NACHTGRENZE: production system for an animated series

A serialized project for YouTube needs more than ideas. It needs a system that records what counts as canon, what has actually been produced, and how an episode is to be executed. This repository is not the film. It is the factory behind it.

What was done

• Canon tree in nine layers: story, characters, world, visual rules, production rules, registries
• Execution package for episode 1: 16 scenes, 170 shots, each with dependencies and approval state
• Asset lifecycle DRAFT → reviewed → APPROVED → LOCKED; automation never promotes on its own
• Validation run before every commit: locked references, duplicate IDs, every shot dependency, manifest and registry parity
• Studio interface on Next.js 16 and React 19, split into domain, storage and a provider SDK
• Trend scanner in Python: market research on YouTube formats, evidence, never canon
• Rules for AI collaborators: reading order, boundaries, what an agent must not touch

Result

• 175 commits, 271 files, about 46,000 lines of rules and documentation
• Canon and production state are separate authorities. Contradictions surface at validation, not in the edit
• The status report is generated from the registries, never maintained by hand

Stack: Next.js 16, React 19, TypeScript, Python, Node.js, JSON Schema, pnpm workspaces