Key requirements: Geometry & maps: calculate building centroids, handle coordinate rotation/scaling (fitting a floor plan to a satellite footprint), and manage true vs. magnetic north.ai integration: use multimodal llms (eg. gemini 1.5 pro/gpt-4.o) to extract (x,y) coordinates from floor plan images into structured json.logic engine: build a deterministic scoring engine in code (python/node) that calculates results based on 3 admin-managed csv files (rules, weights, categories).infrastructure: architect for zero-cost/free-tier (e.g., supabase, vercel). Ownership: strictly use a dedicated project email for all third-party service registrations for a clean handoff. The Goal: A tool where a user aligns a floor plan over a map, ai identifies the rooms, and a code-based engine scores the orientation based on my custom csv logic.