Merge pull request 'Release v1.8.0: prod build fix (re-promote)' (#175) from development into master

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`.

.gitea/workflows/ci.yaml

@@ -26,12 +26,12 @@ on:
   push:
     branches: [development]
 
-# The dictionary release the test suite validates against — the current
+# The dictionary release. One Gitea variable is the single source of truth: the
-# scrabble-dictionary release. Centralised here so a release bump is one edit; the
+# test suite validates against it here (inherited by the unit/integration jobs) and
-# unit/integration jobs inherit it. The deploy job overrides it per contour with
+# both contours' deploy jobs seed a fresh volume with the same value. A release bump
-# vars.TEST_DICT_VERSION (the seed for a fresh volume), see deploy/README.md.
+# is one edit (the variable). See deploy/README.md.
 env:
-  DICT_VERSION: v1.3.1
+  DICT_VERSION: ${{ vars.DICT_VERSION }}
 
 jobs:
   # changes detects which areas a PR/push touched, so the test jobs can skip when
@@ -329,24 +329,42 @@ jobs:
           TELEGRAM_PROMO_BOT_TOKEN: ${{ secrets.TEST_TELEGRAM_PROMO_BOT_TOKEN }}
           # VK Mini App protected key (offline HMAC for the launch-param signature); empty
           # leaves the VK auth path (auth.vk) disabled until the operator sets the secret.
-          GATEWAY_VK_APP_SECRET: ${{ secrets.TEST_GATEWAY_VK_APP_SECRET }}
+          # One VK Mini App serves every contour -> unprefixed secret.
+          GATEWAY_VK_APP_SECRET: ${{ secrets.GATEWAY_VK_APP_SECRET }}
+          # VK ID web login (browser VK-identity linking): the VK ID "Web" app's protected key
+          # for the server-side confidential code exchange — a SEPARATE VK app from the Mini
+          # App above. One VK ID "Web" app serves every contour -> unprefixed secret.
+          GATEWAY_VK_ID_CLIENT_SECRET: ${{ secrets.GATEWAY_VK_ID_CLIENT_SECRET }}
+          # Planted honeytoken bearer: presenting it flags the caller (logs + a ban metric on
+          # test where the IP ban is off; a 24h IP ban on prod). Per-contour secret; empty = trap off.
+          GATEWAY_HONEYTOKEN: ${{ secrets.TEST_GATEWAY_HONEYTOKEN }}
           # Signs the finished-game export download URLs (backend + compose interpolation).
           EXPORT_SIGN_KEY: ${{ secrets.TEST_EXPORT_SIGN_KEY }}
-          # Transactional email via the shared Selectel relay. Empty host leaves the
+          # Transactional email via the shared Selectel relay: one account for every
-          # backend on the log mailer (email disabled) but the contour still boots.
+          # contour -> unprefixed host/port/tls/user/pass. Empty host leaves the backend
-          SMTP_RELAY_USER: ${{ secrets.TEST_SMTP_RELAY_USER }}
+          # on the log mailer (email disabled) but the contour still boots.
-          SMTP_RELAY_PASS: ${{ secrets.TEST_SMTP_RELAY_PASS }}
+          SMTP_RELAY_USER: ${{ secrets.SMTP_RELAY_USER }}
-          SMTP_RELAY_HOST: ${{ vars.TEST_SMTP_RELAY_HOST }}
+          SMTP_RELAY_PASS: ${{ secrets.SMTP_RELAY_PASS }}
-          SMTP_RELAY_PORT: ${{ vars.TEST_SMTP_RELAY_PORT }}
+          SMTP_RELAY_HOST: ${{ vars.SMTP_RELAY_HOST }}
-          SMTP_RELAY_TLS: ${{ vars.TEST_SMTP_RELAY_TLS }}
+          SMTP_RELAY_PORT: ${{ vars.SMTP_RELAY_PORT }}
+          SMTP_RELAY_TLS: ${{ vars.SMTP_RELAY_TLS }}
           SMTP_RELAY_FROM: ${{ vars.TEST_SMTP_RELAY_FROM }}
+          # Operator alerts: backend admin emails (new feedback / complaints) + Grafana
+          # infra alerts. Distinct senders + recipients; Grafana uses the relay's STARTTLS
+          # host:port. Empty leaves the alert worker off and Grafana SMTP disabled.
+          SMTP_RELAY_ADMIN_FROM: ${{ vars.TEST_SMTP_RELAY_ADMIN_FROM }}
+          ADMIN_EMAIL: ${{ vars.TEST_ADMIN_EMAIL }}
+          SMTP_RELAY_SERVICE_FROM: ${{ vars.TEST_SMTP_RELAY_SERVICE_FROM }}
+          SERVICE_EMAIL: ${{ vars.TEST_SERVICE_EMAIL }}
+          GRAFANA_SMTP_PORT: ${{ vars.GRAFANA_SMTP_PORT }}
+          GF_SMTP_ENABLED: ${{ vars.TEST_GF_SMTP_ENABLED }}
           # Canonical public origin for links in the email (this contour's URL);
           # required by the backend whenever SMTP_RELAY_HOST is set.
           PUBLIC_BASE_URL: ${{ vars.TEST_PUBLIC_BASE_URL }}
           GM_BASICAUTH_USER: ${{ vars.TEST_GM_BASICAUTH_USER }}
-          GRAFANA_ROOT_URL: ${{ vars.TEST_GRAFANA_ROOT_URL }}
           CADDY_SITE_ADDRESS: ${{ vars.TEST_CADDY_SITE_ADDRESS }}
-          TELEGRAM_MINIAPP_URL: ${{ vars.TEST_TELEGRAM_MINIAPP_URL }}
+          # TELEGRAM_MINIAPP_URL, GRAFANA_ROOT_URL and VITE_VK_ID_REDIRECT_URL are derived
+          # from PUBLIC_BASE_URL in the run step below, not stored as their own variables.
           TELEGRAM_GAME_CHANNEL_ID: ${{ vars.TEST_TELEGRAM_GAME_CHANNEL_ID }}
           TELEGRAM_CHAT_ID: ${{ vars.TEST_TELEGRAM_CHAT_ID }}
           TELEGRAM_SUPPORT_CHAT_ID: ${{ vars.TEST_TELEGRAM_SUPPORT_CHAT_ID }}
@@ -359,12 +377,16 @@ jobs:
           VITE_TELEGRAM_BOT_ID: ${{ vars.TEST_VITE_TELEGRAM_BOT_ID }}
           VITE_TELEGRAM_LINK: ${{ vars.TEST_VITE_TELEGRAM_LINK }}
           VITE_TELEGRAM_GAME_CHANNEL_NAME: ${{ vars.TEST_VITE_TELEGRAM_GAME_CHANNEL_NAME }}
-          VITE_VK_APP_LINK: ${{ vars.TEST_VITE_VK_APP_LINK }}
+          # VK Mini App landing link + VK ID "Web" app id: one value each serves every
-          VITE_GATEWAY_URL: ${{ vars.TEST_VITE_GATEWAY_URL }}
+          # contour -> unprefixed. VITE_VK_APP_ID also feeds the gateway (GATEWAY_VK_ID_APP_ID);
-          # Unset vars render empty -> the compose ":-" defaults apply.
+          # the VK ID redirect URL is derived from PUBLIC_BASE_URL in the run step below.
+          VITE_VK_APP_LINK: ${{ vars.VITE_VK_APP_LINK }}
+          VITE_VK_APP_ID: ${{ vars.VITE_VK_APP_ID }}
+          # VITE_GATEWAY_URL omitted: the SPA is served same-origin, so it stays the
+          # compose ":-" empty default. Other unset vars likewise fall to their defaults.
           POSTGRES_DB: ${{ vars.TEST_POSTGRES_DB }}
           POSTGRES_USER: ${{ vars.TEST_POSTGRES_USER }}
-          DICT_VERSION: ${{ vars.TEST_DICT_VERSION }}
+          DICT_VERSION: ${{ vars.DICT_VERSION }}
           LOG_LEVEL: ${{ vars.TEST_LOG_LEVEL }}
         run: |
           # Seed the config files to a stable host path. The runner checks out into
@@ -375,8 +397,34 @@ jobs:
           conf="$HOME/.scrabble-deploy"
           rm -rf "$conf"
           mkdir -p "$conf"
-          cp -r caddy otelcol prometheus tempo grafana "$conf"/
+          cp -r caddy otelcol prometheus tempo grafana blackbox "$conf"/
           export SCRABBLE_CONFIG_DIR="$conf"
+          # Maintenance page for the redeploy window, mirroring prod-deploy.sh so the SPA
+          # overlay is exercised on the test contour too (not only prod). Raised just before
+          # the recreate and lowered once caddy is back (below); the trap clears it if the
+          # step fails so the contour never sticks in maintenance (and the reseed above wipes a
+          # stale flag anyway). The caddy-routed probes run in the NEXT step, after it is lowered.
+          maint_flag="$conf/caddy/on"
+          trap 'rm -f "$maint_flag"' EXIT
+          # Derive the public URLs from the one canonical origin instead of storing each as
+          # its own variable (paths are structural SPA routes / the Caddy /_gm sub-path).
+          # Exported before build so the VK ID redirect is baked into the SPA.
+          base="${PUBLIC_BASE_URL%/}"
+          export TELEGRAM_MINIAPP_URL="$base/telegram/"
+          export GRAFANA_ROOT_URL="$base/_gm/grafana/"
+          export VITE_VK_ID_REDIRECT_URL="$base/app/"
+          # Grafana's SMTP from_address must be a BARE address (it rejects the "Name" <addr>
+          # form the backend go-mail accepts) and validates it even when SMTP is disabled — a
+          # bad value crash-loops Grafana. Split the display-format SERVICE From into a bare
+          # address + name for Grafana; the backend keeps the full form.
+          svc_from="${SMTP_RELAY_SERVICE_FROM:-}"
+          case "$svc_from" in
+            *"<"*">"*)
+              export GRAFANA_SMTP_FROM_ADDRESS="$(printf '%s' "$svc_from" | sed -E 's/.*<([^>]+)>.*/\1/')"
+              export GRAFANA_SMTP_FROM_NAME="$(printf '%s' "$svc_from" | sed -E 's/[[:space:]]*<[^>]*>.*$//; s/^"//; s/"$//')" ;;
+            *)
+              export GRAFANA_SMTP_FROM_ADDRESS="$svc_from" ;;
+          esac
           # Bot-link mTLS material for the test contour: a private CA + gateway/bot
           # leaves (CN=gateway, the service name the bot dials). Prod supplies these
           # from PROD_ secrets instead. Regenerated each deploy; both ends redeploy
@@ -389,12 +437,23 @@ jobs:
           # bot on its own host instead (deploy/docker-compose.bot.yml), and the prod
           # main host omits both. Without the profile they would not start here.
           docker compose --ansi never --profile telegram-local build --progress plain
+          # Raise the maintenance page, THEN bring caddy onto the reseeded config mount so it
+          # actually carries the flag: the running caddy sits on the stale pre-reseed mount (the
+          # dir was rm'd + recreated — see the force-recreate note below), so a flag written to
+          # the new dir is invisible until caddy is recreated. With the fresh caddy up, an open
+          # SPA sees the 503 marker + overlay for the whole recreate window, not a bare reconnect.
+          : > "$maint_flag"
+          docker compose --ansi never up -d --force-recreate --no-deps caddy
           docker compose --ansi never --profile telegram-local up -d --remove-orphans
           # The config-only services bind-mount the reseeded config dir. A plain `up -d`
           # leaves them on the previous bind mount (the dir was rm'd + recreated), so a
-          # changed Caddyfile or Grafana dashboard is ignored — force-recreate them to
+          # changed Grafana dashboard is ignored — force-recreate them to pick up the fresh
-          # pick up the fresh config.
+          # config. (Caddy was already recreated above so it would carry the maintenance flag.)
-          docker compose --ansi never up -d --force-recreate --no-deps caddy otelcol prometheus tempo grafana
+          docker compose --ansi never up -d --force-recreate --no-deps otelcol prometheus tempo grafana
+          # Lower the maintenance page: services are back. An open SPA's poll now gets through
+          # (once the gateway finishes booting) and reloads into the fresh client; the caddy
+          # probes in the next step see 200. The EXIT trap is a backstop if we failed earlier.
+          rm -f "$maint_flag"
 
       - name: Probe the landing, gateway and backend
         run: |

.gitea/workflows/prod-deploy.yaml

@@ -40,12 +40,22 @@ jobs:
       VITE_TELEGRAM_BOT_ID: ${{ vars.PROD_VITE_TELEGRAM_BOT_ID }}
       VITE_TELEGRAM_LINK: ${{ vars.PROD_VITE_TELEGRAM_LINK }}
       VITE_TELEGRAM_GAME_CHANNEL_NAME: ${{ vars.PROD_VITE_TELEGRAM_GAME_CHANNEL_NAME }}
-      VITE_VK_APP_LINK: ${{ vars.PROD_VITE_VK_APP_LINK }}
+      # VK Mini App link + VK ID "Web" app id: one value each serves every contour.
-      VITE_GATEWAY_URL: ${{ vars.PROD_VITE_GATEWAY_URL }}
+      VITE_VK_APP_LINK: ${{ vars.VITE_VK_APP_LINK }}
+      VITE_VK_APP_ID: ${{ vars.VITE_VK_APP_ID }}
+      # VITE_GATEWAY_URL omitted: the SPA is served same-origin (compose ":-" default).
       POSTGRES_PASSWORD: ${{ secrets.PROD_POSTGRES_PASSWORD }}
       GM_BASICAUTH_HASH: ${{ secrets.PROD_GM_BASICAUTH_HASH }}
-      TELEGRAM_MINIAPP_URL: ${{ vars.PROD_TELEGRAM_MINIAPP_URL }}
+      # `docker compose build` interpolates the WHOLE compose file, so every :?-guarded
-      DICT_VERSION: ${{ vars.PROD_DICT_VERSION }}
+      # runtime var must be present at build even though it is not a build-arg — incl. the
+      # backend's EXPORT_SIGN_KEY (added with the finished-game export after v1.7.0, which is
+      # why the first v1.8.0 build tripped on it). POSTGRES_PASSWORD/GM_BASICAUTH_HASH above
+      # are here for the same reason; DICT_VERSION + the derived Mini App URL cover the rest.
+      EXPORT_SIGN_KEY: ${{ secrets.PROD_EXPORT_SIGN_KEY }}
+      # PUBLIC_BASE_URL drives the derived VK ID redirect (baked into the SPA) and the
+      # Mini App URL; both are computed in the build step, not stored variables.
+      PUBLIC_BASE_URL: ${{ vars.PROD_PUBLIC_BASE_URL }}
+      DICT_VERSION: ${{ vars.DICT_VERSION }}
     steps:
       - uses: actions/checkout@v4
         with:
@@ -59,6 +69,11 @@ jobs:
         working-directory: deploy
         run: |
           export TAG="${{ steps.ver.outputs.tag }}" APP_VERSION="${{ steps.ver.outputs.tag }}" SCRABBLE_CONFIG_DIR=.
+          # Derive the public URLs from the one canonical origin: the VK ID redirect is a
+          # build-arg baked into the SPA, and the Mini App URL satisfies the (profiled-out)
+          # bot service's compose ":?" guard during parse. See deploy/write-prod-env.sh.
+          base="${PUBLIC_BASE_URL%/}"
+          export VITE_VK_ID_REDIRECT_URL="$base/app/" TELEGRAM_MINIAPP_URL="$base/telegram/"
           # The main-stack images via compose (reuses the build args, incl. VERSION);
           # the bot separately, since it is profiled out of the prod compose.
           docker compose -f docker-compose.yml -f docker-compose.prod.yml build
@@ -83,28 +98,45 @@ jobs:
       GM_BASICAUTH_HASH: ${{ secrets.PROD_GM_BASICAUTH_HASH }}
       GRAFANA_ADMIN_PASSWORD: ${{ secrets.PROD_GRAFANA_ADMIN_PASSWORD }}
       TELEGRAM_BOT_TOKEN: ${{ secrets.PROD_TELEGRAM_BOT_TOKEN }}
-      GATEWAY_VK_APP_SECRET: ${{ secrets.PROD_GATEWAY_VK_APP_SECRET }}
+      GATEWAY_VK_APP_SECRET: ${{ secrets.GATEWAY_VK_APP_SECRET }}
+      # VK ID web login: the "Web" app id (the gateway reuses it as GATEWAY_VK_ID_APP_ID at
+      # runtime) + the app's protected key. Both shared across contours. The redirect URL is
+      # derived from PUBLIC_BASE_URL in deploy/write-prod-env.sh.
+      VITE_VK_APP_ID: ${{ vars.VITE_VK_APP_ID }}
+      GATEWAY_VK_ID_CLIENT_SECRET: ${{ secrets.GATEWAY_VK_ID_CLIENT_SECRET }}
+      # Planted honeytoken bearer: presenting it earns a 24h IP ban + a high-severity alarm.
+      # Per-contour secret; empty = trap off. Rendered by deploy/write-prod-env.sh.
+      GATEWAY_HONEYTOKEN: ${{ secrets.PROD_GATEWAY_HONEYTOKEN }}
       # Signs the finished-game export download URLs (backend BACKEND_EXPORT_SIGN_KEY).
       EXPORT_SIGN_KEY: ${{ secrets.PROD_EXPORT_SIGN_KEY }}
-      # Transactional email via the shared Selectel relay (confirm-codes).
+      # Transactional email via the shared Selectel relay (confirm-codes): one account for
-      SMTP_RELAY_USER: ${{ secrets.PROD_SMTP_RELAY_USER }}
+      # every contour -> unprefixed host/port/tls/user/pass.
-      SMTP_RELAY_PASS: ${{ secrets.PROD_SMTP_RELAY_PASS }}
+      SMTP_RELAY_USER: ${{ secrets.SMTP_RELAY_USER }}
-      SMTP_RELAY_HOST: ${{ vars.PROD_SMTP_RELAY_HOST }}
+      SMTP_RELAY_PASS: ${{ secrets.SMTP_RELAY_PASS }}
-      SMTP_RELAY_PORT: ${{ vars.PROD_SMTP_RELAY_PORT }}
+      SMTP_RELAY_HOST: ${{ vars.SMTP_RELAY_HOST }}
-      SMTP_RELAY_TLS: ${{ vars.PROD_SMTP_RELAY_TLS }}
+      SMTP_RELAY_PORT: ${{ vars.SMTP_RELAY_PORT }}
+      SMTP_RELAY_TLS: ${{ vars.SMTP_RELAY_TLS }}
       SMTP_RELAY_FROM: ${{ vars.PROD_SMTP_RELAY_FROM }}
+      # Operator alerts: backend admin emails + Grafana infra alerts (distinct senders +
+      # recipients; Grafana uses the relay's STARTTLS host:port).
+      SMTP_RELAY_ADMIN_FROM: ${{ vars.PROD_SMTP_RELAY_ADMIN_FROM }}
+      ADMIN_EMAIL: ${{ vars.PROD_ADMIN_EMAIL }}
+      SMTP_RELAY_SERVICE_FROM: ${{ vars.PROD_SMTP_RELAY_SERVICE_FROM }}
+      SERVICE_EMAIL: ${{ vars.PROD_SERVICE_EMAIL }}
+      GRAFANA_SMTP_PORT: ${{ vars.GRAFANA_SMTP_PORT }}
+      GF_SMTP_ENABLED: ${{ vars.PROD_GF_SMTP_ENABLED }}
       PUBLIC_BASE_URL: ${{ vars.PROD_PUBLIC_BASE_URL }}
       PROD_BOTLINK_CA: ${{ secrets.PROD_BOTLINK_CA }}
       PROD_BOTLINK_GATEWAY_CERT: ${{ secrets.PROD_BOTLINK_GATEWAY_CERT }}
       PROD_BOTLINK_GATEWAY_KEY: ${{ secrets.PROD_BOTLINK_GATEWAY_KEY }}
       GM_BASICAUTH_USER: ${{ vars.PROD_GM_BASICAUTH_USER }}
-      GRAFANA_ROOT_URL: ${{ vars.PROD_GRAFANA_ROOT_URL }}
       CADDY_SITE_ADDRESS: ${{ vars.PROD_CADDY_SITE_ADDRESS }}
       LOG_LEVEL: ${{ vars.PROD_LOG_LEVEL }}
-      DICT_VERSION: ${{ vars.PROD_DICT_VERSION }}
+      DICT_VERSION: ${{ vars.DICT_VERSION }}
       POSTGRES_DB: ${{ vars.PROD_POSTGRES_DB }}
       POSTGRES_USER: ${{ vars.PROD_POSTGRES_USER }}
-      TELEGRAM_MINIAPP_URL: ${{ vars.PROD_TELEGRAM_MINIAPP_URL }}
+      # TELEGRAM_MINIAPP_URL and GRAFANA_ROOT_URL are derived from PUBLIC_BASE_URL in
+      # deploy/write-prod-env.sh, not stored variables.
     steps:
       - uses: actions/checkout@v4
         with:
@@ -132,33 +164,8 @@ jobs:
         run: |
           umask 077
           mkdir -p stage/certs-main
-          cat > stage/env.sh <<EOF
+          # Shared with prod-rollback so the two paths render an identical runtime env.
-          export REGISTRY='$REGISTRY'
+          APP_VERSION="$TAG" bash deploy/write-prod-env.sh stage/env.sh
-          export SCRABBLE_CONFIG_DIR='/opt/scrabble'
-          export POSTGRES_DB='${POSTGRES_DB:-scrabble}'
-          export POSTGRES_USER='${POSTGRES_USER:-scrabble}'
-          export POSTGRES_PASSWORD='$POSTGRES_PASSWORD'
-          export GM_BASICAUTH_USER='${GM_BASICAUTH_USER:-gm}'
-          export GM_BASICAUTH_HASH='$GM_BASICAUTH_HASH'
-          export GRAFANA_ADMIN_PASSWORD='$GRAFANA_ADMIN_PASSWORD'
-          export GRAFANA_ROOT_URL='$GRAFANA_ROOT_URL'
-          export CADDY_SITE_ADDRESS='$CADDY_SITE_ADDRESS'
-          export LOG_LEVEL='${LOG_LEVEL:-info}'
-          export DICT_VERSION='$DICT_VERSION'
-          export APP_VERSION='$TAG'
-          export TELEGRAM_BOT_TOKEN='$TELEGRAM_BOT_TOKEN'
-          export TELEGRAM_MINIAPP_URL='$TELEGRAM_MINIAPP_URL'
-          export GATEWAY_VK_APP_SECRET='$GATEWAY_VK_APP_SECRET'
-          export EXPORT_SIGN_KEY='$EXPORT_SIGN_KEY'
-          export SMTP_RELAY_HOST='$SMTP_RELAY_HOST'
-          export SMTP_RELAY_PORT='$SMTP_RELAY_PORT'
-          export SMTP_RELAY_TLS='$SMTP_RELAY_TLS'
-          export SMTP_RELAY_USER='$SMTP_RELAY_USER'
-          export SMTP_RELAY_PASS='$SMTP_RELAY_PASS'
-          export SMTP_RELAY_FROM='$SMTP_RELAY_FROM'
-          export PUBLIC_BASE_URL='$PUBLIC_BASE_URL'
-          export GATEWAY_ABUSE_BAN_ENABLED='true'
-          EOF
           printf '%s\n' "$PROD_BOTLINK_CA"           > stage/certs-main/ca.crt
           printf '%s\n' "$PROD_BOTLINK_GATEWAY_CERT" > stage/certs-main/gateway.crt
           printf '%s\n' "$PROD_BOTLINK_GATEWAY_KEY"  > stage/certs-main/gateway.key
@@ -169,7 +176,7 @@ jobs:
           ssh_main 'mkdir -p /opt/scrabble/compose'
           tar -C deploy -czf - docker-compose.yml docker-compose.prod.yml prod-deploy.sh \
             | ssh_main 'tar -C /opt/scrabble/compose -xzf -'
-          tar -C deploy -czf - caddy otelcol prometheus tempo grafana \
+          tar -C deploy -czf - caddy otelcol prometheus tempo grafana blackbox \
             | ssh_main 'tar -C /opt/scrabble -xzf -'
           tar -C stage -czf - certs-main \
             | ssh_main 'rm -rf /opt/scrabble/certs && mkdir -p /opt/scrabble/certs && tar -C /opt/scrabble/certs --strip-components=1 -xzf -'
@@ -197,7 +204,8 @@ jobs:
       PROD_BOTLINK_BOT_CERT: ${{ secrets.PROD_BOTLINK_BOT_CERT }}
       PROD_BOTLINK_BOT_KEY: ${{ secrets.PROD_BOTLINK_BOT_KEY }}
       LOG_LEVEL: ${{ vars.PROD_LOG_LEVEL }}
-      TELEGRAM_MINIAPP_URL: ${{ vars.PROD_TELEGRAM_MINIAPP_URL }}
+      # PUBLIC_BASE_URL drives the derived Mini App URL in deploy/write-prod-bot-env.sh.
+      PUBLIC_BASE_URL: ${{ vars.PROD_PUBLIC_BASE_URL }}
       TELEGRAM_GAME_CHANNEL_ID: ${{ vars.PROD_TELEGRAM_GAME_CHANNEL_ID }}
       TELEGRAM_CHAT_ID: ${{ vars.PROD_TELEGRAM_CHAT_ID }}
       TELEGRAM_SUPPORT_CHAT_ID: ${{ vars.PROD_TELEGRAM_SUPPORT_CHAT_ID }}
@@ -214,20 +222,8 @@ jobs:
         run: |
           umask 077
           mkdir -p stage/certs-bot
-          cat > stage/env.bot.sh <<EOF
+          # Shared with prod-rollback so the two paths render an identical bot env.
-          export SCRABBLE_CONFIG_DIR='/opt/scrabble'
+          BOT_IMAGE="$REGISTRY/scrabble-telegram-bot:$TAG" bash deploy/write-prod-bot-env.sh stage/env.bot.sh
-          export BOT_IMAGE='$REGISTRY/scrabble-telegram-bot:$TAG'
-          export BOTLINK_GATEWAY_ADDR='$MAIN_HOST:9443'
-          export TELEGRAM_BOT_TOKEN='$TELEGRAM_BOT_TOKEN'
-          export TELEGRAM_MINIAPP_URL='$TELEGRAM_MINIAPP_URL'
-          export TELEGRAM_GAME_CHANNEL_ID='$TELEGRAM_GAME_CHANNEL_ID'
-          export TELEGRAM_CHAT_ID='$TELEGRAM_CHAT_ID'
-          export TELEGRAM_SUPPORT_CHAT_ID='$TELEGRAM_SUPPORT_CHAT_ID'
-          export TELEGRAM_PROMO_BOT_TOKEN='$TELEGRAM_PROMO_BOT_TOKEN'
-          export TELEGRAM_BOT_USERNAME='$TELEGRAM_BOT_USERNAME'
-          export TELEGRAM_BOT_LINK='$TELEGRAM_BOT_LINK'
-          export LOG_LEVEL='${LOG_LEVEL:-info}'
-          EOF
           printf '%s\n' "$PROD_BOTLINK_CA"       > stage/certs-bot/ca.crt
           printf '%s\n' "$PROD_BOTLINK_BOT_CERT" > stage/certs-bot/bot.crt
           printf '%s\n' "$PROD_BOTLINK_BOT_KEY"  > stage/certs-bot/bot.key

.gitea/workflows/prod-rollback.yaml

@@ -51,13 +51,32 @@ jobs:
       PROD_BOTLINK_GATEWAY_CERT: ${{ secrets.PROD_BOTLINK_GATEWAY_CERT }}
       PROD_BOTLINK_GATEWAY_KEY: ${{ secrets.PROD_BOTLINK_GATEWAY_KEY }}
       GM_BASICAUTH_USER: ${{ vars.PROD_GM_BASICAUTH_USER }}
-      GRAFANA_ROOT_URL: ${{ vars.PROD_GRAFANA_ROOT_URL }}
       CADDY_SITE_ADDRESS: ${{ vars.PROD_CADDY_SITE_ADDRESS }}
       LOG_LEVEL: ${{ vars.PROD_LOG_LEVEL }}
-      DICT_VERSION: ${{ vars.PROD_DICT_VERSION }}
+      DICT_VERSION: ${{ vars.DICT_VERSION }}
       POSTGRES_DB: ${{ vars.PROD_POSTGRES_DB }}
       POSTGRES_USER: ${{ vars.PROD_POSTGRES_USER }}
-      TELEGRAM_MINIAPP_URL: ${{ vars.PROD_TELEGRAM_MINIAPP_URL }}
+      PUBLIC_BASE_URL: ${{ vars.PROD_PUBLIC_BASE_URL }}
+      # Full runtime env — parity with prod-deploy's deploy-main so a rollback re-renders
+      # the SAME env.sh (email / VK login / Grafana alerts survive a rollback). TELEGRAM_MINIAPP_URL
+      # and GRAFANA_ROOT_URL are derived from PUBLIC_BASE_URL in deploy/write-prod-env.sh.
+      GATEWAY_VK_APP_SECRET: ${{ secrets.GATEWAY_VK_APP_SECRET }}
+      VITE_VK_APP_ID: ${{ vars.VITE_VK_APP_ID }}
+      GATEWAY_VK_ID_CLIENT_SECRET: ${{ secrets.GATEWAY_VK_ID_CLIENT_SECRET }}
+      GATEWAY_HONEYTOKEN: ${{ secrets.PROD_GATEWAY_HONEYTOKEN }}
+      EXPORT_SIGN_KEY: ${{ secrets.PROD_EXPORT_SIGN_KEY }}
+      SMTP_RELAY_USER: ${{ secrets.SMTP_RELAY_USER }}
+      SMTP_RELAY_PASS: ${{ secrets.SMTP_RELAY_PASS }}
+      SMTP_RELAY_HOST: ${{ vars.SMTP_RELAY_HOST }}
+      SMTP_RELAY_PORT: ${{ vars.SMTP_RELAY_PORT }}
+      SMTP_RELAY_TLS: ${{ vars.SMTP_RELAY_TLS }}
+      SMTP_RELAY_FROM: ${{ vars.PROD_SMTP_RELAY_FROM }}
+      SMTP_RELAY_ADMIN_FROM: ${{ vars.PROD_SMTP_RELAY_ADMIN_FROM }}
+      ADMIN_EMAIL: ${{ vars.PROD_ADMIN_EMAIL }}
+      SMTP_RELAY_SERVICE_FROM: ${{ vars.PROD_SMTP_RELAY_SERVICE_FROM }}
+      SERVICE_EMAIL: ${{ vars.PROD_SERVICE_EMAIL }}
+      GRAFANA_SMTP_PORT: ${{ vars.GRAFANA_SMTP_PORT }}
+      GF_SMTP_ENABLED: ${{ vars.PROD_GF_SMTP_ENABLED }}
       INPUT_TARGET: ${{ inputs.target_version }}
     steps:
       - uses: actions/checkout@v4
@@ -89,24 +108,9 @@ jobs:
         run: |
           umask 077
           mkdir -p stage/certs-main
-          cat > stage/env.sh <<EOF
+          # Same writer as prod-deploy's deploy-main -> the rollback re-renders the FULL
-          export REGISTRY='$REGISTRY'
+          # runtime env (not a subset), so email / VK login / Grafana alerts survive it.
-          export SCRABBLE_CONFIG_DIR='/opt/scrabble'
+          APP_VERSION="$TARGET" bash deploy/write-prod-env.sh stage/env.sh
-          export POSTGRES_DB='${POSTGRES_DB:-scrabble}'
-          export POSTGRES_USER='${POSTGRES_USER:-scrabble}'
-          export POSTGRES_PASSWORD='$POSTGRES_PASSWORD'
-          export GM_BASICAUTH_USER='${GM_BASICAUTH_USER:-gm}'
-          export GM_BASICAUTH_HASH='$GM_BASICAUTH_HASH'
-          export GRAFANA_ADMIN_PASSWORD='$GRAFANA_ADMIN_PASSWORD'
-          export GRAFANA_ROOT_URL='$GRAFANA_ROOT_URL'
-          export CADDY_SITE_ADDRESS='$CADDY_SITE_ADDRESS'
-          export LOG_LEVEL='${LOG_LEVEL:-info}'
-          export DICT_VERSION='$DICT_VERSION'
-          export APP_VERSION='$TARGET'
-          export TELEGRAM_BOT_TOKEN='$TELEGRAM_BOT_TOKEN'
-          export TELEGRAM_MINIAPP_URL='$TELEGRAM_MINIAPP_URL'
-          export GATEWAY_ABUSE_BAN_ENABLED='true'
-          EOF
           printf '%s\n' "$PROD_BOTLINK_CA"           > stage/certs-main/ca.crt
           printf '%s\n' "$PROD_BOTLINK_GATEWAY_CERT" > stage/certs-main/gateway.crt
           printf '%s\n' "$PROD_BOTLINK_GATEWAY_KEY"  > stage/certs-main/gateway.key
@@ -147,9 +151,11 @@ jobs:
       PROD_BOTLINK_BOT_CERT: ${{ secrets.PROD_BOTLINK_BOT_CERT }}
       PROD_BOTLINK_BOT_KEY: ${{ secrets.PROD_BOTLINK_BOT_KEY }}
       LOG_LEVEL: ${{ vars.PROD_LOG_LEVEL }}
-      TELEGRAM_MINIAPP_URL: ${{ vars.PROD_TELEGRAM_MINIAPP_URL }}
+      # PUBLIC_BASE_URL drives the derived Mini App URL; SUPPORT_CHAT_ID for parity with deploy.
+      PUBLIC_BASE_URL: ${{ vars.PROD_PUBLIC_BASE_URL }}
       TELEGRAM_GAME_CHANNEL_ID: ${{ vars.PROD_TELEGRAM_GAME_CHANNEL_ID }}
       TELEGRAM_CHAT_ID: ${{ vars.PROD_TELEGRAM_CHAT_ID }}
+      TELEGRAM_SUPPORT_CHAT_ID: ${{ vars.PROD_TELEGRAM_SUPPORT_CHAT_ID }}
       TELEGRAM_BOT_USERNAME: ${{ vars.PROD_TELEGRAM_BOT_USERNAME }}
       TELEGRAM_BOT_LINK: ${{ vars.PROD_VITE_TELEGRAM_LINK }}
     steps:
@@ -163,19 +169,8 @@ jobs:
         run: |
           umask 077
           mkdir -p stage/certs-bot
-          cat > stage/env.bot.sh <<EOF
+          # Same writer as prod-deploy's deploy-bot (parity; includes TELEGRAM_SUPPORT_CHAT_ID).
-          export SCRABBLE_CONFIG_DIR='/opt/scrabble'
+          BOT_IMAGE="$REGISTRY/scrabble-telegram-bot:$TARGET" bash deploy/write-prod-bot-env.sh stage/env.bot.sh
-          export BOT_IMAGE='$REGISTRY/scrabble-telegram-bot:$TARGET'
-          export BOTLINK_GATEWAY_ADDR='$MAIN_HOST:9443'
-          export TELEGRAM_BOT_TOKEN='$TELEGRAM_BOT_TOKEN'
-          export TELEGRAM_MINIAPP_URL='$TELEGRAM_MINIAPP_URL'
-          export TELEGRAM_GAME_CHANNEL_ID='$TELEGRAM_GAME_CHANNEL_ID'
-          export TELEGRAM_CHAT_ID='$TELEGRAM_CHAT_ID'
-          export TELEGRAM_PROMO_BOT_TOKEN='$TELEGRAM_PROMO_BOT_TOKEN'
-          export TELEGRAM_BOT_USERNAME='$TELEGRAM_BOT_USERNAME'
-          export TELEGRAM_BOT_LINK='$TELEGRAM_BOT_LINK'
-          export LOG_LEVEL='${LOG_LEVEL:-info}'
-          EOF
           printf '%s\n' "$PROD_BOTLINK_CA"       > stage/certs-bot/ca.crt
           printf '%s\n' "$PROD_BOTLINK_BOT_CERT" > stage/certs-bot/bot.crt
           printf '%s\n' "$PROD_BOTLINK_BOT_KEY"  > stage/certs-bot/bot.key

backend/cmd/backend/main.go

@@ -12,6 +12,7 @@ import (
 	"fmt"
 	"log"
 	"os/signal"
+	"strings"
 	"syscall"
 	"time"
 
@@ -20,6 +21,7 @@ import (
 
 	"scrabble/backend/internal/account"
 	"scrabble/backend/internal/accountmerge"
+	"scrabble/backend/internal/adminalert"
 	"scrabble/backend/internal/ads"
 	"scrabble/backend/internal/banview"
 	"scrabble/backend/internal/config"
@@ -44,6 +46,10 @@ import (
 // telemetryShutdownTimeout bounds the OpenTelemetry flush during process exit.
 const telemetryShutdownTimeout = 5 * time.Second
 
+// adminAlertInterval is how often the operator-alert worker checks for new feedback /
+// complaints; a burst within one interval coalesces into a single digest email.
+const adminAlertInterval = 5 * time.Minute
+
 func main() {
 	cfg, err := config.Load()
 	if err != nil {
@@ -207,6 +213,18 @@ func run(ctx context.Context, cfg config.Config, logger *zap.Logger) error {
 	feedbackSvc := feedback.NewService(feedback.NewStore(db), accounts)
 	feedbackSvc.SetNotifier(hub)
 
+	// Operator alert emails on new feedback / word complaints, coalesced into one digest
+	// per interval. Inert unless a distinct admin sender and recipient are configured.
+	if cfg.SMTP.AdminFrom != "" && cfg.SMTP.AdminTo != "" {
+		consoleURL := ""
+		if cfg.PublicBaseURL != "" {
+			consoleURL = strings.TrimRight(cfg.PublicBaseURL, "/") + "/_gm"
+		}
+		alerts := adminalert.New(mailer, feedbackSvc, games, cfg.SMTP.AdminFrom, cfg.SMTP.AdminTo, consoleURL, logger)
+		go alerts.Run(ctx, adminAlertInterval)
+		logger.Info("admin alert worker started", zap.Duration("interval", adminAlertInterval))
+	}
+
 	// Robot opponent: provision its durable account pool (a hard startup
 	// dependency, like the dictionaries) and start its move driver. The matchmaker
 	// substitutes a pooled robot for a missing human after the wait window.

backend/internal/account/email.go

@@ -206,6 +206,11 @@ func (s *EmailService) ConfirmCode(ctx context.Context, accountID uuid.UUID, ema
 	if err := s.store.confirmEmailIdentity(ctx, conf.id, accountID, addr, s.now()); err != nil {
 		return Account{}, err
 	}
+	// Binding the first confirmed email promotes a guest to a durable account, matching the
+	// link and deeplink flows (defence-in-depth: no confirmed-email path leaves is_guest set).
+	if err := s.store.ClearGuest(ctx, accountID); err != nil {
+		return Account{}, err
+	}
 	return s.store.GetByID(ctx, accountID)
 }
 

backend/internal/account/mailer.go

@@ -15,7 +15,11 @@ import (
 // required plain-text body and doubles as the multipart/alternative fallback;
 // HTML, when non-empty, is the preferred body a capable client renders instead.
 type Message struct {
+	// To is the recipient address, or several comma-separated (all get the one message).
 	To string
+	// From, when non-empty, overrides the configured sender for this message — the admin
+	// alert path uses a distinct From from the user-facing confirm-code sender.
+	From    string
 	Subject string
 	Text    string
 	HTML    string
@@ -29,6 +33,18 @@ type Mailer interface {
 	Send(ctx context.Context, msg Message) error
 }
 
+// splitAddrs splits a comma-separated recipient list into trimmed, non-empty addresses.
+func splitAddrs(list string) []string {
+	parts := strings.Split(list, ",")
+	out := make([]string, 0, len(parts))
+	for _, p := range parts {
+		if a := strings.TrimSpace(p); a != "" {
+			out = append(out, a)
+		}
+	}
+	return out
+}
+
 // SMTPConfig configures the SMTP relay. An empty Host selects the LogMailer
 // instead, so a deployment without a relay still runs (the code lands in the log).
 // TLS is always used and no client certificate is required — only the server
@@ -44,6 +60,11 @@ type SMTPConfig struct {
 	// port (implicit TLS on 465, STARTTLS otherwise); set it explicitly for a relay on
 	// a non-standard port (e.g. Selectel's 1127 = SSL, 1126 = STARTTLS).
 	TLS string
+	// AdminFrom / AdminTo drive the operator alert emails (new feedback / word complaints),
+	// distinct from the user-facing confirm-code sender. AdminTo may be several
+	// comma-separated addresses. Both empty disables the alert worker.
+	AdminFrom string
+	AdminTo   string
 }
 
 const (
@@ -110,10 +131,16 @@ func (m SMTPMailer) Send(ctx context.Context, msg Message) error {
 		return fmt.Errorf("account: build mail client: %w", err)
 	}
 	out := mail.NewMsg()
-	if err := out.From(m.cfg.From); err != nil {
+	from := m.cfg.From
-		return fmt.Errorf("account: set From %q: %w", m.cfg.From, err)
+	if msg.From != "" {
+		from = msg.From
 	}
-	if err := out.To(msg.To); err != nil {
+	if err := out.From(from); err != nil {
+		return fmt.Errorf("account: set From %q: %w", from, err)
+	}
+	// To may carry several comma-separated recipients; go-mail wants them as separate
+	// arguments (a single joined string parses as one malformed address).
+	if err := out.To(splitAddrs(msg.To)...); err != nil {
 		return fmt.Errorf("account: set To %q: %w", msg.To, err)
 	}
 	out.Subject(msg.Subject)

backend/internal/account/mailer_test.go

@@ -1,6 +1,28 @@
 package account
 
-import "testing"
+import (
+	"slices"
+	"testing"
+)
+
+// TestSplitAddrs covers the comma-separated recipient parsing used for the admin alert
+// To (several operator mailboxes in one message), including trimming and empty entries.
+func TestSplitAddrs(t *testing.T) {
+	cases := []struct {
+		in   string
+		want []string
+	}{
+		{"a@x.ru", []string{"a@x.ru"}},
+		{"a@x.ru, b@y.ru", []string{"a@x.ru", "b@y.ru"}},
+		{" a@x.ru ,, b@y.ru ,", []string{"a@x.ru", "b@y.ru"}},
+		{"", nil},
+	}
+	for _, c := range cases {
+		if got := splitAddrs(c.in); !slices.Equal(got, c.want) {
+			t.Errorf("splitAddrs(%q) = %v, want %v", c.in, got, c.want)
+		}
+	}
+}
 
 // TestSMTPTLSMode covers the explicit TLS mode and the port-based fallback, including
 // the non-standard Selectel ports (1127 = SSL, 1126 = STARTTLS) that the 465 heuristic

backend/internal/account/retention.go

@@ -21,9 +21,11 @@ import (
 const RetentionTTL = 2 * 365 * 24 * time.Hour
 
 // Reasons recorded on a retained_identities row: what detached the credential from its
-// account. The row is written just before the live identities row is removed, preserving
+// account (unlink / email change / account deletion here; an account merge that drops a
-// the legal dossier (which email/vk/tg was linked, and when) even as the identity frees
+// same-kind colliding identity writes reason "merge" from the accountmerge package). The
-// for reuse. See docs/ARCHITECTURE.md §9.1.
+// row is written just before the live identities row is removed, preserving the legal
+// dossier (which email/vk/tg was linked, and when) even as the identity frees for reuse.
+// See docs/ARCHITECTURE.md §9.1.
 const (
 	retainUnlink = "unlink"
 	retainChange = "change"

backend/internal/account/userlist.go

@@ -19,6 +19,9 @@ type UserListItem struct {
 	PreferredLanguage string
 	IsGuest           bool
 	IsRobot           bool
+	// IsDeleted marks a tombstoned account (deleted_at set), shown as a badge — a search
+	// spans both lists, so a result can be either live or deleted.
+	IsDeleted bool
 	// FlaggedHighRateAt is the soft high-rate marker (zero when unflagged), shown
 	// as a badge in the console list.
 	FlaggedHighRateAt time.Time
@@ -26,12 +29,14 @@ type UserListItem struct {
 }
 
 // UserFilter narrows the admin user list: Robots selects robot accounts (otherwise the
-// non-robot "people"); NameMask and ExternalIDMask are glob masks ('*' = any run, '?' =
+// non-robot "people"); Deleted selects tombstoned accounts (every other scope hides them);
-// one char) matched case-insensitively against the display name / any identity's external
+// NameMask and ExternalIDMask are glob masks ('*' = any run, '?' = one char) matched
-// id; EmailExact is a strict (exact) match against an account's email identity. An empty
+// case-insensitively against the display name / any identity's external id; EmailExact is a
-// value means no filter on that field.
+// strict (exact) match against an account's email identity. An empty value means no filter
+// on that field.
 type UserFilter struct {
 	Robots         bool
+	Deleted        bool
 	NameMask       string
 	ExternalIDMask string
 	EmailExact     string
@@ -53,21 +58,42 @@ func (s *Store) IsRobot(ctx context.Context, accountID uuid.UUID) (bool, error)
 	return ok, nil
 }
 
-// userListWhere builds the shared WHERE clause and its positional args (from $1).
+// userListWhere builds the shared WHERE clause and its positional args (from $1). On the
+// Robots tab it lists/searches robots only. Otherwise a search (any of the name /
+// external-id / email filters) spans live and deleted people alike — never robots — so the
+// operator finds a match from one query regardless of the People / Deleted tab; the search
+// also looks in the retention journal, so a deleted account is still found by the email /
+// external id it held (those rows moved from identities to retained_identities on deletion)
+// and by its retained real name. With no search, the People / Deleted tab scope applies.
 func userListWhere(f UserFilter) (string, []any) {
-	args := []any{f.Robots}
+	name := LikePattern(f.NameMask)
-	where := robotExists + ` = $1`
+	ext := LikePattern(f.ExternalIDMask)
-	if name := LikePattern(f.NameMask); name != "" {
+	email := strings.ToLower(strings.TrimSpace(f.EmailExact))
+	searching := name != "" || ext != "" || email != ""
+
+	var args []any
+	var where string
+	switch {
+	case f.Robots:
+		where = robotExists + ` = true`
+	case searching:
+		where = robotExists + ` = false`
+	case f.Deleted:
+		where = robotExists + ` = false AND a.deleted_at IS NOT NULL`
+	default:
+		where = robotExists + ` = false AND a.deleted_at IS NULL`
+	}
+	if name != "" {
 		args = append(args, name)
-		where += fmt.Sprintf(` AND a.display_name ILIKE $%d ESCAPE '\'`, len(args))
+		where += fmt.Sprintf(` AND (a.display_name ILIKE $%d ESCAPE '\' OR a.deleted_display_name ILIKE $%d ESCAPE '\')`, len(args), len(args))
 	}
-	if ext := LikePattern(f.ExternalIDMask); ext != "" {
+	if ext != "" {
 		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))
 	}
-	if email := strings.ToLower(strings.TrimSpace(f.EmailExact)); email != "" {
+	if email != "" {
 		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)`, len(args))
+		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))
 	}
 	return where, args
 }
@@ -75,7 +101,7 @@ func userListWhere(f UserFilter) (string, []any) {
 // ListUsers returns the filtered admin user list, newest first, paginated.
 func (s *Store) ListUsers(ctx context.Context, f UserFilter, limit, offset int) ([]UserListItem, error) {
 	where, args := userListWhere(f)
-	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)
@@ -88,7 +114,7 @@ FROM backend.accounts a WHERE ` + where +
 	for rows.Next() {
 		var it UserListItem
 		var flagged sql.NullTime
-		if err := rows.Scan(&it.ID, &it.DisplayName, &it.PreferredLanguage, &it.IsGuest, &flagged, &it.CreatedAt, &it.IsRobot); err != nil {
+		if err := rows.Scan(&it.ID, &it.DisplayName, &it.PreferredLanguage, &it.IsGuest, &flagged, &it.CreatedAt, &it.IsRobot, &it.IsDeleted); err != nil {
 			return nil, fmt.Errorf("account: scan user: %w", err)
 		}
 		if flagged.Valid {

backend/internal/accountmerge/merge.go

@@ -27,6 +27,11 @@ import (
 // without taking a dependency on the game package.
 const statusActive = "active"
 
+// retainReasonMerge is the retained_identities.reason for a credential dropped by a merge
+// collision (both accounts held the same kind). It mirrors the account package's retain
+// reasons, kept local to avoid importing that package's unexported constants.
+const retainReasonMerge = "merge"
+
 // Friendship statuses, highest precedence first, mirroring internal/social.
 const (
 	friendAccepted = "accepted"
@@ -75,6 +80,9 @@ func (m *Merger) Merge(ctx context.Context, primary, secondary uuid.UUID) error
 		if err := mergeAccountFields(ctx, tx, primary, secondary, now); err != nil {
 			return err
 		}
+		if err := dedupeIdentities(ctx, tx, primary, secondary); err != nil {
+			return err
+		}
 		if err := reassignColumn(ctx, tx, table.Identities, table.Identities.AccountID, primary, secondary); err != nil {
 			return fmt.Errorf("accountmerge: identities: %w", err)
 		}
@@ -300,6 +308,77 @@ func reassignColumn(ctx context.Context, tx *sql.Tx, tbl postgres.Table, col pos
 	return err
 }
 
+// dedupeIdentities resolves a same-kind identity collision before the blanket identity
+// reassign: when both accounts already hold an identity of the same kind (e.g. each has a
+// confirmed email — reachable when two email-bearing accounts merge), the primary keeps
+// its own and the secondary's is journaled to retained_identities (reason=merge) and
+// removed. Without this the blanket reassign would leave the survivor with two identities
+// of one kind (there is no per-account-kind unique on identities), which the profile and
+// the retention dossier both treat as singular. Non-colliding identities are untouched and
+// move with the blanket reassign.
+func dedupeIdentities(ctx context.Context, tx *sql.Tx, primary, secondary uuid.UUID) error {
+	var prows []model.Identities
+	if err := postgres.SELECT(table.Identities.Kind).
+		FROM(table.Identities).
+		WHERE(table.Identities.AccountID.EQ(postgres.UUID(primary))).
+		QueryContext(ctx, tx, &prows); err != nil && !errors.Is(err, qrm.ErrNoRows) {
+		return fmt.Errorf("accountmerge: primary identity kinds: %w", err)
+	}
+	occupied := make(map[string]struct{}, len(prows))
+	for _, r := range prows {
+		occupied[r.Kind] = struct{}{}
+	}
+	if len(occupied) == 0 {
+		return nil
+	}
+	var srows []model.Identities
+	if err := postgres.SELECT(
+		table.Identities.Kind, table.Identities.ExternalID,
+		table.Identities.Confirmed, table.Identities.CreatedAt,
+	).FROM(table.Identities).
+		WHERE(table.Identities.AccountID.EQ(postgres.UUID(secondary))).
+		QueryContext(ctx, tx, &srows); err != nil && !errors.Is(err, qrm.ErrNoRows) {
+		return fmt.Errorf("accountmerge: secondary identities: %w", err)
+	}
+	for _, s := range srows {
+		if _, dup := occupied[s.Kind]; !dup {
+			continue
+		}
+		if err := retainMergedIdentity(ctx, tx, secondary, s); err != nil {
+			return err
+		}
+		del := table.Identities.DELETE().WHERE(
+			table.Identities.AccountID.EQ(postgres.UUID(secondary)).
+				AND(table.Identities.Kind.EQ(postgres.String(s.Kind))).
+				AND(table.Identities.ExternalID.EQ(postgres.String(s.ExternalID))),
+		)
+		if _, err := del.ExecContext(ctx, tx); err != nil {
+			return fmt.Errorf("accountmerge: drop colliding %s identity: %w", s.Kind, err)
+		}
+	}
+	return nil
+}
+
+// retainMergedIdentity appends a retained_identities row for a secondary identity dropped
+// by a merge collision (reason=merge), preserving it in the legal dossier. It mirrors
+// account.retainIdentityTx, which is unexported; detached_at falls to the column default.
+func retainMergedIdentity(ctx context.Context, tx *sql.Tx, accountID uuid.UUID, id model.Identities) error {
+	rid, err := uuid.NewV7()
+	if err != nil {
+		return fmt.Errorf("accountmerge: new retained id: %w", err)
+	}
+	ins := table.RetainedIdentities.INSERT(
+		table.RetainedIdentities.RetainedID, table.RetainedIdentities.AccountID,
+		table.RetainedIdentities.Kind, table.RetainedIdentities.ExternalID,
+		table.RetainedIdentities.Confirmed, table.RetainedIdentities.LinkedAt,
+		table.RetainedIdentities.Reason,
+	).VALUES(rid, accountID, id.Kind, id.ExternalID, id.Confirmed, id.CreatedAt, retainReasonMerge)
+	if _, err := ins.ExecContext(ctx, tx); err != nil {
+		return fmt.Errorf("accountmerge: retain merged %s identity: %w", id.Kind, err)
+	}
+	return nil
+}
+
 // friendRank ranks a friendship status for dedupe precedence (higher wins).
 func friendRank(status string) int {
 	switch status {

backend/internal/adminalert/notify.go

@@ -0,0 +1,116 @@
+// Package adminalert emails the operator when new player feedback or word complaints
+// arrive, coalescing a burst into a single digest per interval so a flood is one email,
+// not N. It is inert unless an admin sender and recipient are configured. The sender is
+// distinct from the user-facing confirm-code From, and the recipient may be several
+// comma-separated addresses (the mailer splits them).
+package adminalert
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"time"
+
+	"go.uber.org/zap"
+
+	"scrabble/backend/internal/account"
+)
+
+// FeedbackCounter counts feedback created since a time (satisfied by feedback.Service).
+type FeedbackCounter interface {
+	CountSince(ctx context.Context, since time.Time) (int, error)
+}
+
+// ComplaintCounter counts word complaints filed since a time (satisfied by game.Service).
+type ComplaintCounter interface {
+	CountComplaintsSince(ctx context.Context, since time.Time) (int, error)
+}
+
+// Notifier polls for new feedback and complaints and emails the operator a digest.
+type Notifier struct {
+	mailer     account.Mailer
+	feedback   FeedbackCounter
+	complaints ComplaintCounter
+	from       string
+	to         string
+	consoleURL string
+	clock      func() time.Time
+	log        *zap.Logger
+	last       time.Time
+}
+
+// New constructs a Notifier. from and to are the alert sender and recipient(s); consoleURL,
+// when non-empty, is the admin-console link included in the email. log may be nil. The
+// watermark starts at "now", so only items arriving after start-up are reported.
+func New(mailer account.Mailer, fb FeedbackCounter, cp ComplaintCounter, from, to, consoleURL string, log *zap.Logger) *Notifier {
+	if log == nil {
+		log = zap.NewNop()
+	}
+	return &Notifier{
+		mailer: mailer, feedback: fb, complaints: cp, from: from, to: to, consoleURL: consoleURL,
+		clock: func() time.Time { return time.Now().UTC() }, log: log, last: time.Now().UTC(),
+	}
+}
+
+// Run polls on each tick until ctx is cancelled.
+func (n *Notifier) Run(ctx context.Context, interval time.Duration) {
+	ticker := time.NewTicker(interval)
+	defer ticker.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-ticker.C:
+			n.tick(ctx)
+		}
+	}
+}
+
+// tick counts what arrived since the last watermark and, if anything did, emails one
+// digest. The watermark only advances after a successful send (or a quiet tick), so a
+// transient send failure is retried on the next tick — the counts simply grow.
+func (n *Notifier) tick(ctx context.Context) {
+	now := n.clock()
+	fb, err := n.feedback.CountSince(ctx, n.last)
+	if err != nil {
+		n.log.Warn("admin alert: count feedback failed", zap.Error(err))
+		return
+	}
+	cp, err := n.complaints.CountComplaintsSince(ctx, n.last)
+	if err != nil {
+		n.log.Warn("admin alert: count complaints failed", zap.Error(err))
+		return
+	}
+	if fb == 0 && cp == 0 {
+		n.last = now
+		return
+	}
+	if err := n.mailer.Send(ctx, n.digest(fb, cp)); err != nil {
+		n.log.Warn("admin alert: send failed", zap.Error(err))
+		return
+	}
+	n.log.Info("admin alert sent", zap.Int("feedback", fb), zap.Int("complaints", cp))
+	n.last = now
+}
+
+// digest builds the operator alert email for fb new feedback and cp new complaints.
+func (n *Notifier) digest(fb, cp int) account.Message {
+	var parts []string
+	if fb > 0 {
+		parts = append(parts, fmt.Sprintf("%d new feedback message(s)", fb))
+	}
+	if cp > 0 {
+		parts = append(parts, fmt.Sprintf("%d new word complaint(s)", cp))
+	}
+	summary := strings.Join(parts, ", ")
+	text := summary + "."
+	if n.consoleURL != "" {
+		text += "\n\nOpen the admin console: " + n.consoleURL
+	}
+	return account.Message{
+		From:    n.from,
+		To:      n.to,
+		Subject: "Erudit — " + summary,
+		Text:    text,
+	}
+}

backend/internal/adminalert/notify_test.go

@@ -0,0 +1,55 @@
+package adminalert
+
+import (
+	"context"
+	"strings"
+	"testing"
+	"time"
+
+	"scrabble/backend/internal/account"
+)
+
+// The fakes ignore the watermark and return a fixed count, which is all the digest logic
+// needs.
+type fbCounter struct{ n int }
+
+func (f fbCounter) CountSince(context.Context, time.Time) (int, error) { return f.n, nil }
+
+type cpCounter struct{ n int }
+
+func (c cpCounter) CountComplaintsSince(context.Context, time.Time) (int, error) { return c.n, nil }
+
+type recordingMailer struct{ sent []account.Message }
+
+func (m *recordingMailer) Send(_ context.Context, msg account.Message) error {
+	m.sent = append(m.sent, msg)
+	return nil
+}
+
+func TestNotifierSkipsWhenNothingNew(t *testing.T) {
+	mailer := &recordingMailer{}
+	n := New(mailer, fbCounter{0}, cpCounter{0}, "alerts@erudit-game.ru", "op@x.ru", "", nil)
+	n.tick(context.Background())
+	if len(mailer.sent) != 0 {
+		t.Fatalf("sent %d emails, want 0 when nothing is new", len(mailer.sent))
+	}
+}
+
+func TestNotifierDigestsNewItems(t *testing.T) {
+	mailer := &recordingMailer{}
+	n := New(mailer, fbCounter{2}, cpCounter{1}, "alerts@erudit-game.ru", "op@x.ru, two@x.ru", "https://erudit-game.ru/_gm", nil)
+	n.tick(context.Background())
+	if len(mailer.sent) != 1 {
+		t.Fatalf("sent %d emails, want 1 digest", len(mailer.sent))
+	}
+	msg := mailer.sent[0]
+	if msg.From != "alerts@erudit-game.ru" || msg.To != "op@x.ru, two@x.ru" {
+		t.Errorf("digest addressing = From %q To %q", msg.From, msg.To)
+	}
+	if !strings.Contains(msg.Subject, "2 new feedback") || !strings.Contains(msg.Subject, "1 new word complaint") {
+		t.Errorf("digest subject = %q, want the feedback + complaint counts", msg.Subject)
+	}
+	if !strings.Contains(msg.Text, "/_gm") {
+		t.Errorf("digest body = %q, want the console link", msg.Text)
+	}
+}

backend/internal/adminconsole/templates/pages/user_detail.gohtml

@@ -119,11 +119,11 @@
 </tbody>
 </table>
 {{end}}
-{{if not .Deleted}}{{if not .Guest}}
+{{if not .Deleted}}
 <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}}{{end}}
+{{end}}
 </section>
 <section class="panel"><h2>Friends</h2>
 <table class="list">

backend/internal/adminconsole/templates/pages/users.gohtml

@@ -2,11 +2,12 @@
 <h1>Users</h1>
 {{with .Data}}
 <nav class="subnav">
-<a href="/_gm/users"{{if not .Robots}} class="active"{{end}}>People</a> ·
+<a href="/_gm/users"{{if and (not .Robots) (not .Deleted)}} class="active"{{end}}>People</a> ·
+<a href="/_gm/users?kind=deleted"{{if .Deleted}} class="active"{{end}}>Deleted</a> ·
 <a href="/_gm/users?kind=robots"{{if .Robots}} class="active"{{end}}>Robots</a>
 </nav>
 <form class="form" method="get" action="/_gm/users">
-{{if .Robots}}<input type="hidden" name="kind" value="robots">{{end}}
+{{if .Robots}}<input type="hidden" name="kind" value="robots">{{end}}{{if .Deleted}}<input type="hidden" name="kind" value="deleted">{{end}}
 <input name="name" value="{{.NameMask}}" placeholder="display name mask (* ?)">
 <input name="ext" value="{{.ExternalIDMask}}" placeholder="external id mask (* ?)">
 <input name="email" value="{{.EmailExact}}" placeholder="email (exact)" type="search">
@@ -18,7 +19,7 @@
 {{range .Items}}
 <tr>
 <td><a href="/_gm/users/{{.ID}}">{{.ID}}</a></td>
-<td>{{.DisplayName}}{{if .Guest}} <span class="pill">guest</span>{{end}}{{if .FlaggedHighRate}} <span class="pill">high-rate</span>{{end}}</td>
+<td>{{.DisplayName}}{{if .Deleted}} <span class="pill">deleted</span>{{end}}{{if .Guest}} <span class="pill">guest</span>{{end}}{{if .FlaggedHighRate}} <span class="pill">high-rate</span>{{end}}</td>
 <td>{{.Kind}}</td>
 <td>{{.Language}}</td>
 <td>{{.CreatedAt}}</td>

backend/internal/adminconsole/views.go

@@ -60,6 +60,7 @@ type UsersView struct {
 	// be emitted verbatim — interpolated as a plain string it would have its "=" and "&"
 	// percent-encoded again by the contextual escaper.
 	Robots         bool
+	Deleted        bool
 	NameMask       string
 	ExternalIDMask string
 	EmailExact     string
@@ -75,6 +76,7 @@ type UserRow struct {
 	Kind            string
 	Language        string
 	Guest           bool
+	Deleted         bool
 	FlaggedHighRate bool
 	CreatedAt       string
 	HasMoveStats    bool

backend/internal/config/config.go

@@ -149,6 +149,8 @@ func Load() (Config, error) {
 		Password:  os.Getenv("BACKEND_SMTP_PASSWORD"),
 		From:      envOr("BACKEND_SMTP_FROM", "no-reply@localhost"),
 		TLS:       os.Getenv("BACKEND_SMTP_TLS"),
+		AdminFrom: os.Getenv("BACKEND_SMTP_ADMIN_FROM"),
+		AdminTo:   os.Getenv("BACKEND_ADMIN_EMAIL"),
 	}
 
 	c := Config{

backend/internal/feedback/service.go

@@ -195,6 +195,11 @@ func (svc *Service) CountUnread(ctx context.Context) (int, error) {
 	return svc.store.CountUnread(ctx)
 }
 
+// CountSince counts feedback created after since, for the operator alert worker.
+func (svc *Service) CountSince(ctx context.Context, since time.Time) (int, error) {
+	return svc.store.CountSince(ctx, since)
+}
+
 // Attachment returns a message's file name and bytes, reporting false when absent.
 func (svc *Service) Attachment(ctx context.Context, id uuid.UUID) (string, []byte, bool, error) {
 	return svc.store.Attachment(ctx, id)

backend/internal/feedback/store.go

@@ -386,3 +386,15 @@ func (s *Store) CountUnread(ctx context.Context) (int, error) {
 	}
 	return n, nil
 }
+
+// CountSince counts feedback messages created strictly after since — the operator alert
+// worker's "new since the last check" signal.
+func (s *Store) CountSince(ctx context.Context, since time.Time) (int, error) {
+	var n int
+	if err := s.db.QueryRowContext(ctx,
+		`SELECT COUNT(*) FROM backend.feedback_messages WHERE created_at > $1`, since,
+	).Scan(&n); err != nil {
+		return 0, fmt.Errorf("feedback: count since: %w", err)
+	}
+	return n, nil
+}

backend/internal/game/service.go

@@ -1008,6 +1008,12 @@ func (svc *Service) CountComplaints(ctx context.Context, status string) (int, er
 	return svc.store.CountComplaints(ctx, status)
 }
 
+// CountComplaintsSince counts word complaints filed after since, for the operator alert
+// worker.
+func (svc *Service) CountComplaintsSince(ctx context.Context, since time.Time) (int, error) {
+	return svc.store.CountComplaintsSince(ctx, since)
+}
+
 // ResolveComplaint closes a complaint with an operator disposition (reject /
 // accept_add / accept_remove) and an optional note. An accepted complaint then
 // appears in DictionaryChanges until a rebuilt dictionary is loaded and the

backend/internal/game/store.go

@@ -1040,6 +1040,19 @@ func (s *Store) CountComplaints(ctx context.Context, status string) (int, error)
 	return int(dest.Count), nil
 }
 
+// CountComplaintsSince counts word complaints filed strictly after since — the operator
+// alert worker's "new since the last check" signal.
+func (s *Store) CountComplaintsSince(ctx context.Context, since time.Time) (int, error) {
+	stmt := postgres.SELECT(postgres.COUNT(table.Complaints.ComplaintID).AS("count")).
+		FROM(table.Complaints).
+		WHERE(table.Complaints.CreatedAt.GT(postgres.TimestampzT(since)))
+	var dest struct{ Count int64 }
+	if err := stmt.QueryContext(ctx, s.db, &dest); err != nil {
+		return 0, fmt.Errorf("game: count complaints since: %w", err)
+	}
+	return int(dest.Count), nil
+}
+
 // ActiveGames returns the turn clocks of every in-progress game; the sweeper
 // filters them against the per-move deadline and the player's away window.
 func (s *Store) ActiveGames(ctx context.Context) ([]activeGame, error) {

backend/internal/inttest/delete_test.go

@@ -134,6 +134,108 @@ func TestDeletionDossierReaders(t *testing.T) {
 	}
 }
 
+// listHasID reports whether the user list contains accountID.
+func listHasID(items []account.UserListItem, id uuid.UUID) bool {
+	for _, it := range items {
+		if it.ID == id {
+			return true
+		}
+	}
+	return false
+}
+
+// TestUserListDeletedFilter: a tombstoned account is hidden from the default People list and
+// shown only under the Deleted scope.
+func TestUserListDeletedFilter(t *testing.T) {
+	ctx := context.Background()
+	store := account.NewStore(testDB)
+	deleter := accountdelete.NewDeleter(testDB)
+
+	live, _, err := store.ProvisionTelegram(ctx, "tg-"+uuid.NewString(), "en", "", "Live", "")
+	if err != nil {
+		t.Fatalf("provision live: %v", err)
+	}
+	goneTg := "tg-" + uuid.NewString()
+	gone, _, err := store.ProvisionTelegram(ctx, goneTg, "en", "", "Gone", "")
+	if err != nil {
+		t.Fatalf("provision gone: %v", err)
+	}
+	goneEmail := "gone-" + uuid.NewString() + "@example.com"
+	if err := store.AttachIdentity(ctx, gone.ID, account.KindEmail, goneEmail, true); err != nil {
+		t.Fatalf("attach gone email: %v", err)
+	}
+	if err := deleter.AnonymizeAndTombstone(ctx, gone.ID); err != nil {
+		t.Fatalf("delete: %v", err)
+	}
+
+	people, err := store.ListUsers(ctx, account.UserFilter{}, 5000, 0)
+	if err != nil {
+		t.Fatalf("list people: %v", err)
+	}
+	if listHasID(people, gone.ID) {
+		t.Error("a deleted account must not appear in the default People list")
+	}
+	if !listHasID(people, live.ID) {
+		t.Error("a live account must appear in the default People list")
+	}
+	deleted, err := store.ListUsers(ctx, account.UserFilter{Deleted: true}, 5000, 0)
+	if err != nil {
+		t.Fatalf("list deleted: %v", err)
+	}
+	if !listHasID(deleted, gone.ID) {
+		t.Error("a deleted account must appear in the Deleted list")
+	}
+	if listHasID(deleted, live.ID) {
+		t.Error("a live account must not appear in the Deleted list")
+	}
+
+	// A search spans both lists and reaches the retention journal: a deleted account is
+	// still found by the email and external id it held (both moved to retained_identities
+	// on deletion, out of the live identities table).
+	byEmail, err := store.ListUsers(ctx, account.UserFilter{EmailExact: goneEmail}, 5000, 0)
+	if err != nil {
+		t.Fatalf("search by email: %v", err)
+	}
+	if !listHasID(byEmail, gone.ID) {
+		t.Error("a deleted account must be found by the email it held (retention journal)")
+	}
+	byExt, err := store.ListUsers(ctx, account.UserFilter{ExternalIDMask: goneTg}, 5000, 0)
+	if err != nil {
+		t.Fatalf("search by external id: %v", err)
+	}
+	if !listHasID(byExt, gone.ID) {
+		t.Error("a deleted account must be found by the external id it held (retention journal)")
+	}
+}
+
+// TestConfirmCodeClearsGuest: confirming an email on a guest via ConfirmCode promotes it to
+// a durable account (defence-in-depth — no confirmed-email path leaves is_guest set).
+func TestConfirmCodeClearsGuest(t *testing.T) {
+	ctx := context.Background()
+	store := account.NewStore(testDB)
+	mailer := &capturingMailer{}
+	svc := account.NewEmailService(store, mailer, "https://erudit-game.ru")
+
+	guest, err := store.ProvisionGuest(ctx, "")
+	if err != nil {
+		t.Fatalf("provision guest: %v", err)
+	}
+	email := "cc-" + uuid.NewString() + "@example.com"
+	if err := svc.RequestCode(ctx, guest.ID, email); err != nil {
+		t.Fatalf("request code: %v", err)
+	}
+	if _, err := svc.ConfirmCode(ctx, guest.ID, email, sixDigit.FindString(mailer.lastBody)); err != nil {
+		t.Fatalf("confirm code: %v", err)
+	}
+	after, err := store.GetByID(ctx, guest.ID)
+	if err != nil {
+		t.Fatalf("load: %v", err)
+	}
+	if after.IsGuest {
+		t.Error("confirming an email must clear the guest flag")
+	}
+}
+
 // TestDropAllRobotGames drops the deletee's solo vs-AI game but keeps a game with a human
 // opponent.
 func TestDropAllRobotGames(t *testing.T) {

backend/internal/inttest/merge_test.go

@@ -251,6 +251,44 @@ func TestAccountMergeFinishedSharedGameKept(t *testing.T) {
 	}
 }
 
+// TestAccountMergeDedupesEmail keeps the primary's email when both accounts have one and
+// journals the secondary's to the dossier (reason=merge), so the survivor never ends up
+// with two email identities.
+func TestAccountMergeDedupesEmail(t *testing.T) {
+	ctx := context.Background()
+	store := account.NewStore(testDB)
+	merger := accountmerge.NewMerger(testDB)
+
+	primary := provisionAccount(t)
+	secondary := provisionAccount(t)
+	primaryEmail := "keep-" + uuid.NewString() + "@example.com"
+	secondaryEmail := "absorb-" + uuid.NewString() + "@example.com"
+	bindEmailIdentity(t, primary, primaryEmail)
+	bindEmailIdentity(t, secondary, secondaryEmail)
+
+	if err := merger.Merge(ctx, primary, secondary); err != nil {
+		t.Fatalf("merge: %v", err)
+	}
+
+	// The primary keeps its own email; the secondary's is gone from the live identities.
+	if owner, ok, _ := store.AccountIDByIdentity(ctx, account.KindEmail, primaryEmail); !ok || owner != primary {
+		t.Errorf("primary email owner = %s ok=%v, want primary %s", owner, ok, primary)
+	}
+	if _, ok, _ := store.AccountIDByIdentity(ctx, account.KindEmail, secondaryEmail); ok {
+		t.Error("the secondary's email must be removed (no duplicate email on the survivor)")
+	}
+	// The absorbed email stays in the legal dossier, tagged reason=merge.
+	var reason string
+	if err := testDB.QueryRowContext(ctx,
+		`SELECT reason FROM backend.retained_identities WHERE account_id=$1 AND kind='email' AND external_id=$2`,
+		secondary, secondaryEmail).Scan(&reason); err != nil {
+		t.Fatalf("retained email row: %v", err)
+	}
+	if reason != "merge" {
+		t.Errorf("retained reason = %q, want merge", reason)
+	}
+}
+
 // TestAccountLinkFreeEmail binds a free email and promotes a guest to durable.
 func TestAccountLinkFreeEmail(t *testing.T) {
 	ctx := context.Background()
@@ -355,3 +393,64 @@ func TestAccountLinkGuestInversion(t *testing.T) {
 		t.Errorf("email owner = %s, want durable", owner)
 	}
 }
+
+// TestAccountLinkFreeVK binds a free VK identity (gateway-validated, no code) and
+// promotes a guest to durable — the ConfirmVK counterpart of the free-email case.
+func TestAccountLinkFreeVK(t *testing.T) {
+	ctx := context.Background()
+	store := account.NewStore(testDB)
+	links := newLinkService(&capturingMailer{})
+
+	guest := provisionGuest(t)
+	vkID := "vk-" + uuid.NewString()
+	res, err := links.ConfirmVK(ctx, guest, vkID)
+	if err != nil {
+		t.Fatalf("confirm vk: %v", err)
+	}
+	if !res.Linked || res.MergeRequired {
+		t.Fatalf("confirm = %+v, want linked", res)
+	}
+	if acc, _ := store.GetByID(ctx, guest); acc.IsGuest {
+		t.Error("guest flag should clear once VK is linked")
+	}
+	if owner, ok, _ := store.AccountIDByIdentity(ctx, account.KindVK, vkID); !ok || owner != guest {
+		t.Errorf("vk owner = %s, want the promoted guest %s", owner, guest)
+	}
+}
+
+// TestAccountLinkVKMergeIntoCaller merges the account owning a VK identity into the
+// current durable account: ConfirmVK previews the merge, MergeVK folds it, the caller
+// stays primary and keeps its session, and the VK identity repoints to the caller.
+func TestAccountLinkVKMergeIntoCaller(t *testing.T) {
+	ctx := context.Background()
+	store := account.NewStore(testDB)
+	links := newLinkService(&capturingMailer{})
+
+	caller := provisionAccount(t)
+	other := provisionAccount(t)
+	vkID := "vk-" + uuid.NewString()
+	if err := store.AttachIdentity(ctx, other, account.KindVK, vkID, true); err != nil {
+		t.Fatalf("seed vk on other: %v", err)
+	}
+
+	confirm, err := links.ConfirmVK(ctx, caller, vkID)
+	if err != nil {
+		t.Fatalf("confirm vk: %v", err)
+	}
+	if !confirm.MergeRequired || confirm.SecondaryID != other {
+		t.Fatalf("confirm = %+v, want merge_required to other %s", confirm, other)
+	}
+	merge, err := links.MergeVK(ctx, caller, vkID)
+	if err != nil {
+		t.Fatalf("merge vk: %v", err)
+	}
+	if merge.PrimaryID != caller || merge.SwitchedToken != "" {
+		t.Fatalf("merge = %+v, want primary caller and no session switch", merge)
+	}
+	if mergedInto(t, other) != caller {
+		t.Error("other should be tombstoned into caller")
+	}
+	if owner, _, _ := store.AccountIDByIdentity(ctx, account.KindVK, vkID); owner != caller {
+		t.Errorf("vk owner = %s, want caller after merge", owner)
+	}
+}

backend/internal/link/service.go

@@ -133,6 +133,53 @@ func (s *Service) attachTelegram(ctx context.Context, callerID uuid.UUID, extern
 	return s.accounts.ClearGuest(ctx, callerID)
 }
 
+// ConfirmVK attaches a gateway-validated VK identity to the caller (Linked) or
+// reports that it belongs to another account (MergeRequired). The gateway has already
+// completed the VK ID code exchange, so externalID is the trusted vk user id.
+func (s *Service) ConfirmVK(ctx context.Context, callerID uuid.UUID, externalID string) (ConfirmResult, error) {
+	owner, ok, err := s.accounts.AccountIDByIdentity(ctx, account.KindVK, externalID)
+	if err != nil {
+		return ConfirmResult{}, err
+	}
+	if !ok {
+		if err := s.attachVK(ctx, callerID, externalID); err != nil {
+			return ConfirmResult{}, err
+		}
+		return ConfirmResult{Linked: true}, nil
+	}
+	if owner == callerID {
+		return ConfirmResult{Linked: true}, nil
+	}
+	return ConfirmResult{MergeRequired: true, SecondaryID: owner}, nil
+}
+
+// MergeVK merges the account owning a gateway-validated VK identity into the caller's
+// (subject to the guest-primary rule).
+func (s *Service) MergeVK(ctx context.Context, callerID uuid.UUID, externalID string) (MergeResult, error) {
+	owner, ok, err := s.accounts.AccountIDByIdentity(ctx, account.KindVK, externalID)
+	if err != nil {
+		return MergeResult{}, err
+	}
+	if !ok {
+		if err := s.attachVK(ctx, callerID, externalID); err != nil {
+			return MergeResult{}, err
+		}
+		return MergeResult{PrimaryID: callerID}, nil
+	}
+	if owner == callerID {
+		return MergeResult{PrimaryID: callerID}, nil
+	}
+	return s.merge(ctx, callerID, owner)
+}
+
+// attachVK links the identity to the caller and promotes a guest.
+func (s *Service) attachVK(ctx context.Context, callerID uuid.UUID, externalID string) error {
+	if err := s.accounts.AttachIdentity(ctx, callerID, account.KindVK, externalID, true); err != nil {
+		return err
+	}
+	return s.accounts.ClearGuest(ctx, callerID)
+}
+
 // merge decides the primary (the caller, unless it is a guest and the other is
 // durable), runs the data merge, retires the secondary's sessions and mints a new
 // session when the active account switches.

backend/internal/postgres/migrations/00008_retained_identities_merge_reason.sql

@@ -0,0 +1,22 @@
+-- Admit the 'merge' reason into retained_identities. An account merge folds a secondary
+-- account into a primary; when both hold an identity of the same kind (e.g. each has a
+-- confirmed email), the survivor keeps its own and the secondary's is journaled to the
+-- legal dossier and removed — so the survivor never ends up with two identities of one
+-- kind, and the absorbed credential is still retained. That journal row carries reason
+-- 'merge', alongside the existing unlink / change / delete.
+--
+-- Expand-contract: the Up only WIDENS the allowed reason set, so an older backend image
+-- (which writes only unlink/change/delete) still satisfies the constraint — a rollback
+-- stays DB-safe. The Down narrows it again and would reject pre-existing 'merge' rows, so
+-- it is a dev-only convenience, not a production rollback path (image rollback runs old
+-- code against this schema, not the Down migration).
+
+-- +goose Up
+ALTER TABLE backend.retained_identities DROP CONSTRAINT retained_identities_reason_chk;
+ALTER TABLE backend.retained_identities ADD CONSTRAINT retained_identities_reason_chk
+    CHECK ((reason = ANY (ARRAY['unlink'::text, 'change'::text, 'delete'::text, 'merge'::text])));
+
+-- +goose Down
+ALTER TABLE backend.retained_identities DROP CONSTRAINT retained_identities_reason_chk;
+ALTER TABLE backend.retained_identities ADD CONSTRAINT retained_identities_reason_chk
+    CHECK ((reason = ANY (ARRAY['unlink'::text, 'change'::text, 'delete'::text])));

backend/internal/server/handlers.go

@@ -74,6 +74,8 @@ func (s *Server) registerRoutes() {
 		u.POST("/link/email/merge", s.handleLinkEmailMerge)
 		u.POST("/link/telegram", s.handleLinkTelegram)
 		u.POST("/link/telegram/merge", s.handleLinkTelegramMerge)
+		u.POST("/link/vk", s.handleLinkVK)
+		u.POST("/link/vk/merge", s.handleLinkVKMerge)
 		u.POST("/link/unlink", s.handleUnlink)
 		// Change the account's confirmed email: mail a code to the new address, then
 		// atomically switch on confirm (a new address owned by another account is

backend/internal/server/handlers_admin_console.go

@@ -134,6 +134,7 @@ func (s *Server) consoleUsers(c *gin.Context) {
 	page := consolePage(c)
 	filter := account.UserFilter{
 		Robots:         c.Query("kind") == "robots",
+		Deleted:        c.Query("kind") == "deleted",
 		NameMask:       c.Query("name"),
 		ExternalIDMask: c.Query("ext"),
 		EmailExact:     c.Query("email"),
@@ -148,6 +149,9 @@ func (s *Server) consoleUsers(c *gin.Context) {
 	if filter.Robots {
 		q.Set("kind", "robots")
 	}
+	if filter.Deleted {
+		q.Set("kind", "deleted")
+	}
 	if strings.TrimSpace(filter.NameMask) != "" {
 		q.Set("name", filter.NameMask)
 	}
@@ -159,21 +163,24 @@ func (s *Server) consoleUsers(c *gin.Context) {
 	}
 	view := adminconsole.UsersView{
 		Pager:  adminconsole.NewPager(page, adminPageSize, total),
-		Robots: filter.Robots, NameMask: filter.NameMask, ExternalIDMask: filter.ExternalIDMask,
+		Robots: filter.Robots, Deleted: filter.Deleted, NameMask: filter.NameMask, ExternalIDMask: filter.ExternalIDMask,
 		EmailExact:  filter.EmailExact,
 		FilterQuery: template.URL(q.Encode()),
 	}
 	ids := make([]uuid.UUID, 0, len(items))
 	for _, it := range items {
 		kind := "registered"
-		if it.IsRobot {
+		switch {
+		case it.IsRobot:
 			kind = "robot"
-		} else if it.IsGuest {
+		case it.IsDeleted:
+			kind = "deleted"
+		case it.IsGuest:
 			kind = "guest"
 		}
 		view.Items = append(view.Items, adminconsole.UserRow{
 			ID: it.ID.String(), DisplayName: it.DisplayName, Kind: kind,
-			Language: it.PreferredLanguage, Guest: it.IsGuest,
+			Language: it.PreferredLanguage, Guest: it.IsGuest, Deleted: it.IsDeleted,
 			FlaggedHighRate: !it.FlaggedHighRateAt.IsZero(), CreatedAt: fmtTime(it.CreatedAt),
 		})
 		ids = append(ids, it.ID)

backend/internal/server/handlers_link.go

@@ -34,6 +34,12 @@ type linkTelegramBody struct {
 	ExternalID string `json:"external_id"`
 }
 
+// linkVKBody carries a gateway-validated VK identity (the trusted vk user id the
+// gateway resolved from the VK ID code exchange).
+type linkVKBody struct {
+	ExternalID string `json:"external_id"`
+}
+
 // linkResultResponse is the unified result of a confirm or merge step. Status is
 // "linked" (bound to the caller), "merge_required" (the identity belongs to another
 // account — the secondary_* fields summarise it for the irreversible confirmation),
@@ -233,6 +239,48 @@ func (s *Server) handleLinkTelegramMerge(c *gin.Context) {
 	c.JSON(http.StatusOK, s.mergeResultResponse(c, res))
 }
 
+// handleLinkVK attaches a gateway-validated VK identity to the caller or reports a
+// required merge.
+func (s *Server) handleLinkVK(c *gin.Context) {
+	uid, ok := userID(c)
+	if !ok {
+		abortBadRequest(c, "missing identity")
+		return
+	}
+	var req linkVKBody
+	if err := c.ShouldBindJSON(&req); err != nil || req.ExternalID == "" {
+		abortBadRequest(c, "missing external_id")
+		return
+	}
+	res, err := s.links.ConfirmVK(c.Request.Context(), uid, req.ExternalID)
+	if err != nil {
+		s.abortErr(c, err)
+		return
+	}
+	c.JSON(http.StatusOK, s.confirmResultResponse(c, uid, res))
+}
+
+// handleLinkVKMerge merges the account owning a gateway-validated VK identity into
+// the caller's.
+func (s *Server) handleLinkVKMerge(c *gin.Context) {
+	uid, ok := userID(c)
+	if !ok {
+		abortBadRequest(c, "missing identity")
+		return
+	}
+	var req linkVKBody
+	if err := c.ShouldBindJSON(&req); err != nil || req.ExternalID == "" {
+		abortBadRequest(c, "missing external_id")
+		return
+	}
+	res, err := s.links.MergeVK(c.Request.Context(), uid, req.ExternalID)
+	if err != nil {
+		s.abortErr(c, err)
+		return
+	}
+	c.JSON(http.StatusOK, s.mergeResultResponse(c, res))
+}
+
 // confirmResultResponse renders a confirm step: a merge preview (secondary summary)
 // or a completed link (the active account's refreshed profile).
 func (s *Server) confirmResultResponse(c *gin.Context, activeID uuid.UUID, res link.ConfirmResult) linkResultResponse {

deploy/.env.example

@@ -1,6 +1,10 @@
 # Environment for deploy/docker-compose.yml. The CI deploy job (ci.yaml) maps the
 # Gitea TEST_-prefixed secrets/variables onto these unprefixed names; the prod
-# deploy maps the PROD_-prefixed set the same way. Copy to deploy/.env for a local run.
+# deploy maps the PROD_-prefixed set the same way. Values that are identical on every
+# contour (DICT_VERSION, SMTP_RELAY_HOST/PORT/TLS/USER/PASS, GRAFANA_SMTP_PORT,
+# VITE_VK_APP_LINK/ID, the two VK secrets) live as ONE unprefixed Gitea entry, and the
+# deploy derives TELEGRAM_MINIAPP_URL / GRAFANA_ROOT_URL / VITE_VK_ID_REDIRECT_URL from
+# PUBLIC_BASE_URL. Copy to deploy/.env for a local run (set the derived ones directly).
 #
 # Full reference (required vs optional, defaults, secret-vs-variable): deploy/README.md.
 
@@ -15,8 +19,9 @@ POSTGRES_PASSWORD=change-me            # required
 # boot the dawg-data volume preserves versions uploaded through the admin console and
 # the active version lives in the DB. On a live volume a changed value is ignored (the
 # recorded .seed_version marker wins — the seed-drift guard); change a running
-# contour's dictionary through /_gm/dictionary (ARCHITECTURE.md §5).
+# contour's dictionary through /_gm/dictionary (ARCHITECTURE.md §5). One shared Gitea
-DICT_VERSION=v1.3.0
+# variable (DICT_VERSION) seeds both contours + pins the CI test suite.
+DICT_VERSION=v1.3.1
 
 # --- Logging ----------------------------------------------------------------
 LOG_LEVEL=info
@@ -45,6 +50,17 @@ SMTP_RELAY_PASS=                       # secret
 SMTP_RELAY_FROM=no-reply@erudit-game.ru
 PUBLIC_BASE_URL=                       # required when SMTP_RELAY_HOST is set (e.g. https://erudit-game.ru)
 
+# Operator alerts (email). The backend emails the admin on new feedback / word complaints
+# (coalesced), and Grafana emails infra alerts. Distinct senders; recipients may be several
+# comma-separated addresses. Grafana reuses SMTP_RELAY_HOST/USER/PASS but dials the STARTTLS
+# port (it can't do the backend's implicit TLS), GRAFANA_SMTP_PORT. All empty = off.
+SMTP_RELAY_ADMIN_FROM=                 # backend admin-alert From (e.g. alerts@erudit-game.ru)
+ADMIN_EMAIL=                           # backend admin-alert recipient(s), comma-separated
+SMTP_RELAY_SERVICE_FROM=               # Grafana alert From; the deploy derives a bare address for Grafana (it rejects "Name" <addr>)
+SERVICE_EMAIL=                         # Grafana alert recipient(s), comma-separated
+GRAFANA_SMTP_PORT=                     # Grafana STARTTLS port on SMTP_RELAY_HOST (Selectel: 1126)
+GF_SMTP_ENABLED=false                  # set true to enable Grafana alert emails
+
 # --- Edge / caddy -----------------------------------------------------------
 # Test: ":80" (the host caddy terminates TLS and forwards to scrabble:80 on the
 # external `edge` network). Prod: a domain so caddy does its own ACME.
@@ -57,10 +73,12 @@ VITE_TELEGRAM_BOT_ID=
 VITE_TELEGRAM_LINK=                                 # friend-invite Mini App link (full URL, e.g. https://t.me/<bot>/<app>)
 VITE_TELEGRAM_GAME_CHANNEL_NAME=                    # landing "Play in Telegram" link, the bot's game channel
 VITE_VK_APP_LINK=                                   # landing "Play on VK" link (full URL, https://vk.com/app<id>)
+VITE_VK_APP_ID=                                     # VK ID "Web" app id (client_id) for VK web-login linking; also the gateway's GATEWAY_VK_ID_APP_ID
+VITE_VK_ID_REDIRECT_URL=                            # VK ID trusted redirect URL (e.g. https://erudit-game.ru/app/); also the gateway's exchange redirect_uri. Deploy derives it as PUBLIC_BASE_URL + /app/
 VITE_GATEWAY_URL=
 
 # --- Grafana ----------------------------------------------------------------
-GRAFANA_ROOT_URL=/_gm/grafana/         # set the full https URL behind a real domain
+GRAFANA_ROOT_URL=/_gm/grafana/         # deploy derives PUBLIC_BASE_URL + /_gm/grafana/; set the full https URL for a local run
 GRAFANA_ADMIN_PASSWORD=admin
 
 # --- Telegram validator + bot -----------------------------------------------
@@ -77,7 +95,7 @@ TELEGRAM_PROMO_BOT_TOKEN=             # optional standalone promo bot token; emp
 TELEGRAM_BOT_USERNAME=               # main bot @username without the @ (promo message); required when the promo token is set
 TELEGRAM_BOT_LINK=                    # main bot Mini App link for the promo button (reuse VITE_TELEGRAM_LINK); required when the promo token is set
 TELEGRAM_PROMO_START_PARAM=           # promo button startapp payload — a variant-seed deep link (default verudit_ru-scrabble_en) adding English Scrabble for new users; empty forwards the user's /start payload
-TELEGRAM_MINIAPP_URL=                  # required
+TELEGRAM_MINIAPP_URL=                  # required; deploy derives it as PUBLIC_BASE_URL + /telegram/
 TELEGRAM_TEST_ENV=false
 TELEGRAM_API_BASE_URL=
 
@@ -86,3 +104,14 @@ TELEGRAM_API_BASE_URL=
 # launch-parameter signature in-process under it (a pure offline HMAC, no VK API call).
 # Empty disables the VK auth path (auth.vk). Set from the Gitea TEST_/PROD_ secret.
 GATEWAY_VK_APP_SECRET=
+# VK ID web login (browser VK-identity linking) — a SEPARATE VK "Web" app: the gateway runs
+# the confidential OAuth 2.1 code exchange under this protected key. The app id + redirect URL
+# come from VITE_VK_APP_ID / VITE_VK_ID_REDIRECT_URL above. All three empty disables link.vk.*.
+GATEWAY_VK_ID_CLIENT_SECRET=
+
+# --- Gateway anti-abuse ------------------------------------------------------
+# Planted honeytoken bearer value: any request presenting it is flagged — a 24h IP ban
+# where the IP ban is on (prod), logs + a ban metric otherwise (test). Plant the value
+# somewhere an attacker would find it; empty disables the trap. Gitea
+# TEST_/PROD_GATEWAY_HONEYTOKEN (secret).
+GATEWAY_HONEYTOKEN=

deploy/README.md

@@ -46,6 +46,18 @@ runs `docker compose up -d --build` on the runner host. The prod deploy maps the
 **`PROD_`** set the same way. So a Gitea secret named `TEST_POSTGRES_PASSWORD`
 feeds the compose's `POSTGRES_PASSWORD`, etc.
 
+Three naming classes in Gitea:
+- **Per-contour** (`TEST_`/`PROD_<NAME>`) — values that differ between the contours
+  (bots, hosts, public origin, log level, email senders): the common case below.
+- **Shared** (one unprefixed `<NAME>`, no prefix) — values identical on every contour,
+  stored once: `DICT_VERSION`, `SMTP_RELAY_HOST`/`PORT`/`TLS`/`USER`/`PASS`,
+  `GRAFANA_SMTP_PORT`, `VITE_VK_APP_LINK`, `VITE_VK_APP_ID`, `GATEWAY_VK_APP_SECRET`,
+  `GATEWAY_VK_ID_CLIENT_SECRET` (one Selectel relay + one pair of VK apps for all contours).
+- **Derived** — not stored at all; the deploy computes them from `PUBLIC_BASE_URL`
+  (`deploy/write-prod-env.sh`, and the `ci.yaml` deploy step): `TELEGRAM_MINIAPP_URL`
+  (`+ /telegram/`), `GRAFANA_ROOT_URL` (`+ /_gm/grafana/`), `VITE_VK_ID_REDIRECT_URL`
+  (`+ /app/`). Set them directly only for a local `.env` run.
+
 The deploy job also **seeds the config files** (`caddy`, `otelcol`, `prometheus`,
 `tempo`, `grafana`) to a stable host path (`$HOME/.scrabble-deploy`) and sets
 `SCRABBLE_CONFIG_DIR` to it before `up`. The runner's checkout is an ephemeral act
@@ -62,7 +74,7 @@ compose binds from this directory.
 | --- | --- | --- |
 | `POSTGRES_PASSWORD` | secret | Postgres password (also embedded in `BACKEND_POSTGRES_DSN`). |
 | `GM_BASICAUTH_HASH` | secret | bcrypt hash gating `/_gm` (admin console + Grafana). Generate with `docker run --rm caddy:2-alpine caddy hash-password --plaintext '<pw>'`. |
-| `TELEGRAM_MINIAPP_URL` | variable | The Mini App URL the bot hands out in deep links / buttons. |
+| `TELEGRAM_MINIAPP_URL` | derived | The Mini App URL the bot hands out in deep links / buttons. The deploy derives `PUBLIC_BASE_URL + /telegram/`; set it directly only for a local run (compose still `:?`-requires it). |
 | `EXPORT_SIGN_KEY` | secret | HMAC key signing the public finished-game export download URLs (`/dl/*`). Generate with `openssl rand -base64 32`. |
 
 **Plus the bot token** — `TELEGRAM_BOT_TOKEN` (secret), shared by the validator (HMAC
@@ -82,11 +94,11 @@ without it Docker's resolver handles `otelcol`, `gateway` and `api.telegram.org`
 | --- | --- | --- | --- |
 | `POSTGRES_DB` | variable | `scrabble` | Database name. |
 | `POSTGRES_USER` | variable | `scrabble` | Database user. |
-| `DICT_VERSION` | variable | `v1.3.0` | `scrabble-dictionary` release tag baked into the backend image as the **seed for a fresh volume** (build-arg). A live contour changes dictionary through the admin console, not this; on a seeded volume a changed value is ignored (the recorded `.seed_version` marker wins — the seed-drift guard, ARCHITECTURE.md §5). Set per contour as `TEST_`/`PROD_DICT_VERSION`. |
+| `DICT_VERSION` | variable (shared) | `v1.3.1` | `scrabble-dictionary` release tag baked into the backend image as the **seed for a fresh volume** (build-arg). A live contour changes dictionary through the admin console, not this; on a seeded volume a changed value is ignored (the recorded `.seed_version` marker wins — the seed-drift guard, ARCHITECTURE.md §5). One shared unprefixed Gitea variable seeds both contours and pins the CI test suite. |
 | `LOG_LEVEL` | variable | `info` | Shared log level for backend / gateway / validator / bot (`debug\|info\|warn\|error`). |
 | `CADDY_SITE_ADDRESS` | variable | `:80` | Caddy site address. Test: `:80` (host caddy terminates TLS). Prod: a domain, so caddy does its own ACME. |
 | `GM_BASICAUTH_USER` | variable | `gm` | Username for the `/_gm` Basic-Auth. |
-| `GRAFANA_ROOT_URL` | variable | `/_gm/grafana/` | Grafana root URL (sub-path serving). Set the full `https://<domain>/_gm/grafana/` behind a real domain. |
+| `GRAFANA_ROOT_URL` | derived | `/_gm/grafana/` | Grafana root URL (sub-path serving). The deploy derives `PUBLIC_BASE_URL + /_gm/grafana/`; set the full `https://<domain>/_gm/grafana/` only for a local run. |
 | `GRAFANA_ADMIN_PASSWORD` | secret | `admin` | Grafana admin password. Low impact (the login form is disabled, access is anonymous-admin behind caddy) but set it anyway. |
 | `TELEGRAM_GAME_CHANNEL_ID` | variable | _(empty)_ | The bot's game-channel id; empty/`0` disables channel posts. |
 | `TELEGRAM_TEST_ENV` | _pinned_ | `false` | `true` routes the bot through Telegram's test environment (`.../bot<token>/test/METHOD`). **The CI test contour pins this to `true` in `ci.yaml`** (the contour is the test environment) — it is not a Gitea variable. Set it in `.env` for a local run; prod leaves it `false`. |
@@ -94,15 +106,26 @@ without it Docker's resolver handles `otelcol`, `gateway` and `api.telegram.org`
 | `VITE_TELEGRAM_BOT_ID` | variable | _(empty)_ | UI build-arg: numeric bot id for the web Login Widget. |
 | `VITE_TELEGRAM_LINK` | variable | _(empty)_ | UI build-arg: friend-invite Mini App link (full URL, `https://t.me/<bot>/<app>` — `<app>` is the Mini App short name from BotFather). |
 | `VITE_TELEGRAM_GAME_CHANNEL_NAME` | variable | _(empty)_ | UI build-arg: the landing "Play in Telegram" link, the bot's game channel (e.g. `https://t.me/Erudit_Game`). |
-| `VITE_VK_APP_LINK` | variable | _(empty)_ | UI build-arg: the landing "Play on VK" link, the VK Mini App (full URL, `https://vk.com/app<id>`). |
+| `VITE_VK_APP_LINK` | variable (shared) | _(empty)_ | UI build-arg: the landing "Play on VK" link, the VK Mini App (full URL, `https://vk.com/app<id>`). One VK Mini App for all contours. |
+| `VITE_VK_APP_ID` | variable (shared) | _(empty)_ | VK ID "Web" app id (`client_id`) for VK web-login linking. Baked into the SPA authorize URL and reused as the gateway's `GATEWAY_VK_ID_APP_ID`. Empty disables the `link.vk.*` ops. One VK ID "Web" app for all contours. |
+| `VITE_VK_ID_REDIRECT_URL` | derived | _(empty)_ | VK ID trusted redirect URL — must match the app's registered redirect exactly. The deploy derives `PUBLIC_BASE_URL + /app/` (used by the SPA and as the gateway's exchange `redirect_uri`); set it only for a local run. |
+| `GATEWAY_VK_APP_SECRET` | secret (shared) | _(empty)_ | The VK Mini App's protected key (`client_secret`): the gateway verifies the launch-parameter signature in-process under it (offline HMAC, no VK API call). Empty disables the VK auth path (`auth.vk`). One VK Mini App for all contours. |
+| `GATEWAY_VK_ID_CLIENT_SECRET` | secret (shared) | _(empty)_ | The VK ID "Web" app's protected key (`client_secret`) for the gateway's server-side confidential code exchange. A SEPARATE VK app from `GATEWAY_VK_APP_SECRET` (the Mini App). One VK ID "Web" app for all contours. |
+| `GATEWAY_HONEYTOKEN` | secret | _(empty)_ | Planted honeytoken bearer value: presenting it earns a 24h IP ban + a high-severity alarm where the IP ban is on (prod), or logs + a `gateway_abuse_banned_total{reason="honeytoken"}` metric (test, ban off). Plant the value somewhere an attacker would find it; empty disables the trap. Per-contour `TEST_`/`PROD_GATEWAY_HONEYTOKEN`. |
 | `VITE_GATEWAY_URL` | variable | _(empty)_ | UI build-arg: gateway origin; empty = same-origin (the usual single-origin deploy). |
-| `SMTP_RELAY_HOST` | variable | _(empty)_ | Selectel SMTP relay host for confirm-code email. Empty leaves the backend on the log mailer (email disabled) — the contour still boots. One relay for every contour (limit 100 msgs / 5 min). |
+| `SMTP_RELAY_HOST` | variable (shared) | _(empty)_ | Selectel SMTP relay host for confirm-code email. Empty leaves the backend on the log mailer (email disabled) — the contour still boots. One relay for every contour (limit 100 msgs / 5 min). |
-| `SMTP_RELAY_PORT` | variable | `465` | Relay port. No client certificate is needed (the server cert is validated against the system roots). |
+| `SMTP_RELAY_PORT` | variable (shared) | `465` | Relay port. No client certificate is needed (the server cert is validated against the system roots). |
-| `SMTP_RELAY_TLS` | variable | _(empty)_ | Transport security: `ssl` (implicit TLS) or `starttls`. Empty derives it from the port (implicit on `465`, STARTTLS otherwise); required for a relay on a non-standard port (e.g. Selectel's `1127` = SSL, `1126` = STARTTLS). |
+| `SMTP_RELAY_TLS` | variable (shared) | _(empty)_ | Transport security: `ssl` (implicit TLS) or `starttls`. Empty derives it from the port (implicit on `465`, STARTTLS otherwise); required for a relay on a non-standard port (e.g. Selectel's `1127` = SSL, `1126` = STARTTLS). |
-| `SMTP_RELAY_USER` | secret | _(empty)_ | Relay SMTP AUTH username. |
+| `SMTP_RELAY_USER` | secret (shared) | _(empty)_ | Relay SMTP AUTH username. |
-| `SMTP_RELAY_PASS` | secret | _(empty)_ | Relay SMTP AUTH password. |
+| `SMTP_RELAY_PASS` | secret (shared) | _(empty)_ | Relay SMTP AUTH password. |
 | `SMTP_RELAY_FROM` | variable | `no-reply@localhost` | Sender address. **Must use the prod domain** (`no-reply@erudit-game.ru`) — Selectel only accepts the verified sender domain — so it is the same on every contour. |
 | `PUBLIC_BASE_URL` | variable | _(empty)_ | Canonical public https origin for links in the email (the contour's own URL, e.g. `https://erudit-game.ru`). **Required by the backend whenever `SMTP_RELAY_HOST` is set** — it is never derived from a request Host header (anti-injection). |
+| `SMTP_RELAY_ADMIN_FROM` | variable | _(empty)_ | Backend operator-alert sender (new feedback / word complaints), distinct from the confirm-code From. Empty (with `ADMIN_EMAIL`) disables the alert worker. |
+| `ADMIN_EMAIL` | variable | _(empty)_ | Backend operator-alert recipient(s); several comma-separated addresses allowed. |
+| `SMTP_RELAY_SERVICE_FROM` | variable | _(empty)_ | Grafana infra-alert sender address. |
+| `SERVICE_EMAIL` | variable | _(empty)_ | Grafana infra-alert recipient(s); comma-separated allowed (read by the provisioned contact point via `$__env{SERVICE_EMAIL}`). |
+| `GRAFANA_SMTP_PORT` | variable (shared) | _(empty)_ | STARTTLS port Grafana dials on `SMTP_RELAY_HOST` (its client can't do the backend's implicit TLS), e.g. Selectel's `1126`. Reuses `SMTP_RELAY_HOST`/`USER`/`PASS`. |
+| `GF_SMTP_ENABLED` | variable | `false` | Set `true` to enable Grafana alert emails. |
 
 The six `VITE_*` are **build-args** baked into the gateway and landing images at
 build time (both targets share one UI build stage — keep the args identical so it is
@@ -132,14 +155,13 @@ intentionally **unset** — caddy owns `/_gm` in the contour.
 
 The dictionary ships as a versioned **release artifact** (`scrabble-dawg-vX.Y.Z.tar.gz`) from
 [`scrabble-dictionary`](https://gitea.iliadenisov.ru/developer/scrabble-dictionary). The tag is
-a build-time input with **no default** in the images, so it is set in exactly two places to
+a build-time input with **no default** in the images. It is a single Gitea repo variable —
-move the whole stack — change both to a new release:
+**`DICT_VERSION`** (unprefixed, shared across contours) — so a release bump is **one edit**:
 
-1. **CI tests** — `.gitea/workflows/ci.yaml` `env.DICT_VERSION` (the unit/integration jobs
+- the CI test jobs read it via `.gitea/workflows/ci.yaml`'s top-level
-   download that dawg).
+  `env.DICT_VERSION: ${{ vars.DICT_VERSION }}` (the unit/integration jobs download that dawg), and
-2. **Deploy seed** — the Gitea repo variables `TEST_DICT_VERSION` / `PROD_DICT_VERSION` (the tag
+- both deploy jobs feed the same variable to `compose` as the `DICT_VERSION` build-arg that
-   the deploy bakes into a **fresh** volume's image; the deploy job feeds it to `compose` as
+  bakes a **fresh** volume's seed.
-   `DICT_VERSION`).
 
 For local builds set `DICT_VERSION` in `deploy/.env` (template: `.env.example`); a bare
 `docker build` needs `--build-arg DICT_VERSION=vX.Y.Z`. The Dockerfiles and `compose` carry no
@@ -168,6 +190,16 @@ public site. After `master` is green this workflow is the **only** thing that to
 prod — nothing auto-deploys there. It runs four visible jobs: **build → deploy-main →
 deploy-bot → verify** (the per-service rolling shows in the deploy-main log).
 
+**Maintenance page.** During the roll (and any migration window) `prod-deploy.sh` raises a
+flag the edge caddy serves a static 503 "технические работы" page from
+(`deploy/caddy/maintenance.html`), for the user-facing routes only — `/_gm` (Grafana) stays
+reachable. The flag is cleared on any exit (success, a health failure + rollback, or an
+error) by a shell trap, so it can never stick on. It arms from this feature's own deploy
+onward (the caddy carrying the gate must be live first). This is **not** a zero-downtime
+deploy — the backend is a single stateful instance (in-memory game state + push hub, no
+Redis), so its swap still blips (clients auto-reconnect the live stream); the page just
+makes the window graceful instead of raw 502s.
+
 **Versioning.** Each release is a git tag `vX.Y.Z` on `master`; the deploy stamps
 `git describe --tags` into every image tag, every binary (`-ldflags` → `pkg/version` →
 the `service.version` telemetry attribute) and the SPA About screen. Tag the release
@@ -200,18 +232,31 @@ together with the fresh CA.
 **Sizing / monitoring:** the main host launches undersized (2 vCPU / 1.9 GiB); the prod
 overlay trims limits + `GOMAXPROCS=2` + 7d Prometheus retention, and `node_exporter` feeds
 host memory to Grafana (`/_gm/grafana/`). Watch host memory and resize at Selectel when
-players arrive.
+players arrive. The per-container memory caps are enforced (Compose v2 → cgroup), so no one
+service can eat all RAM, but they **overcommit** the host (~2.8 GiB of caps vs 1.9 GiB) — a
+**1 GiB swap file** (Ansible `common` role, `swap_size`, `vm.swappiness=10`) cushions a
+simultaneous spike into swap instead of the kernel OOM-killer. The `host_mem_low` Grafana
+alert fires under 10% available.
 
-**`PROD_` Gitea set** (mirrors `TEST_`, mapped onto the unprefixed names above) — secrets:
+**Shared Gitea set** (one unprefixed entry each, used by every contour) — secrets:
+`GATEWAY_VK_APP_SECRET, GATEWAY_VK_ID_CLIENT_SECRET, SMTP_RELAY_USER, SMTP_RELAY_PASS`;
+variables: `DICT_VERSION, SMTP_RELAY_HOST, SMTP_RELAY_PORT, SMTP_RELAY_TLS, GRAFANA_SMTP_PORT,
+VITE_VK_APP_LINK, VITE_VK_APP_ID`. **Derived from `PUBLIC_BASE_URL` at deploy** (not stored):
+`TELEGRAM_MINIAPP_URL, GRAFANA_ROOT_URL, VITE_VK_ID_REDIRECT_URL`.
+
+**`PROD_` Gitea set** (per-contour, mirrors `TEST_`, mapped onto the unprefixed names above)
+— secrets:
 `PROD_{POSTGRES_PASSWORD, GM_BASICAUTH_HASH, GRAFANA_ADMIN_PASSWORD, TELEGRAM_BOT_TOKEN,
-TELEGRAM_PROMO_BOT_TOKEN, EXPORT_SIGN_KEY, REGISTRY_PASSWORD, SSH_KEY, SSH_KNOWN_HOSTS, BOTLINK_CA,
+TELEGRAM_PROMO_BOT_TOKEN, EXPORT_SIGN_KEY, GATEWAY_HONEYTOKEN, REGISTRY_PASSWORD, SSH_KEY,
-BOTLINK_GATEWAY_CERT, BOTLINK_GATEWAY_KEY, BOTLINK_BOT_CERT, BOTLINK_BOT_KEY,
+SSH_KNOWN_HOSTS, BOTLINK_CA, BOTLINK_GATEWAY_CERT, BOTLINK_GATEWAY_KEY, BOTLINK_BOT_CERT,
-SMTP_RELAY_USER, SMTP_RELAY_PASS}`; variables:
+BOTLINK_BOT_KEY}`; variables:
-`PROD_{REGISTRY_USER, MAIN_HOST, TG_HOST, CADDY_SITE_ADDRESS, GM_BASICAUTH_USER,
+`PROD_{REGISTRY_USER, MAIN_HOST, TG_HOST, CADDY_SITE_ADDRESS, GM_BASICAUTH_USER, LOG_LEVEL,
-GRAFANA_ROOT_URL, LOG_LEVEL, DICT_VERSION, TELEGRAM_MINIAPP_URL, TELEGRAM_GAME_CHANNEL_ID,
+TELEGRAM_GAME_CHANNEL_ID, TELEGRAM_CHAT_ID, TELEGRAM_SUPPORT_CHAT_ID, TELEGRAM_BOT_USERNAME,
-TELEGRAM_CHAT_ID, TELEGRAM_BOT_USERNAME, VITE_TELEGRAM_BOT_ID, VITE_TELEGRAM_LINK,
+VITE_TELEGRAM_BOT_ID, VITE_TELEGRAM_LINK, VITE_TELEGRAM_GAME_CHANNEL_NAME, SMTP_RELAY_FROM,
-VITE_TELEGRAM_GAME_CHANNEL_NAME, VITE_VK_APP_LINK,
+PUBLIC_BASE_URL, SMTP_RELAY_ADMIN_FROM, ADMIN_EMAIL, SMTP_RELAY_SERVICE_FROM, SERVICE_EMAIL,
-SMTP_RELAY_HOST, SMTP_RELAY_PORT, SMTP_RELAY_TLS, SMTP_RELAY_FROM, PUBLIC_BASE_URL}`.
+GF_SMTP_ENABLED}`. The test contour uses the same names under `TEST_`, minus the prod-only
+infra (`MAIN_HOST`/`TG_HOST`/`REGISTRY_*`/`SSH_*`/`BOTLINK_*`, which the test deploy generates
+or runs locally) and plus `TEST_AWG_CONF` (the bot's VPN egress).
 
 ## Host-side setup (outside this repo)
 

deploy/ansible/group_vars/all.yml

@@ -19,3 +19,11 @@ scrabble_base_dir: /opt/scrabble
 # the host's own containers (and any ad-hoc runs) rotate identically.
 docker_log_max_size: "10m"
 docker_log_max_file: "3"
+
+# Swap file as an OOM cushion. The per-container memory caps (docker-compose.prod.yml)
+# sum to more than the tight main host's RAM (2 vCPU / 1.9 GiB), so a simultaneous
+# spike could otherwise hit the kernel OOM-killer and it might pick postgres. A small
+# swap absorbs the overshoot; swappiness stays low so swap is a cushion, not a hot
+# path (a slow service beats a killed one). Set swap_size to "0" to skip provisioning.
+swap_size: "1G"
+swap_swappiness: 10

deploy/ansible/roles/common/tasks/main.yml

@@ -150,6 +150,57 @@
     enabled: true
     state: started
 
+# --- Swap file (OOM cushion) ---------------------------------------------------
+# The prod main host is tight (1.9 GiB) and the per-container memory caps
+# (docker-compose.prod.yml) sum to more than RAM, so a simultaneous spike could hit
+# the kernel OOM-killer (which might pick postgres). A small swap absorbs the
+# overshoot. Idempotent; skipped when swap_size == "0". Builtin-only (no ansible.posix).
+
+- name: Check whether the swap file is already active
+  ansible.builtin.command: swapon --show=NAME --noheadings
+  register: swap_active
+  changed_when: false
+  when: swap_size != "0"
+
+- name: Allocate the swap file
+  ansible.builtin.command:
+    cmd: "fallocate -l {{ swap_size }} /swapfile"
+    creates: /swapfile
+  when: swap_size != "0" and '/swapfile' not in swap_active.stdout
+
+- name: Secure the swap file (0600)
+  ansible.builtin.file:
+    path: /swapfile
+    owner: root
+    group: root
+    mode: "0600"
+  when: swap_size != "0"
+
+- name: Format and enable the swap file
+  ansible.builtin.shell: "mkswap /swapfile && swapon /swapfile"
+  when: swap_size != "0" and '/swapfile' not in swap_active.stdout
+
+- name: Persist the swap file in /etc/fstab
+  ansible.builtin.lineinfile:
+    path: /etc/fstab
+    line: "/swapfile none swap sw 0 0"
+    regexp: '^/swapfile\s'
+    state: present
+  when: swap_size != "0"
+
+- name: Keep swap a cushion, not a hot path (low swappiness)
+  ansible.builtin.copy:
+    dest: /etc/sysctl.d/60-scrabble-swap.conf
+    mode: "0644"
+    content: "vm.swappiness = {{ swap_swappiness }}\n"
+  register: swappiness_conf
+  when: swap_size != "0"
+
+- name: Apply swappiness now
+  ansible.builtin.command: sysctl --system
+  changed_when: false
+  when: swap_size != "0" and swappiness_conf is changed
+
 # --- Deploy directories --------------------------------------------------------
 
 - name: Create the scrabble base directories

deploy/blackbox/blackbox.yml

@@ -0,0 +1,14 @@
+# blackbox_exporter probe modules. tls_cert opens a verified TLS connection to the edge
+# caddy so Prometheus can read probe_ssl_earliest_cert_expiry — the signal behind the
+# certificate-renewal-failure alert. The SNI is the production public host, whose cert the
+# edge caddy serves once it does its own ACME; on the test contour caddy is HTTP-only
+# (behind the host caddy), so the probe finds nothing on :443 and the cert metric is absent.
+modules:
+  tls_cert:
+    prober: tcp
+    timeout: 5s
+    tcp:
+      tls: true
+      tls_config:
+        server_name: erudit-game.ru
+        insecure_skip_verify: false

deploy/caddy/Caddyfile

@@ -33,6 +33,39 @@
 	# is the prod fix. Background + alternatives (incl. serving h3 for real): docs/EDGE_HTTP3.md.
 	header Alt-Svc clear
 
+	# Maintenance gate. While the deploy holds the flag file /srv/maint/on (touched by
+	# prod-deploy.sh around a rolling swap, removed on the script's exit), serve a static
+	# 503 "works in progress" page instead of proxying to a mid-recreate upstream —
+	# a graceful signal rather than raw 502s. Operator surfaces (/_gm) are excluded so
+	# Grafana stays reachable during the window. Caddy stat()s the flag per request, so
+	# toggling it needs no reload; the page + flag live in the caddy config dir bind-mounted
+	# read-only at /srv/maint (docker-compose.yml). The test contour never sets the flag
+	# (only prod-deploy.sh does), so this is inert there.
+	@maintenance {
+		not path /_gm /_gm/*
+		file {
+			root /srv/maint
+			try_files on
+		}
+	}
+	handle @maintenance {
+		error 503
+	}
+	handle_errors {
+		@maint503 expression {err.status_code} == 503
+		handle @maint503 {
+			root * /srv/maint
+			rewrite * /maintenance.html
+			header Retry-After 120
+			# Distinctive maintenance marker so the SPA can tell a planned window apart from
+			# a transient upstream 503 and show its own dimmed overlay (an in-session user
+			# never sees this static page — it only renders on a fresh load). The header
+			# rides every gated response, incl. the Connect/gRPC edge the app polls.
+			header X-Scrabble-Maintenance 1
+			file_server
+		}
+	}
+
 	# Operator surfaces under /_gm: a single shared Basic-Auth, then route.
 	@gm path /_gm /_gm/*
 	handle @gm {

deploy/caddy/maintenance.html

@@ -0,0 +1,44 @@
+<!doctype html>
+<!--
+  Served with 503 by the edge caddy while the deploy holds the maintenance flag
+  (/srv/maint/on, toggled by deploy/prod-deploy.sh around a rolling swap). Static and
+  self-contained (no upstream, no external assets) so it renders while the backend /
+  gateway are mid-recreate.
+-->
+<html lang="ru">
+<head>
+  <meta charset="utf-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1">
+  <meta name="robots" content="noindex">
+  <title>Эрудит — технические работы</title>
+  <style>
+    :root { color-scheme: light dark; }
+    html, body { height: 100%; margin: 0; }
+    body {
+      display: flex; align-items: center; justify-content: center;
+      font-family: system-ui, -apple-system, "Segoe UI", Roboto, sans-serif;
+      background: #f4f1ea; color: #2b2b2b;
+      padding: 24px; box-sizing: border-box;
+    }
+    @media (prefers-color-scheme: dark) { body { background: #1d1b17; color: #e8e4da; } }
+    .card { max-width: 30rem; text-align: center; line-height: 1.5; }
+    .tile {
+      display: inline-block; width: 3.5rem; height: 3.5rem; line-height: 3.5rem;
+      font-size: 2rem; font-weight: 700; border-radius: 10px;
+      background: #d9b451; color: #2b2b2b; box-shadow: 0 2px 4px rgba(0,0,0,.25);
+      margin-bottom: 1.25rem;
+    }
+    h1 { font-size: 1.4rem; margin: 0 0 .5rem; }
+    p { margin: .35rem 0; }
+    .en { opacity: .7; font-size: .95rem; }
+  </style>
+</head>
+<body>
+  <main class="card">
+    <div class="tile">Э</div>
+    <h1>Технические работы</h1>
+    <p>Идёт короткое обновление игры. Мы скоро вернёмся — обновите страницу через пару минут.</p>
+    <p class="en">Scrabble is briefly down for an update. Please refresh in a couple of minutes.</p>
+  </main>
+</body>
+</html>

deploy/docker-compose.yml

@@ -151,6 +151,10 @@ services:
       BACKEND_SMTP_PASSWORD: ${SMTP_RELAY_PASS:-}
       BACKEND_SMTP_FROM: ${SMTP_RELAY_FROM:-no-reply@localhost}
       BACKEND_PUBLIC_BASE_URL: ${PUBLIC_BASE_URL:-}
+      # Operator alert emails (new feedback / word complaints): a distinct sender and the
+      # recipient(s) (comma-separated allowed). Both empty disables the alert worker.
+      BACKEND_SMTP_ADMIN_FROM: ${SMTP_RELAY_ADMIN_FROM:-}
+      BACKEND_ADMIN_EMAIL: ${ADMIN_EMAIL:-}
     # The dictionary lives on a named volume seeded from the image on first boot
     # (the image's /opt/dawg is owned by the nonroot UID, which the fresh volume
     # inherits). The admin console writes new version subdirectories here, and the
@@ -186,6 +190,8 @@ services:
         VITE_TELEGRAM_LINK: ${VITE_TELEGRAM_LINK:-}
         VITE_TELEGRAM_GAME_CHANNEL_NAME: ${VITE_TELEGRAM_GAME_CHANNEL_NAME:-}
         VITE_VK_APP_LINK: ${VITE_VK_APP_LINK:-}
+        VITE_VK_APP_ID: ${VITE_VK_APP_ID:-}
+        VITE_VK_ID_REDIRECT_URL: ${VITE_VK_ID_REDIRECT_URL:-}
         VITE_GATEWAY_URL: ${VITE_GATEWAY_URL:-}
         VITE_APP_VERSION: ${APP_VERSION:-dev}
         # Go binary version (the SPA's VITE_APP_VERSION is the same git tag).
@@ -203,6 +209,14 @@ services:
       # app's protected key (a pure offline HMAC, no VK API round-trip). Empty disables
       # the VK auth path (auth.vk is then unregistered).
       GATEWAY_VK_APP_SECRET: ${GATEWAY_VK_APP_SECRET:-}
+      # VK ID web login (browser VK-identity linking): the confidential OAuth 2.1 code
+      # exchange runs server-side under the VK ID "Web" app's protected key. This is a
+      # SEPARATE VK app from the Mini App above, so the credentials are distinct. The app id
+      # and redirect URL are the same values the SPA builds its authorize URL from (one
+      # source each). All three empty disables the link.vk.* ops.
+      GATEWAY_VK_ID_APP_ID: ${VITE_VK_APP_ID:-}
+      GATEWAY_VK_ID_CLIENT_SECRET: ${GATEWAY_VK_ID_CLIENT_SECRET:-}
+      GATEWAY_VK_ID_REDIRECT_URL: ${VITE_VK_ID_REDIRECT_URL:-}
       # The reverse bot-link: the bot dials :9443 over mTLS; the backend admin relay
       # reaches the gateway at :9092 (plaintext, internal). In the test contour both
       # listeners stay on the internal network (the bot shares the VPN netns); in prod
@@ -213,10 +227,12 @@ services:
       GATEWAY_BOTLINK_TLS_KEY: /certs/gateway.key
       GATEWAY_BOTLINK_TLS_CA: /certs/ca.crt
       # Anti-abuse IP ban (fail2ban-style), fed by rate-limit rejections and the
-      # honeypot/honeytoken. Off by default: it bans by client IP, which is only
+      # honeypot/honeytoken. Off by default: it bans by client IP, which is only real
-      # real in prod — the test contour arrives as one shared NAT address, so a ban
+      # in prod — the test contour arrives as one shared NAT address, so a ban there
-      # there would be self-inflicted (the honeypot/honeytoken still log). Prod sets
+      # would be self-inflicted (the honeypot still logs). The prod deploy forces it on
-      # these from PROD_ inputs; GATEWAY_HONEYTOKEN is the planted bearer trap.
+      # in env.sh (deploy/write-prod-env.sh), not via a Gitea variable. GATEWAY_HONEYTOKEN
+      # is the planted bearer trap, fed from the per-contour TEST_/PROD_GATEWAY_HONEYTOKEN
+      # secret; empty (unset secret) leaves the trap off.
       GATEWAY_ABUSE_BAN_ENABLED: ${GATEWAY_ABUSE_BAN_ENABLED:-false}
       GATEWAY_HONEYTOKEN: ${GATEWAY_HONEYTOKEN:-}
       GATEWAY_LOG_LEVEL: ${LOG_LEVEL:-info}
@@ -260,6 +276,8 @@ services:
         VITE_TELEGRAM_LINK: ${VITE_TELEGRAM_LINK:-}
         VITE_TELEGRAM_GAME_CHANNEL_NAME: ${VITE_TELEGRAM_GAME_CHANNEL_NAME:-}
         VITE_VK_APP_LINK: ${VITE_VK_APP_LINK:-}
+        VITE_VK_APP_ID: ${VITE_VK_APP_ID:-}
+        VITE_VK_ID_REDIRECT_URL: ${VITE_VK_ID_REDIRECT_URL:-}
         VITE_GATEWAY_URL: ${VITE_GATEWAY_URL:-}
         VITE_APP_VERSION: ${APP_VERSION:-dev}
     restart: unless-stopped
@@ -410,6 +428,11 @@ services:
       GM_BASICAUTH_HASH: ${GM_BASICAUTH_HASH:?set GM_BASICAUTH_HASH}
     volumes:
       - ${SCRABBLE_CONFIG_DIR:-.}/caddy/Caddyfile:/etc/caddy/Caddyfile:ro
+      # Maintenance page + toggle flag: the caddy config dir holds maintenance.html and the
+      # `on` flag prod-deploy.sh touches around a rolling swap; the Caddyfile serves a 503
+      # from here while the flag exists (read-only mount — the deploy writes the flag on the
+      # host side). See deploy/caddy/Caddyfile and deploy/prod-deploy.sh.
+      - ${SCRABBLE_CONFIG_DIR:-.}/caddy:/srv/maint:ro
       - caddy-data:/data
     deploy:
       resources:
@@ -497,6 +520,21 @@ services:
       # caddy's Basic-Auth and re-prompts for the password on every dashboard; the
       # dashboards poll and do not need Live.
       GF_LIVE_MAX_CONNECTIONS: "0"
+      # SMTP for alert emails, reusing the shared relay host + credentials + the SERVICE
+      # From/recipient. Grafana's client speaks STARTTLS (not the backend's implicit-TLS
+      # port), so it dials the relay host on its STARTTLS port (GRAFANA_SMTP_PORT).
+      # Disabled unless GF_SMTP_ENABLED.
+      GF_SMTP_ENABLED: ${GF_SMTP_ENABLED:-false}
+      GF_SMTP_HOST: ${SMTP_RELAY_HOST:-}:${GRAFANA_SMTP_PORT:-}
+      GF_SMTP_USER: ${SMTP_RELAY_USER:-}
+      GF_SMTP_PASSWORD: ${SMTP_RELAY_PASS:-}
+      # A BARE address (Grafana rejects the "Name" <addr> form); the deploy derives it from
+      # SMTP_RELAY_SERVICE_FROM, splitting off the display name into GRAFANA_SMTP_FROM_NAME.
+      GF_SMTP_FROM_ADDRESS: ${GRAFANA_SMTP_FROM_ADDRESS:-}
+      GF_SMTP_FROM_NAME: ${GRAFANA_SMTP_FROM_NAME:-Erudit Alerts}
+      GF_SMTP_STARTTLS_POLICY: MandatoryStartTLS
+      # The alert recipient(s), read by the provisioned contact point via $__env{SERVICE_EMAIL}.
+      SERVICE_EMAIL: ${SERVICE_EMAIL:-}
     volumes:
       - ${SCRABBLE_CONFIG_DIR:-.}/grafana/provisioning:/etc/grafana/provisioning:ro
       # Dashboards live under /etc/grafana (NOT /var/lib/grafana, which the
@@ -547,6 +585,24 @@ services:
           memory: 64M
     networks: [internal]
 
+  # blackbox_exporter lets Prometheus alert on TLS certificate expiry (a Caddy ACME
+  # renewal failure) via probe_ssl_earliest_cert_expiry. It probes the edge caddy that
+  # terminates TLS (prod: the published scrabble-caddy; the test contour has no compose
+  # caddy, so the probe simply finds no target and the cert metric is absent — the rule is
+  # absent-safe). See prometheus.yml and grafana alerting rules.
+  blackbox_exporter:
+    container_name: scrabble-blackbox-exporter
+    image: prom/blackbox-exporter:v0.25.0
+    restart: unless-stopped
+    logging: *default-logging
+    volumes:
+      - ${SCRABBLE_CONFIG_DIR:-.}/blackbox/blackbox.yml:/etc/blackbox_exporter/config.yml:ro
+    deploy:
+      resources:
+        limits:
+          memory: 64M
+    networks: [internal, edge]
+
 networks:
   internal:
     name: scrabble-internal

deploy/grafana/provisioning/alerting/contactpoints.yaml

@@ -0,0 +1,13 @@
+# Grafana alerting contact point: the operator's alert mailbox, read from the
+# SERVICE_EMAIL container env (see docker-compose.yml grafana), so it stays per-contour.
+# SERVICE_EMAIL may hold several comma-separated addresses.
+apiVersion: 1
+contactPoints:
+  - orgId: 1
+    name: ops-email
+    receivers:
+      - uid: ops_email
+        type: email
+        settings:
+          addresses: $__env{SERVICE_EMAIL}
+          singleEmail: true

deploy/grafana/provisioning/alerting/policies.yaml

@@ -0,0 +1,10 @@
+# Notification policy: every alert routes to the operator email, grouped so a burst is one
+# message, with a 4-hour re-notify while still firing.
+apiVersion: 1
+policies:
+  - orgId: 1
+    receiver: ops-email
+    group_by: ['alertname']
+    group_wait: 30s
+    group_interval: 5m
+    repeat_interval: 4h

deploy/grafana/provisioning/alerting/rules.yaml

@@ -0,0 +1,214 @@
+# Grafana provisioned alert rules for the Scrabble contour. Each rule is a Prometheus
+# instant query (refId A) fed into a threshold expression (refId C). noDataState/execErrState
+# are OK so an absent metric never raises a false alert — notably cert-expiry, whose blackbox
+# probe has no target on the test contour (caddy is HTTP-only there). Metric names are the
+# real ones Prometheus exposes (edge_request_* from the gateway via the collector,
+# node_*/pg_*/probe_ssl_* from the exporters). All route to the ops-email contact point.
+apiVersion: 1
+groups:
+  - orgId: 1
+    name: scrabble-service
+    folder: Alerts
+    interval: 1m
+    rules:
+      - uid: svc_target_down
+        title: Scrape target down
+        condition: C
+        for: 3m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 300, to: 0 }
+            datasourceUid: prometheus
+            model: { refId: A, expr: up, instant: true }
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: lt, params: [1] }
+        labels: { severity: critical }
+        annotations: { summary: 'A Prometheus scrape target is down (up < 1).' }
+
+      - uid: edge_error_rate
+        title: Gateway internal-error rate high
+        condition: C
+        for: 10m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 600, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: sum by (service_name) (rate(edge_request_duration_count{result="internal"}[5m]))
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: gt, params: [0.05] }
+        labels: { severity: warning }
+        annotations: { summary: 'Sustained internal (5xx-equivalent) errors at the edge.' }
+
+      - uid: edge_latency_p99
+        title: Gateway request latency p99 high
+        condition: C
+        for: 10m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 600, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: histogram_quantile(0.99, sum by (le, service_name) (rate(edge_request_duration_bucket[5m])))
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: gt, params: [1] }
+        labels: { severity: warning }
+        annotations: { summary: 'Edge request p99 latency above 1s.' }
+
+      - uid: tls_cert_expiry
+        title: TLS certificate nearing expiry
+        condition: C
+        for: 15m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 300, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: (probe_ssl_earliest_cert_expiry - time()) / 86400
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: lt, params: [20] }
+        labels: { severity: critical }
+        annotations: { summary: 'Edge TLS cert has under 20 days left — Caddy ACME renewal may have failed.' }
+
+  - orgId: 1
+    name: scrabble-host
+    folder: Alerts
+    interval: 1m
+    rules:
+      - uid: host_mem_low
+        title: Host memory low
+        condition: C
+        for: 5m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 300, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: lt, params: [0.1] }
+        labels: { severity: critical }
+        annotations: { summary: 'Under 10% host memory available.' }
+
+      - uid: host_disk_low
+        title: Host disk low
+        condition: C
+        for: 10m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 300, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: min(node_filesystem_avail_bytes / node_filesystem_size_bytes)
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: lt, params: [0.1] }
+        labels: { severity: critical }
+        annotations: { summary: 'A host filesystem is under 10% free.' }
+
+      - uid: host_cpu_high
+        title: Host CPU saturated
+        condition: C
+        for: 10m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 600, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: 1 - avg(rate(node_cpu_seconds_total{mode="idle"}[5m]))
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: gt, params: [0.9] }
+        labels: { severity: warning }
+        annotations: { summary: 'Host CPU above 90% for 10 minutes.' }
+
+      - uid: pg_connections_high
+        title: Postgres connections high
+        condition: C
+        for: 5m
+        noDataState: OK
+        execErrState: OK
+        data:
+          - refId: A
+            relativeTimeRange: { from: 300, to: 0 }
+            datasourceUid: prometheus
+            model:
+              refId: A
+              expr: sum(pg_stat_activity_count) / max(pg_settings_max_connections)
+              instant: true
+          - refId: C
+            datasourceUid: __expr__
+            model:
+              refId: C
+              type: threshold
+              expression: A
+              conditions:
+                - evaluator: { type: gt, params: [0.8] }
+        labels: { severity: warning }
+        annotations: { summary: 'Postgres using over 80% of max_connections.' }

deploy/prod-deploy.sh

@@ -42,6 +42,15 @@ PREV_STATE_FILE="${PREV_STATE_FILE:-/opt/scrabble/PREVIOUS_TAG}"
 PG_USER="${POSTGRES_USER:-scrabble}"
 PG_DB="${POSTGRES_DB:-scrabble}"
 
+# Edge maintenance page. caddy serves a static 503 "works in progress" page for the
+# user-facing routes while this flag file exists (deploy/caddy/Caddyfile). We hold it
+# across the roll / migration window and clear it on ANY exit — success, a health
+# failure + rollback, or an unexpected error — via the trap, so users get a graceful
+# page instead of raw mid-swap 502s and the flag can never get stuck on.
+MAINT_FLAG="${MAINT_FLAG:-${SCRABBLE_CONFIG_DIR:-/opt/scrabble}/caddy/on}"
+maint_off() { rm -f "$MAINT_FLAG" 2>/dev/null || true; }
+trap maint_off EXIT
+
 cd "$COMPOSE_DIR" || { echo "compose dir $COMPOSE_DIR missing"; exit 1; }
 export REGISTRY
 # otelcol joins the host docker group to read the socket; the GID varies per host.
@@ -119,6 +128,13 @@ if [ -z "$(docker ps -aq -f name=scrabble-backend)" ]; then
   exit 0
 fi
 
+# Existing stack: raise the maintenance page for the whole roll (and any migration
+# window). It shows once caddy carries the gate (from this feature's own deploy
+# onward); the EXIT trap lowers it however this run ends.
+mkdir -p "$(dirname "$MAINT_FLAG")"
+: > "$MAINT_FLAG"
+echo "maintenance page raised ($MAINT_FLAG)"
+
 # Migration deploy: freeze writes and snapshot a consistent dump before migrating.
 if [ "$MIGRATION" = 1 ]; then
   echo "migration deploy: opening maintenance window (stopping the backend = the only writer)"

deploy/prometheus/prometheus.yml

@@ -23,3 +23,21 @@ scrape_configs:
   - job_name: node
     static_configs:
       - targets: ["node_exporter:9100"]
+  # TLS certificate expiry of the edge caddy (probe_ssl_earliest_cert_expiry), for the
+  # certificate-renewal-failure alert. Effective on prod, where caddy terminates TLS on
+  # :443; on the test contour caddy is HTTP-only, so the probe finds nothing and the metric
+  # is absent (the alert rule is absent-safe). The exporter probes the target passed as a
+  # scrape parameter and answers on its own :9115.
+  - job_name: blackbox_tls
+    metrics_path: /probe
+    params:
+      module: [tls_cert]
+    static_configs:
+      - targets: ["caddy:443"]
+    relabel_configs:
+      - source_labels: [__address__]
+        target_label: __param_target
+      - source_labels: [__param_target]
+        target_label: instance
+      - target_label: __address__
+        replacement: blackbox_exporter:9115

deploy/write-prod-bot-env.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+# Render the prod bot-host env.bot.sh from the workflow job environment.
+#
+# Sourced identically by prod-deploy (deploy-bot) and prod-rollback
+# (rollback-bot) so the two paths cannot drift (see deploy/write-prod-env.sh
+# for the same rationale on the main host).
+#
+# Usage:  BOT_IMAGE=<image ref> bash deploy/write-prod-bot-env.sh <out-path>
+#
+# Every other value comes from the caller's environment (the job `env:` block).
+out="${1:?usage: write-prod-bot-env.sh <out-path>}"
+
+# The bot's Mini App URL is the same public origin the SPA serves; derive it
+# rather than storing a second copy.
+base="${PUBLIC_BASE_URL%/}"
+TELEGRAM_MINIAPP_URL="$base/telegram/"
+
+cat > "$out" <<EOF
+export SCRABBLE_CONFIG_DIR='/opt/scrabble'
+export BOT_IMAGE='$BOT_IMAGE'
+export BOTLINK_GATEWAY_ADDR='$MAIN_HOST:9443'
+export TELEGRAM_BOT_TOKEN='$TELEGRAM_BOT_TOKEN'
+export TELEGRAM_MINIAPP_URL='$TELEGRAM_MINIAPP_URL'
+export TELEGRAM_GAME_CHANNEL_ID='$TELEGRAM_GAME_CHANNEL_ID'
+export TELEGRAM_CHAT_ID='$TELEGRAM_CHAT_ID'
+export TELEGRAM_SUPPORT_CHAT_ID='$TELEGRAM_SUPPORT_CHAT_ID'
+export TELEGRAM_PROMO_BOT_TOKEN='$TELEGRAM_PROMO_BOT_TOKEN'
+export TELEGRAM_BOT_USERNAME='$TELEGRAM_BOT_USERNAME'
+export TELEGRAM_BOT_LINK='$TELEGRAM_BOT_LINK'
+export LOG_LEVEL='${LOG_LEVEL:-info}'
+EOF

deploy/write-prod-env.sh

@@ -0,0 +1,76 @@
+#!/usr/bin/env bash
+# Render the prod main-host runtime env.sh from the workflow job environment.
+#
+# Sourced identically by prod-deploy (deploy-main) and prod-rollback
+# (rollback-main) so the two paths cannot drift: a rollback recreates the
+# containers, so it must re-render the SAME runtime env a full deploy does —
+# otherwise transactional email, VK login and Grafana alerts silently go dark
+# after a rollback until the next full deploy.
+#
+# Usage:  APP_VERSION=<tag> bash deploy/write-prod-env.sh <out-path>
+#
+# Every other value comes from the caller's environment (the job `env:` block,
+# vars.* / secrets.*). Mirrors the compose interpolation contract in
+# deploy/.env.example; keep in sync with deploy/docker-compose{,.prod}.yml.
+out="${1:?usage: write-prod-env.sh <out-path>}"
+
+# Derive the public URLs from the one canonical origin instead of storing each
+# under its own variable. The path suffixes are structural (SPA routes / the
+# Caddy /_gm sub-path), identical on every contour.
+base="${PUBLIC_BASE_URL%/}"
+TELEGRAM_MINIAPP_URL="$base/telegram/"
+GRAFANA_ROOT_URL="$base/_gm/grafana/"
+VITE_VK_ID_REDIRECT_URL="$base/app/"
+
+# Grafana needs a BARE from-address (it rejects the "Name" <addr> form the backend
+# go-mail accepts, and validates it even when SMTP is disabled — a bad value
+# crash-loops Grafana). Split the display-format SERVICE From into address + name;
+# the backend keeps the full form.
+svc_from="${SMTP_RELAY_SERVICE_FROM:-}"
+GRAFANA_SMTP_FROM_NAME=''
+case "$svc_from" in
+  *"<"*">"*)
+    GRAFANA_SMTP_FROM_ADDRESS="$(printf '%s' "$svc_from" | sed -E 's/.*<([^>]+)>.*/\1/')"
+    GRAFANA_SMTP_FROM_NAME="$(printf '%s' "$svc_from" | sed -E 's/[[:space:]]*<[^>]*>.*$//; s/^"//; s/"$//')" ;;
+  *) GRAFANA_SMTP_FROM_ADDRESS="$svc_from" ;;
+esac
+
+cat > "$out" <<EOF
+export REGISTRY='$REGISTRY'
+export SCRABBLE_CONFIG_DIR='/opt/scrabble'
+export POSTGRES_DB='${POSTGRES_DB:-scrabble}'
+export POSTGRES_USER='${POSTGRES_USER:-scrabble}'
+export POSTGRES_PASSWORD='$POSTGRES_PASSWORD'
+export GM_BASICAUTH_USER='${GM_BASICAUTH_USER:-gm}'
+export GM_BASICAUTH_HASH='$GM_BASICAUTH_HASH'
+export GRAFANA_ADMIN_PASSWORD='$GRAFANA_ADMIN_PASSWORD'
+export GRAFANA_ROOT_URL='$GRAFANA_ROOT_URL'
+export CADDY_SITE_ADDRESS='$CADDY_SITE_ADDRESS'
+export LOG_LEVEL='${LOG_LEVEL:-info}'
+export DICT_VERSION='$DICT_VERSION'
+export APP_VERSION='$APP_VERSION'
+export TELEGRAM_BOT_TOKEN='$TELEGRAM_BOT_TOKEN'
+export TELEGRAM_MINIAPP_URL='$TELEGRAM_MINIAPP_URL'
+export GATEWAY_VK_APP_SECRET='$GATEWAY_VK_APP_SECRET'
+export VITE_VK_APP_ID='$VITE_VK_APP_ID'
+export VITE_VK_ID_REDIRECT_URL='$VITE_VK_ID_REDIRECT_URL'
+export GATEWAY_VK_ID_CLIENT_SECRET='$GATEWAY_VK_ID_CLIENT_SECRET'
+export EXPORT_SIGN_KEY='$EXPORT_SIGN_KEY'
+export SMTP_RELAY_HOST='$SMTP_RELAY_HOST'
+export SMTP_RELAY_PORT='$SMTP_RELAY_PORT'
+export SMTP_RELAY_TLS='$SMTP_RELAY_TLS'
+export SMTP_RELAY_USER='$SMTP_RELAY_USER'
+export SMTP_RELAY_PASS='$SMTP_RELAY_PASS'
+export SMTP_RELAY_FROM='$SMTP_RELAY_FROM'
+export SMTP_RELAY_ADMIN_FROM='$SMTP_RELAY_ADMIN_FROM'
+export ADMIN_EMAIL='$ADMIN_EMAIL'
+export SMTP_RELAY_SERVICE_FROM='$SMTP_RELAY_SERVICE_FROM'
+export GRAFANA_SMTP_FROM_ADDRESS='$GRAFANA_SMTP_FROM_ADDRESS'
+export GRAFANA_SMTP_FROM_NAME='$GRAFANA_SMTP_FROM_NAME'
+export SERVICE_EMAIL='$SERVICE_EMAIL'
+export GRAFANA_SMTP_PORT='$GRAFANA_SMTP_PORT'
+export GF_SMTP_ENABLED='$GF_SMTP_ENABLED'
+export PUBLIC_BASE_URL='$PUBLIC_BASE_URL'
+export GATEWAY_HONEYTOKEN='$GATEWAY_HONEYTOKEN'
+export GATEWAY_ABUSE_BAN_ENABLED='true'
+EOF

docs/ARCHITECTURE.md

@@ -257,11 +257,19 @@ arrive from a platform rather than completing a mandatory registration).
   control of the identity before attaching it: **email** through the confirm-code
   flow, **Telegram** through the web **Login Widget** (validated by the validator,
   HMAC under `SHA-256(bot_token)` — distinct from Mini App initData; the gateway
-  passes the trusted `external_id` to the backend, as for `auth.telegram`). The
+  passes the trusted `external_id` to the backend, as for `auth.telegram`), and **VK**
+  through **VK ID web login** (`gateway/internal/vkid`): the browser runs the VK ID raw
+  OAuth 2.1 flow with PKCE — a full-page redirect to VK's hosted login, no `@vkid/sdk` —
+  and the gateway completes the **confidential code exchange** server-side under the VK
+  "Web" app's protected key (a distinct VK app from the Mini App) to obtain the trusted
+  `external_id`. A browser has no signed launch parameters, so unlike the VK Mini App
+  (offline HMAC, §12) this makes an outbound call to `id.vk.com`, and it is web-only — a
+  full-page redirect would strand a native Mini App webview. The
   request step **always** sends/accepts the proof (no pre-send "already taken"
   signal, so a probe cannot enumerate registered addresses); a required **merge**
   is revealed **only after** the proof is verified and is performed behind an
-  explicit, irreversible confirmation. A free identity is simply attached (and a
+  explicit, irreversible confirmation (for VK, whose authorization code is single-use,
+  the merge re-authorizes for a fresh code). A free identity is simply attached (and a
   guest is promoted to durable, clearing `is_guest`).
 - **Unlink** detaches a platform identity (`telegram`/`vk`) from the profile. The
   backend **refuses removing the last identity** (`ErrLastIdentity`), so an account
@@ -284,8 +292,11 @@ arrive from a platform rather than completing a mandatory registration).
   invitations / friend-codes / drafts / pending-codes. **Chat, feedback and complaints are
   kept.** The orchestration a layer up resigns the account's active games (so opponents are
   not stranded), **drops its all-robot games** (no human opponent; children cascade), and
-  revokes its sessions. Every credential detachment — **unlink, email change and delete** —
+  revokes its sessions. Every credential detachment — **unlink, email change, delete and a
-  writes a `retained_identities` row (`reason`), so the dossier keeps the full timeline.
+  merge collision** — writes a `retained_identities` row (`reason`), so the dossier keeps the
+  full timeline. (A merge that would otherwise leave the survivor with two identities of one
+  kind — e.g. each account held a confirmed email — keeps the primary's and journals the
+  secondary's with `reason=merge` before dropping it.)
   Step-up is a mailed code (`purpose=delete`, no deeplink — a stray click must not delete;
   `ConfirmByToken` refuses a delete token) for an email account, else a typed phrase.
   `last_login_at` / `last_login_ip` are stamped on the cold-load profile fetch (throttled
@@ -1003,6 +1014,19 @@ browser), so an open in-app session reflects it at once.
   both are surfaced on the **Scrabble — Resources** Grafana dashboard, which captures the
   stress-run resource profile. (`docker_stats` replaced cAdvisor, which on the contour
   host resolved only the root cgroup — a separate-XFS `/var/lib/docker`.)
+- **Alerting.** Grafana emails infra alerts through the shared relay (its own SMTP on the
+  STARTTLS port) to `SERVICE_EMAIL`, from provisioned rules
+  (`deploy/grafana/provisioning/alerting/`): a scrape-target down, the gateway's
+  internal-error rate and p99 latency (`edge_request_*`), host memory/disk/CPU
+  (node_exporter), Postgres connection saturation (postgres_exporter), and **TLS certificate
+  expiry < 20 days** — a Caddy ACME-renewal-failure signal from a **`blackbox_exporter`**
+  probe of the edge caddy (`probe_ssl_earliest_cert_expiry`). Every rule is `noDataState=OK`,
+  so an absent metric never false-alerts — notably the cert probe, which has no TLS target on
+  the HTTP-only contour caddy. Separately, the backend's **admin-alert worker**
+  (`internal/adminalert`, started from `main`) emails the operator (`ADMIN_EMAIL`, from a
+  distinct `SMTP_RELAY_ADMIN_FROM`) when new player feedback or word complaints arrive,
+  **coalescing** a burst into one digest per interval; both recipients may be several
+  comma-separated addresses. Both paths are inert unless configured.
 - Per-request server-side timing via gin middleware from day one (the access log
   carries method, route, status, latency and the active trace id). A
   client-measured RTT piggybacked on the next request is a later enhancement.

docs/FUNCTIONAL.md

@@ -108,8 +108,10 @@ account is kept and the guest's games move into it. A merge is blocked only whil
 two accounts share a game still in progress.
 
 The profile lists the account's **sign-in methods**. On the web a player can add
-Telegram; adding VK on the web is not offered yet, and inside a Mini App the host
+Telegram (a login-widget popup) or VK (VK ID web login — a redirect to VK's sign-in and
-platform is already linked. A linked provider can be **unlinked** — except the last
+back); inside a Mini App the host platform is already linked. Linking a provider that
+already belongs to another account offers the same irreversible **merge** as email
+linking. A linked provider can be **unlinked** — except the last
 remaining sign-in method, which is refused so the account stays reachable. **Email is
 never unlinked; it is changed**: the player enters a new address, confirms a code
 mailed to it, and the account switches to it atomically, freeing the old address. A new

docs/FUNCTIONAL_ru.md

@@ -112,8 +112,10 @@ Telegram держит **единого бота**: все игроки поль
 запрещено, только пока у аккаунтов есть общая незавершённая игра.
 
 В профиле перечислены **способы входа** аккаунта. В вебе игрок может добавить
-Telegram; добавление VK в вебе пока не предлагается, а внутри Mini App платформа-хозяин
+Telegram (попап логин-виджета) или VK (веб-вход VK ID — редирект на страницу входа VK и
-уже привязана. Привязанного провайдера можно **отвязать** — кроме последнего
+обратно); внутри Mini App платформа-хозяин уже привязана. Привязка провайдера, уже
+принадлежащего другому аккаунту, предлагает то же необратимое **слияние**, что и привязка
+email. Привязанного провайдера можно **отвязать** — кроме последнего
 оставшегося способа входа: он не отвязывается, чтобы аккаунт оставался достижимым.
 **Email не отвязывают — его меняют**: игрок вводит новый адрес, подтверждает код,
 отправленный на него, и аккаунт атомарно переключается на новый адрес, освобождая

gateway/Dockerfile

@@ -25,12 +25,16 @@ ARG VITE_TELEGRAM_BOT_ID=
 ARG VITE_TELEGRAM_LINK=
 ARG VITE_TELEGRAM_GAME_CHANNEL_NAME=
 ARG VITE_VK_APP_LINK=
+ARG VITE_VK_APP_ID=
+ARG VITE_VK_ID_REDIRECT_URL=
 ARG VITE_GATEWAY_URL=
 ARG VITE_APP_VERSION=
 ENV VITE_TELEGRAM_BOT_ID=$VITE_TELEGRAM_BOT_ID \
     VITE_TELEGRAM_LINK=$VITE_TELEGRAM_LINK \
     VITE_TELEGRAM_GAME_CHANNEL_NAME=$VITE_TELEGRAM_GAME_CHANNEL_NAME \
     VITE_VK_APP_LINK=$VITE_VK_APP_LINK \
+    VITE_VK_APP_ID=$VITE_VK_APP_ID \
+    VITE_VK_ID_REDIRECT_URL=$VITE_VK_ID_REDIRECT_URL \
     VITE_GATEWAY_URL=$VITE_GATEWAY_URL \
     VITE_APP_VERSION=$VITE_APP_VERSION
 

gateway/README.md

@@ -61,6 +61,12 @@ round-trip, then forwards the trusted `vk_user_id` (and the client-read display
 from the signed params) to the backend. When `GATEWAY_VK_APP_SECRET` is unset VK auth is disabled
 (`auth.vk` is unregistered).
 
+`link.vk.confirm`/`merge` link a VK identity from a **browser** (not a Mini App) via **VK ID web
+login** (`internal/vkid`): the SPA runs the VK ID raw OAuth 2.1 flow (PKCE, no `@vkid/sdk`) and
+the gateway completes the **confidential code exchange** at `id.vk.com` under a SEPARATE VK "Web"
+app's protected key — the only op here that calls VK (the Mini App path above is offline). All
+three `GATEWAY_VK_ID_*` unset leaves the ops unregistered.
+
 The message-type catalog: `auth.telegram`, `auth.vk`, `auth.guest`,
 `auth.email.request`, `auth.email.login`, `profile.get`, `game.submit_play`,
 `game.state`, `lobby.enqueue`, `lobby.poll`, `chat.post`, `chat.read` and the play-loop ops;
@@ -72,9 +78,10 @@ refetch). The social/account/history ops —
 `blocks.*`, `invitation.*` (list/create/accept/decline/cancel), `profile.update`,
 `stats.get`, `game.gcg`, and the `notify` live event — go through the identical
 transcode pattern (`transcode_social.go`). Account linking & merge
-— `link.email.request/confirm/merge` and `link.telegram.confirm/merge`
+— `link.email.request/confirm/merge`, `link.telegram.confirm/merge` and
-(`transcode_link.go`); the telegram ops validate the **Login Widget** payload via the
+`link.vk.confirm/merge` (`transcode_link.go`); the telegram ops validate the **Login
-validator (`ValidateLoginWidget`) and forward the trusted `external_id`. These
+Widget** payload via the validator (`ValidateLoginWidget`), the vk ops complete the VK ID
+web code exchange via `internal/vkid`, and both forward the trusted `external_id`. These
 **superseded** the former `email.bind.*` ops, which were removed.
 
 ## Configuration
@@ -89,6 +96,7 @@ validator (`ValidateLoginWidget`) and forward the trusted `external_id`. These
 | `GATEWAY_ADMIN_USER` / `GATEWAY_ADMIN_PASSWORD` | unset | enable + guard the admin console at `/_gm` |
 | `GATEWAY_VALIDATOR_ADDR` | unset | Telegram validator gRPC address (enables initData / Login Widget validation) |
 | `GATEWAY_VK_APP_SECRET` | unset | VK app protected key; enables in-process VK Mini App launch-signature verification (`auth.vk`) |
+| `GATEWAY_VK_ID_APP_ID` / `GATEWAY_VK_ID_CLIENT_SECRET` / `GATEWAY_VK_ID_REDIRECT_URL` | unset | VK ID "Web" app credentials for VK web-login linking (`link.vk.*`): the server-side confidential OAuth 2.1 code exchange. A separate VK app from `GATEWAY_VK_APP_SECRET`; all three required to enable the ops |
 | `GATEWAY_BOTLINK_ADDR` | unset | reverse mTLS bot-link listener the remote bot dials (enables out-of-app push + admin relay) |
 | `GATEWAY_BOTLINK_RELAY_ADDR` | unset | plaintext internal listener serving the backend admin `SendToUser`/`SendToGameChannel` relay |
 | `GATEWAY_BOTLINK_TLS_CERT` / `_KEY` / `_CA` | unset | gateway server cert, its key, and the CA that signs accepted bot client certs (required when `GATEWAY_BOTLINK_ADDR` is set) |

gateway/cmd/gateway/main.go

@@ -33,6 +33,7 @@ import (
 	"scrabble/gateway/internal/ratelimit"
 	"scrabble/gateway/internal/session"
 	"scrabble/gateway/internal/transcode"
+	"scrabble/gateway/internal/vkid"
 	"scrabble/pkg/mtls"
 	botlinkv1 "scrabble/pkg/proto/botlink/v1"
 	telegramv1 "scrabble/pkg/proto/telegram/v1"
@@ -190,7 +191,15 @@ func run(ctx context.Context, cfg config.Config, logger *zap.Logger) error {
 		logger.Info("admin console disabled (set GATEWAY_ADMIN_USER and GATEWAY_ADMIN_PASSWORD)")
 	}
 
-	registry := transcode.NewRegistry(backend, validator, transcode.WithVKAuth(cfg.VKAppSecret))
+	// VK ID web login (browser VK-identity linking) is optional: build the confidential
+	// code-exchanger only when fully configured, else leave the interface nil so the
+	// link.vk.* ops stay unregistered.
+	var vkidExchanger transcode.VKIDExchanger
+	if cfg.VKID.Enabled() {
+		vkidExchanger = vkid.New(cfg.VKID.AppID, cfg.VKID.ClientSecret, cfg.VKID.RedirectURI)
+		logger.Info("vk id web login enabled")
+	}
+	registry := transcode.NewRegistry(backend, validator, transcode.WithVKAuth(cfg.VKAppSecret), transcode.WithVKLink(vkidExchanger))
 	edge := connectsrv.NewServer(connectsrv.Deps{
 		Registry:     registry,
 		Sessions:     sessions,

gateway/internal/backendclient/api_social.go

@@ -335,6 +335,24 @@ func (c *Client) LinkTelegramMerge(ctx context.Context, userID, externalID strin
 	return out, err
 }
 
+// LinkVK attaches a gateway-validated VK identity to the caller or reports a required
+// merge. externalID is the trusted vk user id resolved from the VK ID code exchange.
+func (c *Client) LinkVK(ctx context.Context, userID, externalID string) (LinkResultResp, error) {
+	var out LinkResultResp
+	err := c.do(ctx, http.MethodPost, "/api/v1/user/link/vk", userID, "",
+		map[string]string{"external_id": externalID}, &out)
+	return out, err
+}
+
+// LinkVKMerge merges the account owning a gateway-validated VK identity into the
+// caller's.
+func (c *Client) LinkVKMerge(ctx context.Context, userID, externalID string) (LinkResultResp, error) {
+	var out LinkResultResp
+	err := c.do(ctx, http.MethodPost, "/api/v1/user/link/vk/merge", userID, "",
+		map[string]string{"external_id": externalID}, &out)
+	return out, err
+}
+
 // ChangeEmailRequest asks the backend to mail a confirm-code to a new address for an
 // authenticated email change.
 func (c *Client) ChangeEmailRequest(ctx context.Context, userID, email string) error {

gateway/internal/config/config.go

@@ -36,6 +36,11 @@ type Config struct {
 	// VK launch-parameter signature in-process under it (a pure offline HMAC, no VK API
 	// round-trip). Empty disables the VK auth path (auth.vk is then unregistered).
 	VKAppSecret string
+	// VKID configures the VK ID web login used to link a VK identity from a browser
+	// (the confidential OAuth 2.1 code exchange against id.vk.com). It belongs to a
+	// separate VK "Web" app from VKAppSecret's Mini App, so its credentials are distinct.
+	// Any field empty disables the VK web-link ops (link.vk.*).
+	VKID VKIDConfig
 	// BotLink configures the reverse mTLS channel to the remote Telegram bot. An
 	// empty BotLink.Addr disables the bot channel (out-of-app push and admin relay).
 	BotLink BotLinkConfig
@@ -161,6 +166,22 @@ func DefaultAbuse() AbuseConfig {
 	}
 }
 
+// VKIDConfig holds the VK ID web-login credentials for the confidential
+// authorization-code exchange. AppID is the VK "Web" app's client id; ClientSecret is
+// its protected key; RedirectURI must exactly match the trusted redirect URL registered
+// with the app and the one the frontend uses. All three are required to enable the flow.
+type VKIDConfig struct {
+	AppID        string
+	ClientSecret string
+	RedirectURI  string
+}
+
+// Enabled reports whether VK ID web login is fully configured. When false the gateway
+// leaves the VK web-link ops (link.vk.*) unregistered.
+func (c VKIDConfig) Enabled() bool {
+	return c.AppID != "" && c.ClientSecret != "" && c.RedirectURI != ""
+}
+
 // Load reads the configuration from the environment, applies defaults, and
 // validates the result.
 func Load() (Config, error) {
@@ -174,6 +195,11 @@ func Load() (Config, error) {
 		AdminPassword:   os.Getenv("GATEWAY_ADMIN_PASSWORD"),
 		ValidatorAddr:   os.Getenv("GATEWAY_VALIDATOR_ADDR"),
 		VKAppSecret:     os.Getenv("GATEWAY_VK_APP_SECRET"),
+		VKID: VKIDConfig{
+			AppID:        os.Getenv("GATEWAY_VK_ID_APP_ID"),
+			ClientSecret: os.Getenv("GATEWAY_VK_ID_CLIENT_SECRET"),
+			RedirectURI:  os.Getenv("GATEWAY_VK_ID_REDIRECT_URL"),
+		},
 		SessionCacheMax: defaultSessionCacheMax,
 		RateLimit:       DefaultRateLimit(),
 		Abuse:           DefaultAbuse(),

gateway/internal/transcode/transcode.go

@@ -14,6 +14,7 @@ import (
 	"scrabble/gateway/internal/backendclient"
 	"scrabble/gateway/internal/connector"
 	"scrabble/gateway/internal/vkauth"
+	"scrabble/gateway/internal/vkid"
 	fb "scrabble/pkg/fbs/scrabblefb"
 )
 
@@ -152,6 +153,15 @@ func WithVKAuth(secret string) Option {
 	}
 }
 
+// WithVKLink registers the VK web-link ops (link.vk.confirm/merge), which complete a
+// browser VK ID authorization via the server-side code exchange. A nil exchanger leaves
+// them unregistered, so the ops are unknown wherever VK ID web login is not configured.
+func WithVKLink(ex VKIDExchanger) Option {
+	return func(r *Registry, backend *backendclient.Client) {
+		registerVKLinkOps(r, backend, ex)
+	}
+}
+
 // Lookup returns the operation for messageType, and whether it is registered.
 func (r *Registry) Lookup(messageType string) (Op, bool) {
 	op, ok := r.ops[messageType]
@@ -172,6 +182,9 @@ func DomainCode(err error) (string, bool) {
 	if errors.Is(err, vkauth.ErrInvalid) {
 		return "invalid_vk_params", true
 	}
+	if errors.Is(err, vkid.ErrInvalid) {
+		return "invalid_vk_id", true
+	}
 	if errors.Is(err, connector.ErrInvalidLoginWidget) {
 		return "invalid_login_widget", true
 	}

gateway/internal/transcode/transcode_link.go

@@ -4,6 +4,7 @@ import (
 	"context"
 
 	"scrabble/gateway/internal/backendclient"
+	"scrabble/gateway/internal/vkid"
 	fb "scrabble/pkg/fbs/scrabblefb"
 )
 
@@ -18,6 +19,8 @@ const (
 	MsgLinkEmailMerge     = "link.email.merge"
 	MsgLinkTelegram       = "link.telegram.confirm"
 	MsgLinkTelegramMerge  = "link.telegram.merge"
+	MsgLinkVK             = "link.vk.confirm"
+	MsgLinkVKMerge        = "link.vk.merge"
 	MsgLinkUnlink         = "link.unlink"
 	MsgEmailChangeRequest = "link.email.change.request"
 	MsgEmailChangeConfirm = "link.email.change.confirm"
@@ -154,3 +157,44 @@ func linkTelegramHandler(backend *backendclient.Client, tg TelegramValidator, me
 		return encodeLinkResult(res), nil
 	}
 }
+
+// VKIDExchanger completes a VK ID web authorization-code exchange, returning the
+// launching user's trusted VK identity. It is satisfied by *vkid.Exchanger and lets the
+// VK web-link ops resolve a browser VK login, which — unlike the Mini App path — has no
+// offline launch signature to verify.
+type VKIDExchanger interface {
+	Exchange(ctx context.Context, code, deviceID, codeVerifier string) (vkid.Identity, error)
+}
+
+// registerVKLinkOps adds the VK web-link ops when a VK ID exchanger is configured; a nil
+// exchanger leaves them unregistered (VK ID web login not configured).
+func registerVKLinkOps(r *Registry, backend *backendclient.Client, ex VKIDExchanger) {
+	if ex == nil {
+		return
+	}
+	r.ops[MsgLinkVK] = Op{Handler: linkVKHandler(backend, ex, false), Auth: true}
+	r.ops[MsgLinkVKMerge] = Op{Handler: linkVKHandler(backend, ex, true), Auth: true}
+}
+
+// linkVKHandler completes the VK ID code exchange (server-side, under the app's
+// protected key) and then calls the backend's link or merge endpoint with the trusted
+// VK external id.
+func linkVKHandler(backend *backendclient.Client, ex VKIDExchanger, merge bool) Handler {
+	return func(ctx context.Context, req Request) ([]byte, error) {
+		in := fb.GetRootAsLinkVKRequest(req.Payload, 0)
+		user, err := ex.Exchange(ctx, string(in.Code()), string(in.DeviceId()), string(in.CodeVerifier()))
+		if err != nil {
+			return nil, err
+		}
+		var res backendclient.LinkResultResp
+		if merge {
+			res, err = backend.LinkVKMerge(ctx, req.UserID, user.ExternalID)
+		} else {
+			res, err = backend.LinkVK(ctx, req.UserID, user.ExternalID)
+		}
+		if err != nil {
+			return nil, err
+		}
+		return encodeLinkResult(res), nil
+	}
+}

gateway/internal/transcode/transcode_vk_link_test.go

@@ -0,0 +1,90 @@
+package transcode_test
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"testing"
+
+	flatbuffers "github.com/google/flatbuffers/go"
+
+	"scrabble/gateway/internal/transcode"
+	"scrabble/gateway/internal/vkid"
+	fb "scrabble/pkg/fbs/scrabblefb"
+)
+
+func linkVKPayload(code, deviceID, verifier string) []byte {
+	b := flatbuffers.NewBuilder(64)
+	c := b.CreateString(code)
+	d := b.CreateString(deviceID)
+	v := b.CreateString(verifier)
+	fb.LinkVKRequestStart(b)
+	fb.LinkVKRequestAddCode(b, c)
+	fb.LinkVKRequestAddDeviceId(b, d)
+	fb.LinkVKRequestAddCodeVerifier(b, v)
+	b.Finish(fb.LinkVKRequestEnd(b))
+	return b.FinishedBytes()
+}
+
+// fakeVKIDExchanger records the exchange inputs and returns a canned identity.
+type fakeVKIDExchanger struct {
+	id                              vkid.Identity
+	err                             error
+	gotCode, gotDevice, gotVerifier string
+}
+
+func (f *fakeVKIDExchanger) Exchange(_ context.Context, code, deviceID, codeVerifier string) (vkid.Identity, error) {
+	f.gotCode, f.gotDevice, f.gotVerifier = code, deviceID, codeVerifier
+	return f.id, f.err
+}
+
+func TestLinkVKExchangesAndForwards(t *testing.T) {
+	var gotExternalID string
+	backend, cleanup := fakeBackend(t, func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path != "/api/v1/user/link/vk" {
+			t.Errorf("path = %q", r.URL.Path)
+		}
+		var body struct {
+			ExternalID string `json:"external_id"`
+		}
+		_ = json.NewDecoder(r.Body).Decode(&body)
+		gotExternalID = body.ExternalID
+		_, _ = w.Write([]byte(`{"status":"linked"}`))
+	})
+	defer cleanup()
+
+	ex := &fakeVKIDExchanger{id: vkid.Identity{ExternalID: "777"}}
+	reg := transcode.NewRegistry(backend, nil, transcode.WithVKLink(ex))
+	op, ok := reg.Lookup(transcode.MsgLinkVK)
+	if !ok {
+		t.Fatal("link.vk.confirm not registered")
+	}
+	payload, err := op.Handler(context.Background(), transcode.Request{UserID: "u-1", Payload: linkVKPayload("the-code", "dev-1", "verifier-1")})
+	if err != nil {
+		t.Fatalf("handler: %v", err)
+	}
+	// The PKCE inputs from the wire must reach the exchanger verbatim...
+	if ex.gotCode != "the-code" || ex.gotDevice != "dev-1" || ex.gotVerifier != "verifier-1" {
+		t.Errorf("exchange got %q/%q/%q", ex.gotCode, ex.gotDevice, ex.gotVerifier)
+	}
+	// ...and the resolved vk id must be the one forwarded to the backend.
+	if gotExternalID != "777" {
+		t.Errorf("backend external_id = %q, want 777", gotExternalID)
+	}
+	if string(fb.GetRootAsLinkResult(payload, 0).Status()) != "linked" {
+		t.Error("expected a linked result")
+	}
+}
+
+func TestLinkVKUnregisteredWithoutExchanger(t *testing.T) {
+	backend, cleanup := fakeBackend(t, func(http.ResponseWriter, *http.Request) {})
+	defer cleanup()
+
+	reg := transcode.NewRegistry(backend, nil)
+	if _, ok := reg.Lookup(transcode.MsgLinkVK); ok {
+		t.Error("link.vk.confirm must be unregistered when no VK ID exchanger is configured")
+	}
+	if _, ok := reg.Lookup(transcode.MsgLinkVKMerge); ok {
+		t.Error("link.vk.merge must be unregistered when no VK ID exchanger is configured")
+	}
+}

gateway/internal/vkid/vkid.go

@@ -0,0 +1,163 @@
+// Package vkid completes VK ID web authorization on the server. A browser linking a
+// VK identity has no signed Mini App launch parameters (that offline HMAC path is
+// vkauth); instead the frontend runs the VK ID raw OAuth 2.1 flow (PKCE) and hands the
+// gateway the authorization code, which the gateway exchanges — confidentially, under
+// the app's protected key — for the launching user's trusted VK id. Unlike the Mini App
+// path this makes an outbound call to VK (there is no offline verification for the web
+// flow). See docs/ARCHITECTURE.md §12.
+package vkid
+
+import (
+	"context"
+	"encoding/base64"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"strings"
+	"time"
+)
+
+const (
+	// tokenEndpoint is VK ID's OAuth 2.1 token endpoint. The frontend obtains the
+	// authorization code against the same host, so the confidential exchange targets it
+	// too. It is a fixed constant (not user input), so the outbound request carries no
+	// SSRF risk.
+	tokenEndpoint = "https://id.vk.com/oauth2/auth"
+	// exchangeTimeout bounds one code-for-token exchange. Linking is interactive, so an
+	// unreachable VK must fail fast rather than hold the request open.
+	exchangeTimeout = 10 * time.Second
+	// maxResponseBytes caps the token-response read to bound memory on an oversized body.
+	maxResponseBytes = 1 << 16
+)
+
+// ErrInvalid reports an authorization fault: the exchange was rejected or yielded no vk
+// user id (a bad or expired code, a mismatched verifier or redirect). It is distinct
+// from a transport failure reaching VK, which surfaces as a wrapped error.
+var ErrInvalid = errors.New("vkid: vk id authorization exchange failed")
+
+// Identity is the user resolved from a completed VK ID exchange. ExternalID is the vk
+// user id, used as the identities external_id.
+type Identity struct {
+	ExternalID string
+}
+
+// numericID accepts a VK user id that the token endpoint returns inconsistently as a
+// JSON string or a JSON number, normalising both to their decimal string form (empty
+// for a null or absent field).
+type numericID string
+
+func (n *numericID) UnmarshalJSON(b []byte) error {
+	s := strings.Trim(string(b), `"`)
+	if s == "null" {
+		s = ""
+	}
+	*n = numericID(s)
+	return nil
+}
+
+// Exchanger completes VK ID confidential authorization-code exchanges for one app.
+type Exchanger struct {
+	appID        string
+	clientSecret string
+	redirectURI  string
+	endpoint     string
+	httpClient   *http.Client
+}
+
+// New constructs an Exchanger for the app credentials. redirectURI must equal the
+// trusted redirect URL registered with the app and the one the frontend used, or VK
+// rejects the exchange.
+func New(appID, clientSecret, redirectURI string) *Exchanger {
+	return &Exchanger{
+		appID:        appID,
+		clientSecret: clientSecret,
+		redirectURI:  redirectURI,
+		endpoint:     tokenEndpoint,
+		httpClient:   &http.Client{Timeout: exchangeTimeout},
+	}
+}
+
+// Exchange completes the authorization-code grant and returns the trusted vk user id.
+// code, deviceID and codeVerifier are the PKCE inputs the frontend obtained from the VK
+// ID authorization redirect; the exchange authenticates with the app's protected key.
+func (e *Exchanger) Exchange(ctx context.Context, code, deviceID, codeVerifier string) (Identity, error) {
+	if code == "" || deviceID == "" || codeVerifier == "" {
+		return Identity{}, ErrInvalid
+	}
+	form := url.Values{
+		"grant_type":    {"authorization_code"},
+		"code":          {code},
+		"code_verifier": {codeVerifier},
+		"device_id":     {deviceID},
+		"client_id":     {e.appID},
+		"client_secret": {e.clientSecret},
+		"redirect_uri":  {e.redirectURI},
+	}
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, e.endpoint, strings.NewReader(form.Encode()))
+	if err != nil {
+		return Identity{}, fmt.Errorf("vkid: build exchange request: %w", err)
+	}
+	req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
+	req.Header.Set("Accept", "application/json")
+	resp, err := e.httpClient.Do(req)
+	if err != nil {
+		return Identity{}, fmt.Errorf("vkid: exchange request: %w", err)
+	}
+	defer resp.Body.Close()
+	body, err := io.ReadAll(io.LimitReader(resp.Body, maxResponseBytes))
+	if err != nil {
+		return Identity{}, fmt.Errorf("vkid: read exchange response: %w", err)
+	}
+	if resp.StatusCode != http.StatusOK {
+		// VK answers 400 with {error, error_description} on a bad/expired code or a
+		// verifier/redirect mismatch — an authorization fault, not a transport error.
+		return Identity{}, ErrInvalid
+	}
+	var out struct {
+		UserID  numericID `json:"user_id"`
+		IDToken string    `json:"id_token"`
+		Error   string    `json:"error"`
+	}
+	if err := json.Unmarshal(body, &out); err != nil {
+		return Identity{}, ErrInvalid
+	}
+	if out.Error != "" {
+		return Identity{}, ErrInvalid
+	}
+	uid := string(out.UserID)
+	if uid == "" || uid == "0" {
+		uid = subjectFromIDToken(out.IDToken)
+	}
+	if uid == "" {
+		return Identity{}, ErrInvalid
+	}
+	return Identity{ExternalID: uid}, nil
+}
+
+// subjectFromIDToken extracts the OIDC subject (the vk user id) from the payload of a
+// VK ID id_token. The token arrives inside a direct TLS response from VK, so its
+// signature is not re-verified here; the claim is only a fallback for a response that
+// omits an explicit user_id.
+func subjectFromIDToken(idToken string) string {
+	parts := strings.Split(idToken, ".")
+	if len(parts) != 3 {
+		return ""
+	}
+	payload, err := base64.RawURLEncoding.DecodeString(parts[1])
+	if err != nil {
+		return ""
+	}
+	var claims struct {
+		Sub numericID `json:"sub"`
+	}
+	if err := json.Unmarshal(payload, &claims); err != nil {
+		return ""
+	}
+	if s := string(claims.Sub); s != "0" {
+		return s
+	}
+	return ""
+}

gateway/internal/vkid/vkid_test.go

@@ -0,0 +1,131 @@
+package vkid
+
+import (
+	"context"
+	"encoding/base64"
+	"errors"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"testing"
+)
+
+// testExchanger builds an Exchanger pointed at a test server.
+func testExchanger(endpoint string) *Exchanger {
+	return &Exchanger{
+		appID:        "app-1",
+		clientSecret: "secret-1",
+		redirectURI:  "https://example.test/app/",
+		endpoint:     endpoint,
+		httpClient:   &http.Client{},
+	}
+}
+
+func TestExchangeSuccessSendsConfidentialPKCE(t *testing.T) {
+	var gotForm url.Values
+	var gotContentType string
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotContentType = r.Header.Get("Content-Type")
+		body, _ := io.ReadAll(r.Body)
+		gotForm, _ = url.ParseQuery(string(body))
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = io.WriteString(w, `{"user_id":"12345","access_token":"a","id_token":"h.e.s"}`)
+	}))
+	defer srv.Close()
+
+	id, err := testExchanger(srv.URL).Exchange(context.Background(), "the-code", "dev-9", "verifier-xyz")
+	if err != nil {
+		t.Fatalf("Exchange: %v", err)
+	}
+	if id.ExternalID != "12345" {
+		t.Fatalf("ExternalID = %q, want 12345", id.ExternalID)
+	}
+	if gotContentType != "application/x-www-form-urlencoded" {
+		t.Errorf("Content-Type = %q", gotContentType)
+	}
+	// The confidential exchange must carry the PKCE inputs and the app credentials.
+	want := map[string]string{
+		"grant_type":    "authorization_code",
+		"code":          "the-code",
+		"code_verifier": "verifier-xyz",
+		"device_id":     "dev-9",
+		"client_id":     "app-1",
+		"client_secret": "secret-1",
+		"redirect_uri":  "https://example.test/app/",
+	}
+	for k, v := range want {
+		if gotForm.Get(k) != v {
+			t.Errorf("form[%s] = %q, want %q", k, gotForm.Get(k), v)
+		}
+	}
+}
+
+func TestExchangeAcceptsNumericUserID(t *testing.T) {
+	// VK returns user_id as a bare JSON number in some responses; it must parse too.
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		_, _ = io.WriteString(w, `{"user_id":1234567890,"access_token":"a"}`)
+	}))
+	defer srv.Close()
+
+	id, err := testExchanger(srv.URL).Exchange(context.Background(), "c", "d", "v")
+	if err != nil {
+		t.Fatalf("Exchange: %v", err)
+	}
+	if id.ExternalID != "1234567890" {
+		t.Fatalf("ExternalID = %q, want 1234567890", id.ExternalID)
+	}
+}
+
+func TestExchangeFallsBackToIDTokenSub(t *testing.T) {
+	sub := "987654321"
+	payload := base64.RawURLEncoding.EncodeToString([]byte(`{"sub":"` + sub + `"}`))
+	idToken := "header." + payload + ".sig"
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		_, _ = io.WriteString(w, `{"access_token":"a","id_token":"`+idToken+`"}`)
+	}))
+	defer srv.Close()
+
+	id, err := testExchanger(srv.URL).Exchange(context.Background(), "c", "d", "v")
+	if err != nil {
+		t.Fatalf("Exchange: %v", err)
+	}
+	if id.ExternalID != sub {
+		t.Fatalf("ExternalID = %q, want %q (from id_token sub)", id.ExternalID, sub)
+	}
+}
+
+func TestExchangeRejectedIsErrInvalid(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusBadRequest)
+		_, _ = io.WriteString(w, `{"error":"invalid_grant","error_description":"code expired"}`)
+	}))
+	defer srv.Close()
+
+	_, err := testExchanger(srv.URL).Exchange(context.Background(), "c", "d", "v")
+	if !errors.Is(err, ErrInvalid) {
+		t.Fatalf("err = %v, want ErrInvalid", err)
+	}
+}
+
+func TestExchangeEmptyInputsFailFast(t *testing.T) {
+	// Missing PKCE inputs are rejected without any network call.
+	ex := testExchanger("http://127.0.0.1:0/never")
+	for _, args := range [][3]string{{"", "d", "v"}, {"c", "", "v"}, {"c", "d", ""}} {
+		if _, err := ex.Exchange(context.Background(), args[0], args[1], args[2]); !errors.Is(err, ErrInvalid) {
+			t.Errorf("Exchange%v err = %v, want ErrInvalid", args, err)
+		}
+	}
+}
+
+func TestExchangeTransportErrorIsNotErrInvalid(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(http.ResponseWriter, *http.Request) {}))
+	srv.Close() // closed listener → connection refused
+	_, err := testExchanger(srv.URL).Exchange(context.Background(), "c", "d", "v")
+	if err == nil {
+		t.Fatal("want a transport error")
+	}
+	if errors.Is(err, ErrInvalid) {
+		t.Fatalf("transport error must not be ErrInvalid: %v", err)
+	}
+}

pkg/fbs/scrabble.fbs

@@ -501,6 +501,17 @@ table LinkTelegramRequest {
   data:string;
 }
 
+// LinkVKRequest carries a VK ID web authorization for attaching a VK identity to the
+// current account. The fields are the PKCE code-exchange inputs the frontend obtains
+// from the VK ID SDK (One Tap): the authorization code, the device id issued with it,
+// and the PKCE code verifier. The gateway completes the confidential code exchange
+// server-side (under the app's protected key) to obtain the trusted vk user id.
+table LinkVKRequest {
+  code:string;
+  device_id:string;
+  code_verifier:string;
+}
+
 // LinkUnlinkRequest detaches a platform identity (kind = "telegram" | "vk") from the
 // caller's account; email is never unlinked (it is changed).
 table LinkUnlinkRequest {

pkg/fbs/scrabblefb/LinkVKRequest.go

@@ -0,0 +1,82 @@
+// Code generated by the FlatBuffers compiler. DO NOT EDIT.
+
+package scrabblefb
+
+import (
+	flatbuffers "github.com/google/flatbuffers/go"
+)
+
+type LinkVKRequest struct {
+	_tab flatbuffers.Table
+}
+
+func GetRootAsLinkVKRequest(buf []byte, offset flatbuffers.UOffsetT) *LinkVKRequest {
+	n := flatbuffers.GetUOffsetT(buf[offset:])
+	x := &LinkVKRequest{}
+	x.Init(buf, n+offset)
+	return x
+}
+
+func FinishLinkVKRequestBuffer(builder *flatbuffers.Builder, offset flatbuffers.UOffsetT) {
+	builder.Finish(offset)
+}
+
+func GetSizePrefixedRootAsLinkVKRequest(buf []byte, offset flatbuffers.UOffsetT) *LinkVKRequest {
+	n := flatbuffers.GetUOffsetT(buf[offset+flatbuffers.SizeUint32:])
+	x := &LinkVKRequest{}
+	x.Init(buf, n+offset+flatbuffers.SizeUint32)
+	return x
+}
+
+func FinishSizePrefixedLinkVKRequestBuffer(builder *flatbuffers.Builder, offset flatbuffers.UOffsetT) {
+	builder.FinishSizePrefixed(offset)
+}
+
+func (rcv *LinkVKRequest) Init(buf []byte, i flatbuffers.UOffsetT) {
+	rcv._tab.Bytes = buf
+	rcv._tab.Pos = i
+}
+
+func (rcv *LinkVKRequest) Table() flatbuffers.Table {
+	return rcv._tab
+}
+
+func (rcv *LinkVKRequest) Code() []byte {
+	o := flatbuffers.UOffsetT(rcv._tab.Offset(4))
+	if o != 0 {
+		return rcv._tab.ByteVector(o + rcv._tab.Pos)
+	}
+	return nil
+}
+
+func (rcv *LinkVKRequest) DeviceId() []byte {
+	o := flatbuffers.UOffsetT(rcv._tab.Offset(6))
+	if o != 0 {
+		return rcv._tab.ByteVector(o + rcv._tab.Pos)
+	}
+	return nil
+}
+
+func (rcv *LinkVKRequest) CodeVerifier() []byte {
+	o := flatbuffers.UOffsetT(rcv._tab.Offset(8))
+	if o != 0 {
+		return rcv._tab.ByteVector(o + rcv._tab.Pos)
+	}
+	return nil
+}
+
+func LinkVKRequestStart(builder *flatbuffers.Builder) {
+	builder.StartObject(3)
+}
+func LinkVKRequestAddCode(builder *flatbuffers.Builder, code flatbuffers.UOffsetT) {
+	builder.PrependUOffsetTSlot(0, flatbuffers.UOffsetT(code), 0)
+}
+func LinkVKRequestAddDeviceId(builder *flatbuffers.Builder, deviceId flatbuffers.UOffsetT) {
+	builder.PrependUOffsetTSlot(1, flatbuffers.UOffsetT(deviceId), 0)
+}
+func LinkVKRequestAddCodeVerifier(builder *flatbuffers.Builder, codeVerifier flatbuffers.UOffsetT) {
+	builder.PrependUOffsetTSlot(2, flatbuffers.UOffsetT(codeVerifier), 0)
+}
+func LinkVKRequestEnd(builder *flatbuffers.Builder) flatbuffers.UOffsetT {
+	return builder.EndObject()
+}

ui/e2e/maintenance.spec.ts

@@ -0,0 +1,38 @@
+import { expect, test } from './fixtures';
+
+// The maintenance overlay is raised in prod by the edge 503 marker (X-Scrabble-Maintenance);
+// the mock transport never produces one, so — like the offline indicator — the e2e drives it
+// through the window.__maint hook (gateway.ts, mock-only). It is app-global (mounted outside
+// the route blocks in App.svelte), so it shows without a session.
+test('maintenance overlay covers the app and lifts on recovery', async ({ page }) => {
+  await page.goto('/');
+  await expect(page.getByRole('alertdialog')).toHaveCount(0);
+
+  // A planned deploy window begins: a non-dismissable dimmed overlay covers the app.
+  await page.evaluate(() => (window as unknown as { __maint: { on(): void } }).__maint.on());
+  const overlay = page.getByRole('alertdialog');
+  await expect(overlay).toBeVisible();
+  // The retry button is present (EN "Try again" / RU "Повторить" depending on locale).
+  await expect(overlay.getByRole('button', { name: /again|Повторить/i })).toBeVisible();
+
+  // The window ends — in prod the store's poll lifts it on the first successful read; the mock
+  // has no probe, so it clears explicitly. The overlay disappears with no page reload.
+  await page.evaluate(() => (window as unknown as { __maint: { off(): void } }).__maint.off());
+  await expect(page.getByRole('alertdialog')).toHaveCount(0);
+});
+
+test('recovery reloads the SPA to pick up the fresh client', async ({ page }) => {
+  await page.goto('/');
+  await page.evaluate(() => (window as unknown as { __maint: { on(): void } }).__maint.on());
+  await expect(page.getByRole('alertdialog')).toBeVisible();
+
+  // Real recovery reloads the page (the deploy may have shipped an incompatible client, and the
+  // running bundle is the old one). Wait for the forced navigation, then confirm the app
+  // re-bootstrapped: the overlay is gone (the store reset by the reload) and the login is back.
+  await Promise.all([
+    page.waitForEvent('load'),
+    page.evaluate(() => (window as unknown as { __maint: { recover(): void } }).__maint.recover()),
+  ]);
+  await expect(page.getByRole('alertdialog')).toHaveCount(0);
+  await expect(page.getByRole('button', { name: /guest/i })).toBeVisible();
+});

ui/src/App.svelte

@@ -9,6 +9,7 @@
   import StaleInviteModal from './components/StaleInviteModal.svelte';
   import WelcomeRedeemModal from './components/WelcomeRedeemModal.svelte';
   import Coachmark from './components/Coachmark.svelte';
+  import MaintenanceOverlay from './components/MaintenanceOverlay.svelte';
   import DebugPanel from './components/DebugPanel.svelte';
   import Login from './screens/Login.svelte';
   import Lobby from './screens/Lobby.svelte';
@@ -128,6 +129,7 @@
 <StaleInviteModal />
 <WelcomeRedeemModal />
 <Coachmark />
+<MaintenanceOverlay />
 
 {#if routeIsLobby && !app.splashDone && !app.blocked && !app.bootError && !app.launchError}
   <Splash />

ui/src/components/MaintenanceOverlay.svelte

@@ -0,0 +1,83 @@
+<script lang="ts">
+  // Non-dismissable maintenance cover. Shown while the edge is in a planned deploy window
+  // (maintenance.svelte.ts, raised from a caddy 503 carrying X-Scrabble-Maintenance). It has
+  // no close affordance — an in-session user cannot dismiss it — but the store's poll lifts
+  // it automatically the moment the gateway answers again; the "retry" button just forces an
+  // immediate re-check. Mirrors the static caddy maintenance page (deploy/caddy/maintenance.html)
+  // so a fresh load and an in-session user see the same thing.
+  import { maintenance, retryNow } from '../lib/maintenance.svelte';
+  import { t } from '../lib/i18n/index.svelte';
+</script>
+
+{#if maintenance.active}
+  <div class="scrim" role="alertdialog" aria-modal="true" aria-labelledby="maint-title" aria-describedby="maint-body">
+    <div class="card">
+      <div class="tile" aria-hidden="true">Э</div>
+      <h1 id="maint-title">{t('maintenance.title')}</h1>
+      <p id="maint-body">{t('maintenance.body')}</p>
+      <button type="button" onclick={retryNow}>{t('maintenance.retry')}</button>
+    </div>
+  </div>
+{/if}
+
+<style>
+  .scrim {
+    position: fixed;
+    inset: 0;
+    /* Above the game (z 60) and toasts (z 50) — a planned window covers everything the user
+       could otherwise interact with; below the dev DebugPanel (z 10000). */
+    z-index: 100;
+    background: rgba(0, 0, 0, 0.55);
+    display: grid;
+    place-items: center;
+    padding: 24px;
+    box-sizing: border-box;
+  }
+  .card {
+    max-width: 22rem;
+    text-align: center;
+    background: var(--surface);
+    color: var(--text);
+    border-radius: var(--radius);
+    box-shadow: var(--shadow);
+    padding: 2rem 1.5rem;
+  }
+  /* Mirrors a placed board tile (as Splash.svelte / the caddy page do). */
+  .tile {
+    display: inline-grid;
+    place-items: center;
+    width: 3.5rem;
+    height: 3.5rem;
+    font-size: 2rem;
+    font-weight: 700;
+    border-radius: 4px;
+    background: var(--tile-bg);
+    color: var(--tile-text);
+    box-shadow:
+      inset 0 -2px 0 var(--tile-edge),
+      2px 0 3px -1px rgba(0, 0, 0, 0.4);
+    margin-bottom: 1.25rem;
+  }
+  h1 {
+    font-size: 1.25rem;
+    margin: 0 0 0.5rem;
+  }
+  p {
+    margin: 0 0 1.5rem;
+    color: var(--text-muted);
+    line-height: 1.5;
+  }
+  button {
+    font: inherit;
+    padding: 0.55rem 1.4rem;
+    border: 1px solid var(--border);
+    border-radius: var(--radius);
+    background: transparent;
+    color: var(--text);
+    cursor: pointer;
+  }
+  button:hover {
+    border-color: var(--accent);
+    color: var(--accent);
+  }
+</style>

ui/src/gen/fbs/scrabblefb.ts

@@ -54,6 +54,7 @@ export { LinkEmailRequest } from './scrabblefb/link-email-request.js';
 export { LinkResult } from './scrabblefb/link-result.js';
 export { LinkTelegramRequest } from './scrabblefb/link-telegram-request.js';
 export { LinkUnlinkRequest } from './scrabblefb/link-unlink-request.js';
+export { LinkVKRequest } from './scrabblefb/link-vkrequest.js';
 export { MatchFoundEvent } from './scrabblefb/match-found-event.js';
 export { MatchResult } from './scrabblefb/match-result.js';
 export { MoveRecord } from './scrabblefb/move-record.js';

ui/src/gen/fbs/scrabblefb/link-vkrequest.ts

@@ -0,0 +1,72 @@
+// automatically generated by the FlatBuffers compiler, do not modify
+
+import * as flatbuffers from 'flatbuffers';
+
+export class LinkVKRequest {
+  bb: flatbuffers.ByteBuffer|null = null;
+  bb_pos = 0;
+  __init(i:number, bb:flatbuffers.ByteBuffer):LinkVKRequest {
+  this.bb_pos = i;
+  this.bb = bb;
+  return this;
+}
+
+static getRootAsLinkVKRequest(bb:flatbuffers.ByteBuffer, obj?:LinkVKRequest):LinkVKRequest {
+  return (obj || new LinkVKRequest()).__init(bb.readInt32(bb.position()) + bb.position(), bb);
+}
+
+static getSizePrefixedRootAsLinkVKRequest(bb:flatbuffers.ByteBuffer, obj?:LinkVKRequest):LinkVKRequest {
+  bb.setPosition(bb.position() + flatbuffers.SIZE_PREFIX_LENGTH);
+  return (obj || new LinkVKRequest()).__init(bb.readInt32(bb.position()) + bb.position(), bb);
+}
+
+code():string|null
+code(optionalEncoding:flatbuffers.Encoding):string|Uint8Array|null
+code(optionalEncoding?:any):string|Uint8Array|null {
+  const offset = this.bb!.__offset(this.bb_pos, 4);
+  return offset ? this.bb!.__string(this.bb_pos + offset, optionalEncoding) : null;
+}
+
+deviceId():string|null
+deviceId(optionalEncoding:flatbuffers.Encoding):string|Uint8Array|null
+deviceId(optionalEncoding?:any):string|Uint8Array|null {
+  const offset = this.bb!.__offset(this.bb_pos, 6);
+  return offset ? this.bb!.__string(this.bb_pos + offset, optionalEncoding) : null;
+}
+
+codeVerifier():string|null
+codeVerifier(optionalEncoding:flatbuffers.Encoding):string|Uint8Array|null
+codeVerifier(optionalEncoding?:any):string|Uint8Array|null {
+  const offset = this.bb!.__offset(this.bb_pos, 8);
+  return offset ? this.bb!.__string(this.bb_pos + offset, optionalEncoding) : null;
+}
+
+static startLinkVKRequest(builder:flatbuffers.Builder) {
+  builder.startObject(3);
+}
+
+static addCode(builder:flatbuffers.Builder, codeOffset:flatbuffers.Offset) {
+  builder.addFieldOffset(0, codeOffset, 0);
+}
+
+static addDeviceId(builder:flatbuffers.Builder, deviceIdOffset:flatbuffers.Offset) {
+  builder.addFieldOffset(1, deviceIdOffset, 0);
+}
+
+static addCodeVerifier(builder:flatbuffers.Builder, codeVerifierOffset:flatbuffers.Offset) {
+  builder.addFieldOffset(2, codeVerifierOffset, 0);
+}
+
+static endLinkVKRequest(builder:flatbuffers.Builder):flatbuffers.Offset {
+  const offset = builder.endObject();
+  return offset;
+}
+
+static createLinkVKRequest(builder:flatbuffers.Builder, codeOffset:flatbuffers.Offset, deviceIdOffset:flatbuffers.Offset, codeVerifierOffset:flatbuffers.Offset):flatbuffers.Offset {
+  LinkVKRequest.startLinkVKRequest(builder);
+  LinkVKRequest.addCode(builder, codeOffset);
+  LinkVKRequest.addDeviceId(builder, deviceIdOffset);
+  LinkVKRequest.addCodeVerifier(builder, codeVerifierOffset);
+  return LinkVKRequest.endLinkVKRequest(builder);
+}
+}

ui/src/lib/app.svelte.ts

@@ -34,6 +34,7 @@ import {
   telegramCloudSet,
 } from './telegram';
 import { onVKPath, insideVK, vkInit, vkClose, vkDisableSwipeBack, vkLaunchParams, vkUserName, vkStartParam, vkOnScheme, vkOnInsets, vkSetViewSettings, appearanceForBg } from './vk';
+import { pendingVKLink, type VKLinkCallback } from './vkid';
 import { haptic } from './haptics';
 import { CLOUD_PREFS_KEY, decodeClientPrefs, encodeClientPrefs } from './cloudprefs';
 import { parseStartParam } from './deeplink';
@@ -96,6 +97,10 @@ export const app = $state<{
   /** Set once the account has been deleted: App.svelte shows the terminal "account deleted"
    *  screen and all push/poll is stopped. Reopening the app just creates a fresh account. */
   accountDeleted: boolean;
+  /** A VK ID web-link authorization callback captured on boot (see lib/vkid): the browser
+   *  returned from VK with an auth code. Profile consumes it on mount to finish the link or
+   *  merge (a full-page redirect loses the route, so boot routes here). Null on a normal load. */
+  vkLinkPending: VKLinkCallback | null;
   toast: Toast | null;
   lastEvent: PushEvent | null;
   theme: ThemePref;
@@ -145,6 +150,7 @@ export const app = $state<{
   profile: null,
   blocked: null,
   accountDeleted: false,
+  vkLinkPending: null,
   toast: null,
   lastEvent: null,
   theme: 'auto',
@@ -778,9 +784,19 @@ export async function bootstrap(): Promise<void> {
   }
 
   const saved = await loadSession();
+  // A VK ID web-link callback (?code&device_id&state) rides the URL after the redirect back
+  // from VK; capture and clear it here (it needs the restored session to link against).
+  const vkcb = pendingVKLink();
   if (saved) {
     await adoptSession(saved);
-    if (router.route.name === 'login') navigate('/');
+    if (vkcb) {
+      // The full-page redirect lost the in-app route, so hand the callback to Profile, which
+      // finishes the link or merge on mount.
+      app.vkLinkPending = vkcb;
+      navigate('/profile');
+    } else if (router.route.name === 'login') {
+      navigate('/');
+    }
   } else if (router.route.name !== 'login' && router.route.name !== 'confirm') {
     navigate('/login');
   }

ui/src/lib/client.ts

@@ -177,6 +177,10 @@ export interface GatewayClient {
   linkEmailMerge(email: string, code: string): Promise<LinkResult>;
   linkTelegram(data: string): Promise<LinkResult>;
   linkTelegramMerge(data: string): Promise<LinkResult>;
+  /** Link a VK identity from the web: the args are the PKCE outputs of the VK ID
+   *  authorization callback, exchanged for the trusted vk id on the gateway. */
+  linkVK(code: string, deviceId: string, codeVerifier: string): Promise<LinkResult>;
+  linkVKMerge(code: string, deviceId: string, codeVerifier: string): Promise<LinkResult>;
   /** Detach a platform identity (kind 'telegram' | 'vk') from the account; email is never
    *  unlinked. Returns the refreshed link result (status 'unlinked'). */
   linkUnlink(kind: string): Promise<LinkResult>;

ui/src/lib/codec.test.ts

@@ -31,6 +31,7 @@ import {
   decodeDeleteRequestResult,
   encodeGuestLogin,
   encodeLinkUnlink,
+  encodeLinkVK,
   encodeStateRequest,
   encodeSubmitPlay,
   encodeTarget,
@@ -466,6 +467,13 @@ describe('codec', () => {
     expect(r.kind()).toBe('telegram');
   });
 
+  it('encodes a VK link request carrying the PKCE code-exchange inputs', () => {
+    const r = fb.LinkVKRequest.getRootAsLinkVKRequest(new ByteBuffer(encodeLinkVK('the-code', 'dev-1', 'verifier-1')));
+    expect(r.code()).toBe('the-code');
+    expect(r.deviceId()).toBe('dev-1');
+    expect(r.codeVerifier()).toBe('verifier-1');
+  });
+
   it('round-trips the account-delete confirm + request-result wire', () => {
     const c = fb.AccountDeleteConfirm.getRootAsAccountDeleteConfirm(
       new ByteBuffer(encodeAccountDeleteConfirm('123456', 'DELETE')),

ui/src/lib/codec.ts

@@ -739,6 +739,18 @@ export function encodeLinkUnlink(kind: string): Uint8Array {
   return finish(b, fb.LinkUnlinkRequest.endLinkUnlinkRequest(b));
 }
 
+export function encodeLinkVK(code: string, deviceId: string, codeVerifier: string): Uint8Array {
+  const b = new Builder(256);
+  const c = b.createString(code);
+  const d = b.createString(deviceId);
+  const v = b.createString(codeVerifier);
+  fb.LinkVKRequest.startLinkVKRequest(b);
+  fb.LinkVKRequest.addCode(b, c);
+  fb.LinkVKRequest.addDeviceId(b, d);
+  fb.LinkVKRequest.addCodeVerifier(b, v);
+  return finish(b, fb.LinkVKRequest.endLinkVKRequest(b));
+}
+
 export function encodeAccountDeleteConfirm(code: string, phrase: string): Uint8Array {
   const b = new Builder(64);
   const c = b.createString(code);

ui/src/lib/gateway.ts

@@ -6,6 +6,7 @@ import type { GatewayClient } from './client';
 import { MockGateway } from './mock/client';
 import { createTransport } from './transport';
 import { reportOffline, reportOnline } from './connection.svelte';
+import { clearMaintenance, maintenanceRecovered, reportMaintenance } from './maintenance.svelte';
 
 const isMock = import.meta.env.MODE === 'mock';
 
@@ -20,6 +21,15 @@ if (isMock && typeof window !== 'undefined') {
     offline: reportOffline,
     online: reportOnline,
   };
+  // The mock never produces a real 503, so the e2e drives the maintenance overlay directly
+  // (the store's poll is inert in mock — no probe is registered — so `off` clears it).
+  (
+    window as unknown as { __maint?: { on(retryAfterMs?: number): void; off(): void; recover(): void } }
+  ).__maint = {
+    on: (retryAfterMs = 15000) => reportMaintenance(retryAfterMs),
+    off: clearMaintenance,
+    recover: maintenanceRecovered,
+  };
   // Drive the auto-match opponent join deterministically from the e2e (the mock otherwise
   // attaches a robot on a timer).
   (

ui/src/lib/i18n/en.ts

@@ -7,6 +7,9 @@ export const en = {
   'app.brand': 'Erudit',
   'dict.loading': 'Loading…',
   'connection.connecting': 'Connecting…',
+  'maintenance.title': 'Under maintenance',
+  'maintenance.body': 'Scrabble is briefly down for an update. It will be back in a moment — no need to reload the page.',
+  'maintenance.retry': 'Try again',
 
   'blocked.title': 'Account blocked',
   'blocked.permanent': 'Your account is blocked.',
@@ -173,6 +176,7 @@ export const en = {
   'profile.guestLocked': 'Sign in with email to manage your profile.',
   'profile.linkAccount': 'Link an account',
   'profile.linkTelegram': 'Link Telegram',
+  'profile.linkVK': 'Link VK',
   'profile.linked': 'Account linked.',
   'profile.merged': 'Accounts merged.',
   'profile.mergeTitle': 'Merge accounts?',

ui/src/lib/i18n/ru.ts

@@ -8,6 +8,9 @@ export const ru: Record<MessageKey, string> = {
   'app.brand': 'Эрудит',
   'dict.loading': 'Загрузка…',
   'connection.connecting': 'Подключение…',
+  'maintenance.title': 'Технические работы',
+  'maintenance.body': 'Идёт короткое обновление игры. Мы скоро вернёмся — страницу перезагружать не нужно.',
+  'maintenance.retry': 'Повторить',
 
   'blocked.title': 'Учётная запись заблокирована',
   'blocked.permanent': 'Ваша учётная запись заблокирована.',
@@ -173,6 +176,7 @@ export const ru: Record<MessageKey, string> = {
   'profile.guestLocked': 'Войдите по почте, чтобы управлять профилем.',
   'profile.linkAccount': 'Привязать аккаунт',
   'profile.linkTelegram': 'Привязать Telegram',
+  'profile.linkVK': 'Привязать VK',
   'profile.linked': 'Аккаунт привязан.',
   'profile.merged': 'Аккаунты объединены.',
   'profile.mergeTitle': 'Объединить аккаунты?',

ui/src/lib/maintenance.svelte.ts

@@ -0,0 +1,99 @@
+// Global maintenance signal + self-clearing poll. `active` is true while the edge is in a
+// planned deploy window (a caddy 503 carried `X-Scrabble-Maintenance`; see maintenance.ts).
+// A non-dismissable overlay (MaintenanceOverlay.svelte) covers the app while active, and a
+// cheap read is retried on a capped-backoff cadence (mirroring connection.svelte.ts) — the
+// first success lifts the overlay, so it can never get stuck even with no other traffic.
+// Mock mode never reports maintenance, so it simply stays inactive.
+
+import { backoffMs } from './retry';
+
+let active = $state(false);
+let retryHintMs = $state(0);
+let pollTimer: ReturnType<typeof setTimeout> | null = null;
+let attempt = 0;
+let probing = false;
+let probe: (() => Promise<void>) | null = null;
+
+export const maintenance = {
+  /** active is true while the edge is in a planned maintenance window. */
+  get active(): boolean {
+    return active;
+  },
+  /** retryHintMs is the last Retry-After hint in milliseconds (0 when unknown). */
+  get retryHintMs(): number {
+    return retryHintMs;
+  },
+};
+
+/** registerMaintenanceProbe installs the reachability read the poll fires while active
+ *  (the transport wires a cheap authenticated read). */
+export function registerMaintenanceProbe(fn: () => Promise<void>): void {
+  probe = fn;
+}
+
+/** reportMaintenance raises the overlay and starts the self-clearing poll. Idempotent: a
+ *  repeat hit only refreshes the hint, it never restarts a running poll. */
+export function reportMaintenance(retryAfterMs: number): void {
+  retryHintMs = retryAfterMs;
+  if (active) return;
+  active = true;
+  attempt = 0;
+  if (probe) schedulePoll();
+}
+
+/** clearMaintenance lowers the overlay and stops the poll without reloading (a reset, or the
+ *  e2e hook). Prod recovery goes through maintenanceRecovered instead. */
+export function clearMaintenance(): void {
+  active = false;
+  if (pollTimer) {
+    clearTimeout(pollTimer);
+    pollTimer = null;
+  }
+}
+
+/** maintenanceRecovered runs when the edge answers again after a window we were showing.
+ *  The deploy that ended the window may have shipped a client-incompatible change (a wire /
+ *  schema bump), and the running bundle is the OLD one — so reload to pick up the fresh
+ *  client instead of merely hiding the overlay. A no-op unless the overlay was up; the cover
+ *  stays over the brief reload flash. In prod recovery is the only way the overlay clears. */
+export function maintenanceRecovered(): void {
+  if (!active) return;
+  active = false;
+  if (pollTimer) {
+    clearTimeout(pollTimer);
+    pollTimer = null;
+  }
+  if (typeof location !== 'undefined') location.reload();
+}
+
+/** retryNow forces an immediate re-check (the overlay's button). It never closes the
+ *  overlay itself — only a successful probe does. A no-op while a probe is in flight. */
+export function retryNow(): void {
+  if (!active || !probe || probing) return;
+  if (pollTimer) {
+    clearTimeout(pollTimer);
+    pollTimer = null;
+  }
+  attempt = 0;
+  runProbe();
+}
+
+function schedulePoll(): void {
+  pollTimer = setTimeout(runProbe, backoffMs(++attempt));
+}
+
+function runProbe(): void {
+  pollTimer = null;
+  if (!active || !probe || probing) return;
+  probing = true;
+  probe().then(
+    () => {
+      probing = false;
+      maintenanceRecovered();
+    },
+    () => {
+      probing = false;
+      if (active) schedulePoll();
+    },
+  );
+}

ui/src/lib/maintenance.test.ts

@@ -0,0 +1,44 @@
+import { describe, expect, it } from 'vitest';
+import { Code, ConnectError } from '@connectrpc/connect';
+import { maintenanceRetryMs, parseRetryAfterMs } from './maintenance';
+
+// A raw ConnectError as connect-web builds it from a non-Connect HTTP 503: the response
+// headers are preserved as `.metadata` (verified against @connectrpc/connect-web).
+const err503 = (headers: Record<string, string>): ConnectError =>
+  new ConnectError('HTTP 503', Code.Unavailable, new Headers(headers));
+
+describe('maintenanceRetryMs', () => {
+  it('detects the marker and parses Retry-After', () => {
+    expect(maintenanceRetryMs(err503({ 'x-scrabble-maintenance': '1', 'retry-after': '120' }))).toBe(120_000);
+  });
+
+  it('matches the header case-insensitively, defaulting the delay when Retry-After is absent', () => {
+    expect(maintenanceRetryMs(err503({ 'X-Scrabble-Maintenance': '1' }))).toBe(15_000);
+  });
+
+  it('returns null for a 503 without the marker (a plain transient failure)', () => {
+    expect(maintenanceRetryMs(err503({ 'retry-after': '30' }))).toBeNull();
+  });
+
+  it('returns null for non-Connect errors', () => {
+    expect(maintenanceRetryMs(new Error('boom'))).toBeNull();
+    expect(maintenanceRetryMs(undefined)).toBeNull();
+    expect(maintenanceRetryMs('nope')).toBeNull();
+  });
+});
+
+describe('parseRetryAfterMs', () => {
+  it('clamps into the 3s–120s band', () => {
+    expect(parseRetryAfterMs('120')).toBe(120_000);
+    expect(parseRetryAfterMs('1')).toBe(3_000); // floor
+    expect(parseRetryAfterMs('9999')).toBe(120_000); // ceiling
+  });
+
+  it('falls back to the default on absent / garbage / non-positive', () => {
+    expect(parseRetryAfterMs(null)).toBe(15_000);
+    expect(parseRetryAfterMs(undefined)).toBe(15_000);
+    expect(parseRetryAfterMs('soon')).toBe(15_000);
+    expect(parseRetryAfterMs('0')).toBe(15_000);
+    expect(parseRetryAfterMs('-5')).toBe(15_000);
+  });
+});

ui/src/lib/maintenance.ts

@@ -0,0 +1,37 @@
+// Detection of a planned edge maintenance window. During a prod deploy roll the edge
+// caddy returns HTTP 503 with an `X-Scrabble-Maintenance: 1` header (and a `Retry-After`)
+// for every gated route. The SPA keys STRICTLY on that marker — not on any transport
+// `unavailable`, which the "Connecting…" indicator already covers — so a planned window
+// raises the maintenance overlay while an ordinary blip stays a soft reconnect.
+//
+// These helpers are pure (no DOM, no runes) so they unit-test in the node vitest env. The
+// maintenance marker + Retry-After ride on the raw ConnectError's `metadata`, which
+// toGatewayError (retry.ts) discards — so detection must run on the raw error.
+
+import { ConnectError } from '@connectrpc/connect';
+
+const DEFAULT_RETRY_MS = 15_000;
+const MIN_RETRY_MS = 3_000;
+const MAX_RETRY_MS = 120_000;
+
+/**
+ * parseRetryAfterMs converts a `Retry-After` header (delta-seconds — the edge emits
+ * seconds, never an HTTP-date) into a millisecond hint clamped to a sane band. An absent
+ * or unparseable value falls back to a default.
+ */
+export function parseRetryAfterMs(header: string | null | undefined): number {
+  const secs = header == null ? NaN : Number(header);
+  if (!Number.isFinite(secs) || secs <= 0) return DEFAULT_RETRY_MS;
+  return Math.min(MAX_RETRY_MS, Math.max(MIN_RETRY_MS, Math.round(secs * 1000)));
+}
+
+/**
+ * maintenanceRetryMs returns the maintenance `Retry-After` hint in milliseconds when the
+ * thrown transport error is the edge maintenance 503 (carries `X-Scrabble-Maintenance: 1`),
+ * or null for any other error. Header lookup is case-insensitive (Headers.get).
+ */
+export function maintenanceRetryMs(e: unknown): number | null {
+  if (!(e instanceof ConnectError)) return null;
+  if (e.metadata?.get('x-scrabble-maintenance') !== '1') return null;
+  return parseRetryAfterMs(e.metadata.get('retry-after'));
+}

ui/src/lib/mock/client.ts

@@ -674,6 +674,16 @@ export class MockGateway implements GatewayClient {
     this.profile.telegramLinked = true;
     return { ...emptyLinked(), status: 'merged' };
   }
+  async linkVK(_code: string, _deviceId: string, _codeVerifier: string): Promise<LinkResult> {
+    this.profile.isGuest = false;
+    this.profile.vkLinked = true;
+    return emptyLinked();
+  }
+  async linkVKMerge(_code: string, _deviceId: string, _codeVerifier: string): Promise<LinkResult> {
+    this.profile.isGuest = false;
+    this.profile.vkLinked = true;
+    return { ...emptyLinked(), status: 'merged' };
+  }
   async linkUnlink(kind: string): Promise<LinkResult> {
     if (kind === 'telegram') this.profile.telegramLinked = false;
     if (kind === 'vk') this.profile.vkLinked = false;

ui/src/lib/transport.ts

@@ -12,6 +12,8 @@ import { GatewayError, type GatewayClient } from './client';
 import * as codec from './codec';
 import { browserOffset } from './profileValidation';
 import { registerProbe, reportOffline, reportOnline } from './connection.svelte';
+import { maintenanceRecovered, registerMaintenanceProbe, reportMaintenance } from './maintenance.svelte';
+import { maintenanceRetryMs } from './maintenance';
 import { backoffMs, isConnectionCode, retryable, toGatewayError } from './retry';
 
 const MAX_RETRIES = 6;
@@ -28,10 +30,14 @@ export function createTransport(baseUrl: string): GatewayClient {
 
   // The reachability probe the connection watcher fires while offline: a cheap authenticated read
   // (it must reject when there is no session, so the watcher keeps waiting rather than reporting up).
-  registerProbe(async () => {
+  const reachabilityProbe = async (): Promise<void> => {
     if (!token) throw new Error('no session');
     await client.execute({ messageType: 'profile.get', payload: codec.empty(), requestId: '' }, { headers: headers() });
-  });
+  };
+  registerProbe(reachabilityProbe);
+  // The maintenance overlay reuses the same cheap read to poll for the end of a deploy
+  // window; the first success lifts the overlay (maintenance.svelte.ts).
+  registerMaintenanceProbe(reachabilityProbe);
 
   // exec runs one unary op, auto-retrying transient transport failures with capped backoff (so a
   // dropped connection or a rate-limit recovers seamlessly) and driving the global Connecting
@@ -43,6 +49,14 @@ export function createTransport(baseUrl: string): GatewayClient {
         res = await client.execute({ messageType, payload, requestId: '' }, { headers: headers(), signal });
       } catch (e) {
         if (signal?.aborted) throw e; // an intentional cancel (e.g. the tiles moved) — do not retry
+        const maintMs = maintenanceRetryMs(e);
+        if (maintMs !== null) {
+          // A planned deploy window (the edge 503 carried X-Scrabble-Maintenance): raise the
+          // overlay and fail fast — its own poll drives recovery — instead of burning the
+          // retry budget on every read for the length of the window.
+          reportMaintenance(maintMs);
+          throw toGatewayError(e);
+        }
         const err = toGatewayError(e);
         if (retryable(err.code, messageType) && attempt < MAX_RETRIES) {
           reportOffline();
@@ -53,6 +67,9 @@ export function createTransport(baseUrl: string): GatewayClient {
         throw err;
       }
       reportOnline();
+      // A read got through: if the maintenance overlay was up, the deploy window has ended —
+      // reload to pick up the (possibly incompatible) fresh client (maintenance.svelte.ts).
+      maintenanceRecovered();
       if (res.resultCode && res.resultCode !== 'ok') throw new GatewayError(res.resultCode);
       return res.payload;
     }
@@ -268,6 +285,12 @@ export function createTransport(baseUrl: string): GatewayClient {
     async linkTelegramMerge(data) {
       return codec.decodeLinkResult(await exec('link.telegram.merge', codec.encodeLinkTelegram(data)));
     },
+    async linkVK(code, deviceId, codeVerifier) {
+      return codec.decodeLinkResult(await exec('link.vk.confirm', codec.encodeLinkVK(code, deviceId, codeVerifier)));
+    },
+    async linkVKMerge(code, deviceId, codeVerifier) {
+      return codec.decodeLinkResult(await exec('link.vk.merge', codec.encodeLinkVK(code, deviceId, codeVerifier)));
+    },
     async linkUnlink(kind) {
       return codec.decodeLinkResult(await exec('link.unlink', codec.encodeLinkUnlink(kind)));
     },
@@ -304,7 +327,11 @@ export function createTransport(baseUrl: string): GatewayClient {
             if (pe) onEvent(pe);
           }
         } catch (e) {
-          if (!ctrl.signal.aborted) onError?.(toGatewayError(e));
+          if (!ctrl.signal.aborted) {
+            const maintMs = maintenanceRetryMs(e);
+            if (maintMs !== null) reportMaintenance(maintMs);
+            onError?.(toGatewayError(e));
+          }
         }
       })();
       return () => ctrl.abort();

ui/src/lib/vkid.ts

@@ -0,0 +1,122 @@
+// VK ID web login for linking a VK identity to the current account from a browser.
+// Unlike the VK Mini App (whose signed launch params authenticate offline), a browser
+// has no launch signature, so it runs the VK ID raw OAuth 2.1 flow (PKCE, no @vkid/sdk):
+// a full-page redirect to VK's hosted login, then a callback to our redirect URL carrying
+// the authorization code. The gateway completes the confidential code exchange. This is
+// web-only — a full-page redirect would strand a native Mini App webview, where the
+// platform identity is already linked anyway.
+
+import { insideTelegram } from './telegram';
+import { insideVK } from './vk';
+
+function isMock(): boolean {
+  return import.meta.env.MODE === 'mock';
+}
+
+function vkAppId(): string {
+  return ((import.meta.env.VITE_VK_APP_ID as string | undefined) ?? '').trim();
+}
+
+function vkRedirectUrl(): string {
+  return ((import.meta.env.VITE_VK_ID_REDIRECT_URL as string | undefined) ?? '').trim();
+}
+
+/**
+ * vkWebLinkAvailable reports whether the "Link VK" control should be shown: on the plain
+ * web (not inside a Telegram/VK Mini App, where a redirect would break the webview) and
+ * either the mock build or a configured VK ID app id and redirect URL.
+ */
+export function vkWebLinkAvailable(): boolean {
+  if (insideTelegram() || insideVK()) return false;
+  if (isMock()) return true;
+  return vkAppId() !== '' && vkRedirectUrl() !== '';
+}
+
+const STORAGE_KEY = 'vkid.link';
+
+// PendingState is stashed before the redirect and consumed on the callback: the PKCE
+// verifier (needed for the gateway exchange), the CSRF state, and whether this is an
+// initial link or the re-authorization for a merge (VK codes are single-use, so a merge
+// cannot reuse the link's code — it re-authorizes for a fresh one).
+type PendingState = { verifier: string; state: string; mode: 'link' | 'merge' };
+
+/** VKLinkCallback carries a verified VK ID authorization callback for the gateway exchange. */
+export type VKLinkCallback = { code: string; deviceId: string; verifier: string; mode: 'link' | 'merge' };
+
+// b64url encodes bytes as base64url without padding (the PKCE / VK ID alphabet).
+function b64url(bytes: Uint8Array): string {
+  let s = '';
+  for (const b of bytes) s += String.fromCharCode(b);
+  return btoa(s).replace(/\+/g, '-').replace(/\//g, '_').replace(/=+$/, '');
+}
+
+function randomToken(bytes: number): string {
+  const a = new Uint8Array(bytes);
+  crypto.getRandomValues(a);
+  return b64url(a);
+}
+
+async function challengeFrom(verifier: string): Promise<string> {
+  const digest = await crypto.subtle.digest('SHA-256', new TextEncoder().encode(verifier));
+  return b64url(new Uint8Array(digest));
+}
+
+/**
+ * startVKLink begins VK ID web authorization by redirecting the whole tab to VK's hosted
+ * login. The PKCE verifier and CSRF state survive the redirect in sessionStorage; VK
+ * returns to VITE_VK_ID_REDIRECT_URL with ?code&device_id&state, processed on the next
+ * boot via pendingVKLink. mode distinguishes an initial link from a merge re-auth. The
+ * mock build never leaves the SPA.
+ */
+export async function startVKLink(mode: 'link' | 'merge'): Promise<void> {
+  if (isMock()) return;
+  const verifier = randomToken(48); // 64 base64url chars (VK ID requires 43..128)
+  const state = randomToken(24); // 32 base64url chars (VK ID requires >= 32)
+  const challenge = await challengeFrom(verifier);
+  const pending: PendingState = { verifier, state, mode };
+  sessionStorage.setItem(STORAGE_KEY, JSON.stringify(pending));
+  const url = new URL('https://id.vk.com/authorize');
+  url.searchParams.set('response_type', 'code');
+  url.searchParams.set('client_id', vkAppId());
+  url.searchParams.set('redirect_uri', vkRedirectUrl());
+  url.searchParams.set('state', state);
+  url.searchParams.set('code_challenge', challenge);
+  url.searchParams.set('code_challenge_method', 'S256');
+  url.searchParams.set('scope', 'vkid.personal_info');
+  location.assign(url.toString());
+}
+
+// stripCallbackQuery removes the OAuth query params from the URL so a refresh does not
+// re-process the callback, keeping the path and any hash route intact.
+function stripCallbackQuery(): void {
+  if (typeof history !== 'undefined' && history.replaceState) {
+    history.replaceState(null, '', location.pathname + location.hash);
+  }
+}
+
+/**
+ * pendingVKLink extracts and clears a VK ID authorization callback from the current URL,
+ * verifying the returned state against the one stored before the redirect (CSRF). It
+ * returns null on a normal load or when the state does not match, and always strips the
+ * query so a refresh does not re-run it.
+ */
+export function pendingVKLink(): VKLinkCallback | null {
+  if (typeof location === 'undefined') return null;
+  const q = new URLSearchParams(location.search);
+  const code = q.get('code');
+  const deviceId = q.get('device_id');
+  const returnedState = q.get('state');
+  if (!code || !deviceId || !returnedState) return null;
+  const raw = sessionStorage.getItem(STORAGE_KEY);
+  sessionStorage.removeItem(STORAGE_KEY);
+  stripCallbackQuery();
+  if (!raw) return null;
+  let pending: PendingState;
+  try {
+    pending = JSON.parse(raw) as PendingState;
+  } catch {
+    return null;
+  }
+  if (pending.state !== returnedState) return null;
+  return { code, deviceId, verifier: pending.verifier, mode: pending.mode };
+}

ui/src/screens/Profile.svelte

@@ -14,6 +14,7 @@
   import { connection } from '../lib/connection.svelte';
   import { gateway } from '../lib/gateway';
   import { loginWidgetAvailable, requestTelegramLogin } from '../lib/telegram';
+  import { startVKLink, vkWebLinkAvailable } from '../lib/vkid';
   import { t } from '../lib/i18n/index.svelte';
   import {
     awayDurationOk,
@@ -45,7 +46,7 @@
   let emailSent = $state(false);
   // A pending irreversible merge surfaced after the code/widget was verified; the
   // dialog confirms it. tgData holds the Telegram widget payload for the merge step.
-  let pendingMerge = $state<null | { kind: 'email' | 'telegram'; name: string; games: number; friends: number }>(null);
+  let pendingMerge = $state<null | { kind: 'email' | 'telegram' | 'vk'; name: string; games: number; friends: number }>(null);
   let tgData = '';
   // The change-email sub-form is open (a new address is being entered/confirmed).
   let changingEmail = $state(false);
@@ -56,6 +57,7 @@
   let deleteCode = $state('');
   let deletePhrase = $state('');
   const telegramLinkable = loginWidgetAvailable();
+  const vkLinkable = vkWebLinkAvailable();
 
   function defaultTz(): string {
     const b = browserOffset();
@@ -82,7 +84,12 @@
     notificationsInAppOnly = p.notificationsInAppOnly;
     variantPrefs = [...p.variantPreferences];
   }
-  onMount(populate);
+  onMount(() => {
+    populate();
+    // A VK ID web-link callback captured on boot (app.vkLinkPending) is finished here, since
+    // the redirect back from VK lost the in-app route and lands the app on a fresh mount.
+    if (app.vkLinkPending) void processVKLink();
+  });
 
   const awayStart = $derived(`${startH}:${startM}`);
   const awayEnd = $derived(`${endH}:${endM}`);
@@ -180,8 +187,48 @@
     }
   }
 
+  // addVK starts VK ID web login for linking: it redirects the whole tab to VK's hosted login,
+  // returning to the app with an auth code that boot hands back to processVKLink.
+  function addVK() {
+    void startVKLink('link');
+  }
+
+  // processVKLink finishes a VK ID web-link callback captured on boot: it exchanges the code via
+  // the gateway and either completes the link or opens the merge dialog. VK's authorization code
+  // is single-use, so a required merge re-authorizes for a fresh code (see confirmMerge); that
+  // returning 'merge' callback lands back here and completes.
+  async function processVKLink() {
+    const cb = app.vkLinkPending;
+    app.vkLinkPending = null;
+    if (!cb) return;
+    try {
+      if (cb.mode === 'merge') {
+        await applyLinkResult(await gateway.linkVKMerge(cb.code, cb.deviceId, cb.verifier));
+        populate();
+        showToast(t('profile.merged'));
+        return;
+      }
+      const r = await gateway.linkVK(cb.code, cb.deviceId, cb.verifier);
+      if (r.status === 'merge_required') {
+        pendingMerge = { kind: 'vk', name: r.secondaryDisplayName, games: r.secondaryGames, friends: r.secondaryFriends };
+        return;
+      }
+      await applyLinkResult(r);
+      populate();
+      showToast(t('profile.linked'));
+    } catch (e) {
+      handleError(e);
+    }
+  }
+
   async function confirmMerge() {
     if (!pendingMerge) return;
+    // A VK merge cannot reuse VK's single-use authorization code, so it re-authorizes for a fresh
+    // one; the returning callback completes it (processVKLink). Email/Telegram re-send their proof.
+    if (pendingMerge.kind === 'vk') {
+      void startVKLink('merge');
+      return;
+    }
     try {
       const r =
         pendingMerge.kind === 'email'
@@ -344,7 +391,7 @@
 
     <!-- Sign-in methods: bind/change an email and link/unlink providers. A returning email
          (add) triggers the merge dialog below. On the web an account can add Telegram via the
-         login widget; adding VK on the web is deferred (no VK OAuth) and inside a Mini App the
+         login widget or VK via VK ID web login (a full-page redirect); inside a Mini App the
          host provider is already linked. Email is never unlinked — it is changed. Unlink is
          offered only when another identity remains (the backend also refuses the last one). -->
     <section class="accounts">
@@ -413,6 +460,10 @@
             <button class="ghost danger" onclick={() => (confirmUnlink = { kind: 'vk', label: 'VK' })} disabled={!connection.online}>{t('profile.unlink')}</button>
           {/if}
         </div>
+      {:else if vkLinkable}
+        <button class="ghost tg" onclick={addVK} disabled={!connection.online}>
+          <img src="vk-logo.svg" alt="" width="18" height="18" />{t('profile.linkVK')}
+        </button>
       {/if}
     </section>
 
@@ -644,6 +695,12 @@
     display: flex;
     gap: 8px;
   }
+  /* When a button row sits below its own field (the change-email and delete dialogs),
+     separate them and keep the actions right-aligned. */
+  .addrow.end {
+    margin-top: 14px;
+    justify-content: flex-end;
+  }
   .addrow input {
     flex: 1;
     min-width: 0;