GeoTrade

A market-intelligence terminal for Indian equities that links real-time news to simulated price movement.

Real headlines are pulled from twelve Indian publisher RSS feeds, tagged with the NSE symbols each story should move, and fed into a local market simulator. Prices are simulated; the news is real.

Research project. Prices are synthetic and the positioning notes are generated, not researched. Nothing here is investment advice.

---

What it does

News → symbols → impact. A headline like "India-EU FTA: tariff quota for 1 lakh cars" is tagged to NIFTYAUTO, MARUTI and TATAMOTORS with a signed expected move for each. The simulator applies those impacts gradually, so the tape visibly reacts to the story.

Desk View. Each story carries a positioning note on the two or three stocks it most plausibly moves — action, conviction, CMP/target/stop levels, horizon, and a one-line reason drawn from the sector's transmission channel. Derived from the tagger's impact map rather than rolled at random, so a story that pushes a stock up never produces a sell call on it.

Single source of truth. One simulator owns all market state. Every panel, ticker, sparkline and chart reads the same snapshot, so a symbol never shows two different prices.

Panels

Markets · News → Markets · India Market Map · NSE Gainers & Losers · NSE Sector Heatmap · FII/DII Flows · MCX Commodities · Energy Complex · Forex & Rupee · G-Sec & Rates · Fear & Greed · Financial Stress · Market Breadth · India Macro

Stack

Running locally

npm install
npm run dev

Vite serves the frontend on :5273 and proxies /api to the news API on :5274.

Production build (one process serves both):

npm run build && npm start

Architecture

server/
  feeds.ts     Indian publisher RSS sources
  tagger.ts    headline → NSE symbols + sentiment → impact %
  poller.ts    poll every 5 min, dedupe across publishers, cache
  index.ts     REST API + static hosting in production

src/
  simulation/  market engine — the single source of truth
  data/        instrument universe, news adapter, types
  lib/         formatting (₹, lakh/crore), desk-view generator
  components/  panels, charts, terminal shell

The frontend polls /api/news every 90 seconds. If the API is unreachable it falls back to an offline generator of plausible Indian headlines, so the demo never depends on the network.

API

Design notes

The UI deliberately follows a dense terminal idiom — monospace throughout, 12px base, dark ground, semantic green/red reserved for direction. Rupee values use Indian lakh/crore grouping (₹1,43,174, not ₹143,174), and the header tracks the actual NSE session state.

Documentation

docs/GeoTrade-Technical-Design-Report.docx — a 50-page technical design report covering the intended production architecture: entity resolution, sentiment vs. surprise modelling, abnormal-return labelling, the model ladder, and walk-forward validation. docs/geotrade-signal-architecture.html is a shorter web version.

Known limitations

• Prices are simulated, not real market data.
• The tagger is a rule engine, not a trained model. It mislabels edge cases.
• Publisher feeds bot-filter datacentre IPs; feeds that work locally may return 403 from a cloud host.
• English-language sources only.

Licence

Not yet specified.