Pump-Short — Live Trade Monitor

A dependency-free, single-page dashboard over the Hyperliquid pump-short paper trader. Every setup the bot sees — taken, no-capacity (qualified but unfunded), and skipped (filter-rejected ghost) — with entry/SL/TP, an interactive what-if chart, live mark-to-market, and a taken-vs-ghost filter-edge verdict.

No build step, no libraries. Pure HTML/CSS/canvas — works offline and behind strict CDNs.

Files

File	Role
index.html	Layout + modal markup
styles.css	Dark trading-terminal theme
chart.js	Candlestick rendering, candle fetch (cached + rate-limited), draggable-level Interactive, short what-if simulator
app.js	Loads trades.json, renders KPIs / funnel / grid / open positions, wires the modal
trades.json	Committed data snapshot — produced by the exporter below

Refreshing the data

trades.json is a flattened snapshot of the vault's paper-trade-log.jsonl (the bot's source of truth). Open-position mark-to-market is always live (polled from the Hyperliquid API in the browser); the closed-trade history is only as fresh as the last export. To refresh it:

python scripts/export_dashboard_data.py        # auto-locates the vault log
python scripts/export_dashboard_data.py --log <path>   # explicit log

The portfolio / win-rate / PnL math mirrors lib/portfolio.py so the dashboard numbers match the Telegram alerts exactly.

Keeping the deployed dashboard live (run on the bot's box)

The dashboard only matches the Telegram alerts if it exports from the same vault log the live bot writes. So run the refresh on the machine running paper_trader.py — the exporter resolves VAULT_PATH from .env, the same path the bot writes to:

.\refresh_dashboard.ps1          # export trades.json, commit + push if it changed
.\refresh_dashboard.ps1 -NoPush  # export only (local preview)

Schedule it (every 15 min shown — tune it; each change is one commit). Push the branch your Vercel project deploys, and the dashboard mirrors the live bot:

schtasks /Create /SC MINUTE /MO 15 /TN "hl-dashboard-refresh" /F `
  /TR "powershell -NoProfile -ExecutionPolicy Bypass -File C:\path\to\repo\refresh_dashboard.ps1"

Requires git push auth on that box (the operator's GitHub access). If the bot moves machines, the refresh task moves with it — wherever the bot writes the log is where the dashboard must export from.

Run locally (one command)

From the repo root, double-click run-dashboard.bat (Windows) or run:

python run_dashboard.py

It refreshes trades.json from a local vault if one is present (otherwise it serves the committed snapshot), starts a local web server, and opens the dashboard in your browser. Requires Python 3.7+ — standard library only.

A web server is required: opening index.html directly via file:// won't work, because the browser blocks fetch("trades.json") from the filesystem.

Manual equivalent:

cd dashboard && python -m http.server 8765   # then open http://localhost:8765

Hand-off to a teammate (view-only, no vault)

Zip the dashboard/ folder plus run_dashboard.py and run-dashboard.bat, send it over, and they double-click the .bat. They'll see the snapshot you exported plus live open-position prices (pulled from the Hyperliquid API in their browser). Re-send a fresh zip to update the history.

Deploy (Vercel)

vercel.json at the repo root points Vercel at this dashboard/ directory as a static site (no build). trades.json is served no-store so a redeploy always surfaces the latest snapshot. Push a refreshed trades.json to redeploy the history.