MenuMenu — the whole outegro.dev stack

House special: cooked on the server, served with a fresh nonce.

In production Every web app of the platform. Server rendering, standalone builds and a strict content security policy with a nonce per request.

Taste it liveoutegro.devid.outegro.devbattleship.outegro.devedu.outegro.dev

A classic. The compiler decides what to memoize, so the cook can rest.

In production Components in all six apps, compiled by the React Compiler everywhere except the game, whose MobX observers opt out.

Strict diet: no any.

In production One language across the monorepo — apps, services and the contracts between them.

Utility-class canapés with accessibility in the filling.

In production The shared @outegro/ui kit: tokens, components and self-hosted fonts for every app; Radix takes care of keyboard and focus.

A multilingual menu: your language is remembered at every table.

In production Every app is multilingual; the chosen language lives in a cookie shared by all outegro.dev subdomains.

Reactive sauce: change an ingredient and the plate updates itself.

In production State of the game, the payment pages, the books and the operator console, in small stores with one job each.

Taste it livebattleship.outegro.dev

A paper diorama at twelve frames a second, like an old cartoon.

In production The paper diorama and the paper boat on this site: frames render only on demand, and without a real GPU or with reduced motion the plain layout stays.

Taste it liveoutegro.dev

SQL prepared right at your table.

In production SQL exercises in the books run right in the browser, inside a Web Worker; the server checks the answer.

Taste it liveedu.outegro.dev

The stock every other dish is made from.

In production The runtime of every service and web app, shipped in slim container images.

Served in five pots, one per service.

In production The five services: identity, notifications, payments, the game server and Education.

The bouncer at the door of every request.

In production The same schemas validate requests and define the contracts shared by apps and services.

A wristband at the door: short shelf life, forgeries spotted at once.

In production Short-lived access tokens signed by identity and verified by every service against the keys identity publishes.

Served live: fire a shot and it lands on the table.

In production The live game: turns, shots and reconnects, opened with a one-time ticket.

Taste it livebattleship.outegro.dev

The waiter: walks to the kitchen himself and never hands the guest a token.

In production Web apps call the services from the server; the browser holds only HttpOnly cookies, never tokens.

Bill, please. The webhook knocks twice; the charge happens once.

In production Checkout, subscriptions and refunds; webhooks are verified, stored and applied once, even when they arrive again.

Taste it livebattleship.outegro.dev/shop

A note from the chef, in the guest’s language.

In production Sign-in codes, receipts and security notices by email, in the user’s language.

The bell on the counter, only in Telegram.

In production The Telegram bot for notifications and linking a Telegram account.

Walk in with your Google card; same table as before.

In production Sign-in with Google next to email codes, linked to the same account.

Taste it liveid.outegro.dev

Fingerprint entry, like a restaurant for spies.

In production Sign-in with a passkey (fingerprint, face or device PIN) into the same account, no password needed.

Taste it liveid.outegro.dev

A sous-chef who explains the recipe, just not a hundred times a night.

In production Explains chapters and gives hints on exercises; the model plugs in through its own adapter, with limits per reader and per day.

Taste it liveedu.outegro.dev

Aged 30 days in storage; restores in 81 seconds.

In production A database and a role per service, continuous WAL archiving and a daily backup to Cloudflare R2, kept for 30 days. Restore is drilled: four databases were rebuilt from the backups in 81 seconds, with the same row counts and constraints.

A light drizzle of SQL over the types.

In production Typed schemas and SQL migrations; migrations run before every rollout.

The fridge by the stove: everything needed right now.

In production Refresh-token rotation, rate limits, the matchmaking queue, game rooms and socket tickets.

A rabbit courier: nobody home, he comes back later; hopeless cases go to dead-letter.

In production A payment grants Premium in the game and sends a receipt; failed deliveries retry through delay queues, and those that keep failing wait in dead-letter queues.

A letter goes out only once the till has rung.

In production An event leaves a service only if its transaction committed, and is applied once even when it arrives twice.

One stove, every burner on.

In production Single-node Kubernetes running every app, service and database.

The doorman out front; nobody gets in through the back.

In production DNS, TLS and protection in front; the server accepts web traffic only from Cloudflare.

The maître d’: seats the guests and renews the certificates himself.

In production Ingress and TLS certificates, issued and renewed automatically.

The kitchen cooks strictly from the recipe book in Git. Not tasty? git revert.

In production GitOps: the cluster follows a Git repository; migrations run before each rollout, and a rollback is a git revert.

A dishwasher that tastes every dish on the way.

In production Lint, type checks, tests against real PostgreSQL, Valkey and RabbitMQ, and builds on every pull request and merge; images go to GitHub’s registry.

One oven, one baking schedule for the whole menu.

In production One build graph for every app and service image.

The secret sauce stays in a safe, even in Git.

In production Secrets are kept in Git encrypted; plain values never leave the cluster.

One pantry, shelves in dependency order.

In production One repository: apps, services and shared packages build in dependency order, with caching.

A thermometer in every pot.

In production Metrics from the node, Kubernetes, Traefik, PostgreSQL, RabbitMQ and every service, with alert rules for disks, errors, queues, backups and payments.

The big board in the kitchen: shows what is burning.

In production Metric and log dashboards behind the operator console’s sign-in: Traefik asks the console before letting each request through.

A week of the kitchen log, tokens blotted out with correction fluid.

In production Logs of every pod for a week, searchable by request and correlation id; tokens are masked before they are stored.

The order number on every ticket.

In production Structured JSON logs; one correlation id follows a request through every service it touches.

The watchdog: patrols the kitchen every five minutes and barks into Telegram.

In production Every five minutes: pods, the node, certificates, backups, Argo CD and HTTP checks, with alerts and a daily digest in Telegram.

Every ingredient tasted on its own.

In production Unit and integration tests for the services, shared packages and app stores.

Tasted on real ingredients, no plastic food.

In production Integration tests against real PostgreSQL, Valkey and RabbitMQ, locally and in CI.

A mystery diner in three browsers at once.

In production End-to-end tests for every web app; this site runs them in Chromium, Firefox and WebKit.

Makes sure any guest can sit at any table.

In production Accessibility scans inside the end-to-end tests.

One plating standard: every plate looks the same.

In production One linter and formatter for the whole monorepo.

Robot cooks, each at their own station.

In production Coding agents work in parallel, each in its own git worktree; I set the task, review the result and merge.

A fresh cookbook instead of memories.

In production Current library documentation for the agents, instead of stale memory.

The head chef tastes every plate before it goes out.

In production Agents are fast; typed contracts, tests, lint and review decide what ships.