A self-hosted market dashboard that needs zero API keys (A-shares, HK, US, gold, commodities, treasury yields)

The one-screen market cockpit with no API bills. Live demo: https://mrd.hermes.cc.cd - no keys, no login, works instantly. Source: github.com/theBigGavin/marketingdashboard.

Most market-data APIs charge $20-80/mo for real-time feeds. This dashboard pulls everything from free public endpoints (Tencent, Sina, Eastmoney, Wallstreetcn, CNBC, Binance...) through a zero-dependency Node proxy with in-memory TTL caching. It costs nothing to run.

What's on screen

• Global indices side by side - SSE / SZSE / Hang Seng / Dow / Nasdaq / S&P 500 / VIX, minute-level charts
• Commodities & crypto - NY/London gold, SHFE gold, LME copper, crude, BTC - live + intraday curves
• US Treasury monitor - 10Y/2Y yields, 2s10s spread, yield-curve history back to 2001
• Sector money-flow - minute-level cumulative net-inflow curves; click a sector and the stock ranking panel switches to that sector's constituents instantly
• Industry-chain panorama - semiconductors, AI compute, EVs, robotics, innovative drugs... upstream/midstream/downstream tickers linked to live quotes
• 7×24 news flash - auto-highlighted macro keywords
• Earnings window - disclosure calendar, profit forecasts, 12-quarter per-company trends

All charts are hand-written SVG (no charting library), which keeps the bundle small and the rendering fast even on a 2015 TV.

Deployment

docker build -t mrd . && docker run -p 3000:3000 mrd
# or just open the live demo

Runs on a single Node process - API + static frontend on one port. There's also a macOS app (thin Swift WKWebView shell) and an Android TV build with D-pad navigation.

What I learned

• Public endpoints are good enough for research-grade data - GBK-encoded Tencent quotes, Eastmoney's clist API for sector rankings, Sina for futures. Each needs its own decoding/retry/rate-limit handling.
• A tiny in-memory cache with LRU eviction beats any external cache - 1.5s-24h TTLs per endpoint, no Redis needed.
• Hand-written SVG charts are underrated - no 200KB charting dep, full control over zoom/overlay behavior, and they survive CSS-zoom panel overlays on TV.

Star it if you find it useful, and PRs are welcome.