Live Alpha·Discovery + Earn on Telegram·Last update 25 Apr 2026

Under the Hood

Technology & Security

How LOA is built, secured, observed, and scaled.

Architecture

Next.js 14 App Router miniapp (Vercel edge) → Express backend (Railway staging, Cloud Run production target Q3 2026) → PostgreSQL via Prisma → Redis cache → TON mainnet (Tact smart contracts).

AI Operations

Gemini 2.0-flash as primary inference (Google Generative AI SDK) with Claude hot-swap fallback. Migrating to Vertex AI in Q3 2026 alongside three new user-facing flows: AI quiz generation, anti-bot scoring, personalized daily missions.

Security

HMAC-SHA256 Telegram initData verification, JWT sessions, IP + behavior-aware rate limiting (smartRateLimit), timing-safe admin auth, secrets in Secret Manager, HSTS preload, strict CSP.

Observability

pino structured logging, PostHog product analytics, uptime checks, Cloud Monitoring dashboards for p99 latency, error rate, SQL connection pool, Redis hit rate. Target SLO: 99.9% uptime, p99 <200ms.

Compliance

Wyoming LLC structure; KVKK data protection framework (Turkey); MiCA readiness target Q4 2026 (EU); LOA token classified as utility (Jetton TEP-74), not a security.

Cloud Migration

Four-product Google Cloud stack targeted Q3 2026: BigQuery (analytics warehouse), GKE Autopilot (backend orchestration), Cloud SQL HA (primary Postgres), Vertex AI (7 GM agent endpoints). Six-week migration plan · same $1.8K/mo baseline.