// every measurable property of your files maps to a fixed visual element. same file → same building, every time. //
| LOG FEATURE | VISUAL ELEMENT |
|---|---|
| timestamp (earliest event) | ──▶ X position along the street (west → east = older → newer) |
| file kind (log/csv/json/text) | ──▶ which street row (Z axis) the building sits on |
| line count | ──▶ rooftop antenna height |
| avg line length | ──▶ windows per story (facade columns) |
| byte size | ──▶ building footprint (width × depth) |
| entropy | ──▶ lit-window ratio (more entropy → more lights on) |
| token uniqueness | ──▶ window warmth (cool cyan → warm magenta) |
| null / empty rate | ──▶ forced-dark windows |
| numeric variance | ──▶ number of stories (building height) |
| error rate | ──▶ red rooftop beacons (count) + red pulse + red wire-edges if anomaly>70 |
| warning rate | ──▶ amber base glow around the building |
| line count (per street) | ──▶ vehicle count on that street |
| error rate (per street) | ──▶ red car appears when avg street error rate > 5% |
| inner anomalies (line-level) | ──▶ red trash cans around the building base |
| anomaly score (z-vector L2) | ──▶ building edge color + vertical pulse amplitude |
// selecting a building computes cosine similarity of its z-vector against every other file. top-5 shown with a one-line reason derived from the dominant feature dimension. //