every stripe tilted 4°.</text>`;
Animated logo for the **VK app loading screen** (shown before the SPA assets load).
The Erudit «Э» tile (score `8`) **drops in under gravity, lands with a soft squash —
its left/right edges bulging into a cushion — then springs back up**, looping. A light
"glint" is caught at the moment of the bounce.
| File | Purpose |
|------|---------|
| `erudit-loader.json` | The Lottie to upload to VK. |
| `erudit-loader-preview.gif` | Looping preview. Rendered on a green "baize" background **only in the preview** — the real asset has a **transparent** background. |
| `build/` | The reproducible build pipeline (see *Regenerate*). |
| `BACKEND_SMTP_FROM` | `no-reply@localhost` | Envelope/From address for confirm-codes. |
| `BACKEND_SMTP_HOST` | — | Confirm-code relay host. **Empty selects the development log mailer** (the code is logged, not sent). |
| `BACKEND_SMTP_PORT` | `587` | Relay port. No client certificate is needed (the server cert is validated against the system roots). |
| `BACKEND_SMTP_TLS` | — | Transport security: `ssl` (implicit TLS from connect) or `starttls`. Empty derives it from the port (implicit on `465`, STARTTLS otherwise); set it for a relay on a non-standard port (e.g. Selectel's `1127` = SSL, `1126` = STARTTLS). |
| `BACKEND_SMTP_FROM` | `no-reply@localhost` | From address. A deployed contour must use the prod domain (the relay only accepts its verified sender domain). |
| `BACKEND_PUBLIC_BASE_URL` | — | Canonical public origin (scheme + host) for links in the email. **Required when `BACKEND_SMTP_HOST` is set.** Never derived from a request Host header (anti-injection). |
| `BACKEND_CONNECTOR_ADDR` | — | the gateway bot-link relay gRPC address for admin-console operator broadcasts. Empty disables broadcasts. |
| `BACKEND_GUEST_REAP_INTERVAL` | `1h` | How often the abandoned-guest reaper sweeps. |
| `BACKEND_GUEST_RETENTION` | `720h` | Account age past which a guest with no game seat is deleted. |
// sampleFixture is the JSON committed with the .dawg so the TypeScript cursor test
// knows the exact word set (as alphabet indexes) to expect from enumeration.
typesampleFixturestruct{
Alphabetstring`json:"alphabet"`
NumAddedint`json:"numAdded"`
Words[]string`json:"words"`
Indexes[][]int`json:"indexes"`
}
// genFixture is the move-generation golden set for one sample dictionary.
typegenFixturestruct{
RulesetgenRuleset`json:"ruleset"`
Cases[]genCase`json:"cases"`
}
// genRuleset is the scoring data the TS side rebuilds so evaluate() matches the Go
// solver: letter values, premium multipliers per square, the centre, rack size and bonus.
typegenRulesetstruct{
Sizeint`json:"size"`
Colsint`json:"cols"`
Centerint`json:"center"`
RackSizeint`json:"rackSize"`
Bingoint`json:"bingo"`
Values[]int`json:"values"`
LetterMult[][]int`json:"letterMult"`
WordMult[][]int`json:"wordMult"`
}
// genTile is one placed tile (a board tile or a move placement).
typegenTilestruct{
Rowint`json:"row"`
Colint`json:"col"`
Letterint`json:"letter"`
Blankbool`json:"blank"`
}
// genRack is a rack as a multiset of letter indexes plus a blank count.
typegenRackstruct{
Letters[]int`json:"letters"`
Blanksint`json:"blanks"`
}
// genMove is one ranked generated play: its orientation, placed tiles and total score.
typegenMovestruct{
Dirint`json:"dir"`
Tiles[]genTile`json:"tiles"`
Scoreint`json:"score"`
}
// genCase is one generation position: the tiles already on the board (empty when
// none), the rack, the mode/rule and the ranked moves the solver returns.
typegenCasestruct{
Namestring`json:"name"`
Placed[]genTile`json:"placed"`
RackgenRack`json:"rack"`
Modeint`json:"mode"`
IgnoreCrossWordsbool`json:"ignoreCrossWords"`
Moves[]genMove`json:"moves"`
}
funcmain(){
out:=flag.String("out","ui/src/lib/dict/testdata","output directory for fixtures")
dawgDir:=flag.String("dawg-dir","","when set, emit real-dictionary move-gen golden from the .dawg files in this dir (conformance mode) instead of the committed samples")
where=robotExists+` = false AND a.deleted_at IS NOT NULL`
default:
where=robotExists+` = false AND a.deleted_at IS NULL`
}
ifext:=LikePattern(f.ExternalIDMask);ext!=""{
ifname!=""{
args=append(args,name)
where+=fmt.Sprintf(` AND (a.display_name ILIKE $%d ESCAPE '\' OR a.deleted_display_name ILIKE $%d ESCAPE '\')`,len(args),len(args))
}
ifext!=""{
args=append(args,ext)
where+=fmt.Sprintf(` AND EXISTS (SELECT 1 FROM backend.identities i WHERE i.account_id = a.account_id AND i.external_id ILIKE $%d ESCAPE '\')`,len(args))
where+=fmt.Sprintf(` AND (EXISTS (SELECT 1 FROM backend.identities i WHERE i.account_id = a.account_id AND i.external_id ILIKE $%d ESCAPE '\') OR EXISTS (SELECT 1 FROM backend.retained_identities r WHERE r.account_id = a.account_id AND r.external_id ILIKE $%d ESCAPE '\'))`,len(args),len(args))
}
ifemail!=""{
args=append(args,email)
where+=fmt.Sprintf(` AND (EXISTS (SELECT 1 FROM backend.identities i WHERE i.account_id = a.account_id AND i.kind = 'email' AND i.external_id = $%d) OR EXISTS (SELECT 1 FROM backend.retained_identities r WHERE r.account_id = a.account_id AND r.kind = 'email' AND r.external_id = $%d))`,len(args),len(args))
q:=`SELECT a.account_id, a.display_name, a.preferred_language, a.is_guest, a.flagged_high_rate_at, a.created_at, `+robotExists+` AS is_robot
q:=`SELECT a.account_id, a.display_name, a.preferred_language, a.is_guest, a.flagged_high_rate_at, a.created_at, `+robotExists+` AS is_robot, (a.deleted_at IS NOT NULL) AS is_deleted
FROM backend.accounts a WHERE `+where+
fmt.Sprintf(` ORDER BY a.created_at DESC LIMIT $%d OFFSET $%d`,len(args)+1,len(args)+2)
args=append(args,limit,offset)
@@ -82,7 +114,7 @@ FROM backend.accounts a WHERE ` + where +
<p class="note">Shows to <em>everyone</em>, always — bypassing paid accounts, hint wallets and the no-banner role. While any urgent campaign is live it is the only thing the strip shows (other campaigns and the default are suppressed). For system alerts.</p>
<p class="note">Live preview of the strip on both themes. An empty override falls back to the neutral theme colours; the top/bottom border is derived from the background.</p>
var TOK={light:{bg:'#e3e7ee',fg:'#6b7280',link:'#2f6df6'},dark:{bg:'#272f3c',fg:'#9aa3b2',link:'#5b8cff'}};
function byName(n){return document.querySelector('[name="'+n+'"]');}
var allOn=byName('override_all_on'), darkOn=byName('override_dark_on');
if(!allOn||!darkOn){return;}
var f={ab:byName('override_bg'),af:byName('override_fg'),al:byName('override_link'),db:byName('override_bg_dark'),df:byName('override_fg_dark'),dl:byName('override_link_dark')};
var lightEl=document.getElementById('prev-light'), darkEl=document.getElementById('prev-dark');
function hexToRgb(h){h=h.replace('#','');return [parseInt(h.slice(0,2),16),parseInt(h.slice(2,4),16),parseInt(h.slice(4,6),16)];}
function pad(x){x=Math.max(0,Math.min(255,Math.round(x))).toString(16);return x.length<2?'0'+x:x;}
function rgbToHex(r){return '#'+pad(r[0])+pad(r[1])+pad(r[2]);}
function mix(a,b,t){return [a[0]+(b[0]-a[0])*t,a[1]+(b[1]-a[1])*t,a[2]+(b[2]-a[2])*t];}
function lum(r){return (0.2126*r[0]+0.7152*r[1]+0.0722*r[2])/255;}
// Derived border: nudge the background 14% toward black on a light bg, toward white on a dark bg.
function border(bg){var r=hexToRgb(bg);return rgbToHex(mix(r, lum(r)>0.5?[0,0,0]:[255,255,255], 0.14));}
<p class="note">A <strong>pack</strong> funds chips (a money price per rail — RUB via direct, VOTE via vk, XTR via telegram); a <strong>value</strong> buys benefits with chips (a CHIP price). Archived products are hidden from players but still credit an in-flight payment and can be granted. A product with transactions can only be archived, not deleted. Amounts are in minor units (RUB kopecks; VOTE/XTR/CHIP whole). The <code>tournament</code> atom is not sellable yet — keep such a product archived.</p>
<label>CHIP — value <input type="number" name="price_chip" min="0"></label>
</fieldset>
<label><input type="checkbox" name="active" value="true"> Active (on sale)</label>
<div><button type="submit">Add</button></div>
</form>
</section>
<section class="panel"><h2>Rewarded ads (watch for chips)</h2>
<p class="note">The chips a player earns per rewarded-video view (VK only), plus the per-day and per-hour caps that bound free chips — <strong>0 payout turns rewarded off</strong>. This is the shared reward config, not a product.</p>
<p class="note">Turn purchases off on a rail/channel and show the user a localized reason on their next attempt. Unchecked = off; an empty message shows a built-in “temporarily unavailable”. Fail-open: a rail you never touch stays on. A per-user “allow” override (on a user card) bypasses this switch.</p>
{{if not .Transacted}}<form class="form" method="post" action="/_gm/catalog/{{.ID}}/delete" onsubmit="return confirm('Delete this product? It has never been transacted, so this is safe and permanent.')"><button type="submit">Delete</button></form>{{end}}
<p class="note">Per-tier, per-kind caps on a player's simultaneous unfinished games. <strong>-1</strong> = unlimited, <strong>0</strong> = the kind is blocked, a positive number caps concurrent games of that kind. Guests are additionally blocked from friend games outright. Changes apply immediately (no redeploy); games already in progress are never affected.</p>
<p class="note">A zero quantity / blank price removes that atom / price. Amounts are in minor units. Saving revalidates the sellable shape when the product is active. Archive / unarchive from the <a href="/_gm/catalog">catalog list</a>.</p>
<td>{{if and (eq .Kind "fund") .Order}}<form class="form" method="post" action="/_gm/users/{{$uid}}/refund" onsubmit="return confirm('Refund this order in full? Record the money refund on the rail first; this revokes the chips (floored at 0).')"><input type="hidden" name="order_id" value="{{.Order}}"><button type="submit">Refund</button></form>{{end}}</td></tr>
{{end}}
</tbody>
</table>
<p class="note"><a href="/_gm/ledger.csv">Export the full ledger (CSV)</a> — all accounts, for tax + reconciliation.</p>
{{else}}<p class="note">payments not enabled</p>{{end}}
</section>
<section class="panel"><h2>Grant benefits</h2>
{{if .Grant.Present}}
<p class="note">A zero-price admin sale of a value — <strong>never chips</strong>. The origin is your compliance choice. The by-product grant applies a defined bundle, including an archived reward product.</p>
<form class="form" method="post" action="/_gm/users/{{.ID}}/remove-email" onsubmit="return confirm('Erase the email identity from this account? The address will be freed.')">
<button type="submit">Erase email</button>
</form>
{{end}}
</section>
<section class="panel"><h2>Deletion & retention</h2>
<form class="form" method="post" action="/_gm/users/{{.ID}}/delete" onsubmit="return confirm('Delete this account? Its credentials are journalled and freed, its data anonymised, and its sessions revoked. This cannot be undone.')">
<button type="submit">Delete user</button>
</form>
{{end}}
</section>
<section class="panel"><h2>Friends</h2>
<table class="list">
@@ -142,13 +214,18 @@
{{else}}<p class="note">connector not configured (set BACKEND_CONNECTOR_ADDR)</p>{{end}}
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
developer
Ilia Denisov
