jev-shopper

# Jev Shopper shop design Jev Shopper runs a live shop window: a few products stream price ticks and Jev (TypeSafe's System One model) reads the window each tick and calls the best buy — the product furthest below its own usual price — showing confidence and spending on paper when the signal is strong. The agent shapes `shopper.json` — the products, the price noise, the budget and the deal bar. The shop and its prices are made up. ## The loop 1. Update `shopper.json` (title, products with price/drift, vol, tickMs, and a `style` line that tells Jev a buying strategy). The viewer watches it and Jev adapts live — no restart, no second server. 2. `node "$JEV_DSH/toolchain/check.mjs"` verifies the workspace's `shopper.json` is valid. Run it before you call a shop done. 3. Watch the board. The top bar shows the share of deal calls Jev got right and what its buys saved against the usual price. Does Jev buy real sales, or wobbles (little saved), or nothing (dull)? That observation is the finding. ## Reading the shop The viewer shows each product as a price card with a live chart, a spotlight on Jev's "best buy", a probability bar per product, purchase stamps, the budget left and receipts. Every product has a hidden fair price (list price, slow drift, now and then a real sale). Jev reads that price plus noise. `vol` is the noise and the difficulty dial: at `0.01` Jev's deal calls are right almost every time and buys save about 16%; at `0.15` wobbles pass for deals,...