Newsroom platform: a whole newsroom, from draft to audience

Who it is for

News agencies, news sites and press offices that publish dozens of stories a day and need more than a blog with plugins: a newsroom with several desks and several roles, photography and video that must be managed, reporters who need files and press cards, and publication that has to reach the site, Telegram, WhatsApp, RSS and email at once.

This is a finished product, not a project started from zero: 16 independent modules, 293 HTTP endpoints, 135,000 lines of source and 1,120 automated tests, written and proven before you arrive. What is done for you is deployment, archive migration, the public site's design and training your newsroom.

The side paths are defined too, and every one of them leaves a trace: needs revision with a recorded reason for the writer, rejected with a mandatory reason, recall, back to draft, unpublish, and archive and restore.

What matters is that the states are defined in the domain, not in the interface. Nobody routes around a stage with a direct URL or a forgotten button; without the permission, the story does not move. Sixteen organisational roles — from editor-in-chief to desk editor, reporter, photographer and SEO specialist — and permissions at three scopes: mine, my desk, everyone.

Each module owns its domain, application layer and infrastructure, and talks to the others only through a contract: the discipline of microservice boundaries with the simplicity of a single deployment. A layer of architecture tests checks those boundaries at build time — the design rules are not documentation, they are part of the build, and breaking one turns it red.

Persian-first, not translated

Localisation here is not a skin over a foreign product; it is in the architecture.

• The Jalali calendar throughout. Every date — the scheduling picker, the public archive, the event log — is Jalali, with Persian numerals everywhere. The conversion happens in one place, not in twenty places with twenty results.
• The newsroom clock. Time is stored in UTC and shown in Tehran time everywhere, so a 00:30 publication does not appear a day earlier to an editor who is travelling.
• Persian-aware search. The index covers headline, standfirst and body, normalising the two forms of ye and kaf, the digits and the zero-width non-joiner — so two spellings of the same word find the same story.
• Real RTL. The layout is built on CSS logical properties, not fixed left and right; Latin runs inside Persian text are isolated islands and URLs do not scramble.
• Persian slugs. A story's URL is built from its Persian headline without transliteration, and changing it creates a redirect automatically — a link published elsewhere does not break.

The stack

Three applications, one contract: a single API feeds both the admin panel and the public site, with types generated from that contract so neither drifts ahead of the other.

• Back end — .NET 10 with Minimal API, CQRS, Entity Framework Core and PostgreSQL. 84 projects, 1,029 files, 34 migrations, 89,515 lines.
• Admin panel — React 19 with Vite, Tailwind, TanStack Query, Radix UI, the TipTap editor and Zod validation. 248 files, 36 routes, 33,465 lines.
• Public site — Next.js 15 with server rendering and JSON-LD structured data, so a story is read correctly by Google and by news surfaces. 79 files, 17 pages.
• Deployment— Docker Compose with automatic database migration and health checks.

Proven, not claimed

824 unit tests, 127 architecture tests, 13 integration tests, 8 end-to-end acceptance tests and 148 client-side tests. Three more things were in this product's definition of done: a capability whose infrastructure is not yet decided answers with a plain "not enabled yet" rather than a vague error; no counter ever shows an invented number; and colour contrast is verified by real WCAG computation, locked in tests, in both the light and the dark theme.