I need a compact proof-of-concept backend that can use small AI agents (Python is my preferred stack) to generate and maintain structured “world” data. The system only has to cover the essentials; it is not meant to evolve into a full game. Core features • JSON generation: create initial regions, cities, and natural resources. • Event engine: propose simple, logical world events. • Validation layer: inspect any update request and approve or reject it before the data store is touched. • Versioning: keep snapshots, expose a rollback endpoint, and make restores trivial. • REST API: clean endpoints for world creation, event suggestions, updates, snapshot retrieval, and rollback. Tech notes Python is the target language, but I am open to whichever lightweight web framework you feel keeps things simple. Storage can be SQLite during development and easily swapped for Postgres or Supabase later; please abstract that layer so the switch is painless. Deliverables 1. Running prototype (CLI and HTTP endpoints) 2. Clear folder structure and requirements.txt/pyproject.toml 3. README with local setup, run, and test instructions 4. Brief architecture write-up describing AI agent flow, validation logic, and snapshot strategy Timeline I’d like an initial version as soon as reasonably possible—feel free to outline milestones so we can move quickly yet cleanly.