AIIA Dashboard

Open-source US-stock AI-sector intelligence — real-time monitoring, factor analysis, and LLM daily briefings.

AIIA · 美股 AI 板块情报监测系统 — a real-time intelligence platform tracking 10 sub-sectors and 84 tickers (76 stocks + 8 ETFs) across the US AI complex: quotes, volume, approximate money-flow, and news, surfaced as a sector thermometer, rotation table, heatmap, K-line charts, and an LLM-generated Chinese daily briefing.

🔗 Open the live AIIA Dashboard → View source on GitHub →
Open source on GitHub · live deployment self-hosted at /AIIA-Dashboard/

Architecture

A single Cloudflare Worker (5 cron pipelines + a Hono API + static hosting) backed by a D1 database, with a React SPA front end — running entirely on free data sources and free-tier infrastructure, 7×24.

Twelve Data (daily)   ┐
Finnhub (quotes/news) │→  Worker (cron × 5)  →  D1  →  Hono API  →  React SPA (7 views)
RSS / geo intel feeds │
OpenAI-compatible LLM ┘   (Chinese daily brief + research)

Views

Route What it shows
/ Overview 5-factor sector thermometer, market-cap-weighted heatmap, rotation table, today’s AI brief, Top Movers, earnings calendar
/map Global map worldmonitor-style intel map: HQ bubbles, AI-infra nodes, geo news events, supply-chain arcs
/sector/:id · /stock/:ticker Sector equal-weight index & constituents; daily K-line + indicators + 15s quasi-real-time quotes
/rank Leaderboard Multi-factor ranking across all 84 tickers with 30-day sparklines
/news Intel feed LLM structured event cards (signal / market reaction / historical baseline) + event dictionary (CAR stats)
/team Research team ai-berkshire-style multi-agent pipeline running four investing-master methodologies

Highlights

  • Intel factorization: each headline is parsed into a structured event (10-type taxonomy × polarity × strength × novelty × supply-chain transmission); EOD computes ticker-level factors and abnormal returns (β-stripped OLS residual vs QQQ + leave-one-out peers), with multi-horizon rank-IC.
  • Sector thermometer (0–100): momentum, breadth, volume, money-flow, and news-sentiment sub-scores, z-scored over 252 trading days and logistic-mapped.
  • ~330 automated tests across Worker and web (metric lock-values, pipeline contracts, D1, budget guardrails).

Stack: Cloudflare Workers · D1 · Hono · React · TypeScript · lightweight-charts · Vite

Open source and continuously running — use the live deployment above or explore the code on GitHub.