jevbooks

← All projects

jev-map-sandbox

jumboly/jev-map-sandbox

Testing Jev's typed judgments on map layer importance

What it is

A browser-only sandbox experimenting with Jev on maps. Experiment 01 classifies the importance of 24 groups from the 123-layer 地理院地図Vector style catalog based on spoken instructions, then reflects the result on a MapLibre GL JS map.

How it uses Jev

The utterance becomes the state, and each of the 24 layer groups is sent as one score question in a single parallel request. Scores from 0.0–4.0 are mapped to four levels (強調 / 通常 / 淡色 / 非表示) and applied to the map; each judgment is shown with its confidence and can be manually overridden.

Primitives:score

Technique worth stealing

Send one score question per layer group in a single parallel request, then bucket the 0.0–4.0 scores into four display levels.

Try it

npm install && npm run dev, open http://localhost:5173/, enter a Vercel AI Gateway API key (vck_…).

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricyes0.94
  • Shows a System One patternno0.19
  • Handles uncertaintyno0.04
  • Measuredno0.05
  • Runnableunclear0.44
  • Worth recommendingno0.34
  • Model replicano0.03
  • Problem scopescore on a 0–2 scale0.06
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.