Ordo Kino

Ordo Kino was a weekend MVP. The question was simple: what if I could browse my media collection from a Netflix-like interface, with an AI chat helping me find what I'm looking for?

The standalone version was built on Next.js 16 with Groq Llama 3.3 70B and Vercel AI SDK 6. The AI got 4 tools: nCore search (split by category: movies, series, music, games — Hungarian and international), torrent detail fetching with IMDB links, download triggering, and local media library browsing. The library scanner recursively walks directories and extracts year, quality (1080p/2160p), codec (x264/x265/AV1), source (BluRay/WEB-DL/AMZN/NF), and language from folder names. There's also an MCP server with StdioServerTransport, so any MCP client can connect to it.

It worked so well that it didn't stay standalone. I ported it into Ordo Centralis as a full film module on .NET 8: CQRS pattern with MediatR, EF Core, Angular frontend. Chat history now persists to the database, streaming runs on token-based auth (5-minute TTL, ConcurrentDictionary), with Range request support (HTTP 206). nCore credentials are encrypted in the database, and permissions run through project-level RBAC.

The MVP → production transition is a good example of how I validate ideas: build something fast, try it out, and if it works, integrate it properly into the system.