Skip to main content

​
Data plane overview

QuantDesk’s market data stack is split into three layers:
  1. Collectors ingest raw market, protocol, and article events.
  2. The data plane normalizes those events into a shared envelope and writes authoritative hub tables.
  3. Backend APIs and websockets expose the hub to frontend surfaces, backtesting workflows, and external clients.

​
Core principles

  • Registry first: market identity comes from the market registry, not ticker guesses.
  • One envelope: Redis producers publish the same event shape before any downstream fanout.
  • Postgres hub: hub_market_stats, hub_candles, hub_trades, hub_news, hub_reference_perps, and hub_collector_status are the backend source of truth.
  • REST bootstrap, websocket freshness: clients load the latest snapshot over HTTP, then stay current with Socket.IO updates such as market_stats_update.

​
Services

Article news ownership lives in the Sentiment Feed Aggregator only; the Market Data Ingestion Pipeline no longer acts as the first-class article collector in the standard local development configuration. For local end-to-end verification, the standard Local development configuration is expected to bring up the Market Data Ingestion Pipeline, Sentiment Feed Aggregator, and Social Stream Ingestion Matrix alongside the gateway/frontend services.

​
Public read paths

​
Why this exists

Epic 15 removes market-facing mock data from production surfaces and gives every consumer the same canonical input set. That keeps:
  • the Lite and Pro terminals aligned,
  • backtest dataset exports reproducible,
  • TradFi session handling explicit,
  • and collector health observable.