feat: aggregate dashboard metrics server-side

stack.yml

@@ -26,6 +26,8 @@ services:
       NEXT_PUBLIC_APP_URL: "${NEXT_PUBLIC_APP_URL}"
       BETTER_AUTH_URL: "${BETTER_AUTH_URL}"
       BETTER_AUTH_SECRET: "${BETTER_AUTH_SECRET}"
+      REPORTS_CRON_SECRET: "${REPORTS_CRON_SECRET}"
+      REPORTS_CRON_BASE_URL: "${REPORTS_CRON_BASE_URL}"
       # Mantém o SQLite fora do repositório
       DATABASE_URL: "file:/app/data/db.sqlite"
       # Usado para forçar novo rollout a cada deploy (setado pelo CI)
@@ -76,6 +78,9 @@ services:
       - MACHINE_PROVISIONING_SECRET=${MACHINE_PROVISIONING_SECRET}
       - MACHINE_TOKEN_TTL_MS=${MACHINE_TOKEN_TTL_MS:-2592000000}
       - FLEET_SYNC_SECRET=${FLEET_SYNC_SECRET:-}
+      - REPORTS_CRON_SECRET=${REPORTS_CRON_SECRET}
+      - REPORTS_CRON_BASE_URL=${REPORTS_CRON_BASE_URL}
+      - REPORTS_CRON_ENABLED=${REPORTS_CRON_ENABLED:-false}
     deploy:
       mode: replicated
       replicas: 1
@@ -85,8 +90,9 @@ services:
         failure_action: rollback
       resources:
         limits:
-          # Convex vinha sendo finalizado por OOM (ExitCode 137) com limite de 5G.
-          # Aumentamos para 6G para aproveitar melhor os 8G da VPS sem mudar a infra.
+          # Limite conservador para evitar que o backend afete o host inteiro.
+          memory: "12G"
+        reservations:
           memory: "6G"
       restart_policy:
         condition: any
@@ -105,10 +111,10 @@ services:
       - traefik_public
     healthcheck:
       test: ["CMD-SHELL", "curl -sf http://localhost:3210/version >/dev/null || exit 1"]
-      interval: 10s
+      interval: 30s
       timeout: 10s
-      retries: 5
-      start_period: 20s
+      retries: 10
+      start_period: 120s
 
   convex_dashboard:
     image: ghcr.io/get-convex/convex-dashboard:latest