Skip to main content

​
System Overview

Noise is built on a hybrid architecture: off-chain order matching for speed, onchain settlement for verifiability. The system has three main layers:
  • API — handles user requests, authentication, and real-time data delivery via REST and WebSocket
  • Matching Engine — manages the order book, matches orders, and calculates mark prices and funding rates
  • Blockchain Settlement — smart contracts on Base that hold collateral and settle matched trades onchain

​
Order Lifecycle

When you submit an order, it moves through a defined set of states:
StatusDescription
NewOrder submitted and written to the database
PendingMatching Engine has picked up the order
AcceptedOrder is live on the order book (limit orders) or being matched (market orders)
FilledOrder matched and settled onchain
CancelledOrder cancelled by user or system
RejectedMatching Engine rejected the order (insufficient margin, invalid parameters, etc.)

​
Wallets & Custody

Noise is non-custodial. When you sign up, an embedded wallet is created for you through Privy. Your collateral is held in smart contracts on Base — never by Noise. All deposits and withdrawals are onchain and verifiable.
You can connect an external wallet instead of using the embedded wallet if you prefer.
Order signing uses EIP-712 typed data, executed through the Privy server SDK on behalf of your embedded wallet.

​
Onchain Settlement

All collateral is held in smart contracts on Base. When a trade is matched off-chain, settlement happens onchain through two contracts:
  • SettlementEngine — validates signatures from both sides of a trade and orchestrates atomic settlement
  • PositionManager — manages isolated margin positions, enforces collateral requirements, and executes position updates

​
Oracle Infrastructure

The relevance engine runs as a separate service outside the main platform. It queries configured data sources on a regular tick interval, computes the Relevance Index for each market, and publishes the results. The Matching Engine consumes the Relevance Index, normalizes it into the Attention Factor, and combines it with trading volume data to produce the oracle price. The oracle price then feeds into funding rate calculations. See How It Works for detail on each of these components.
The oracle price is not the Relevance Index. It is a derived value that blends relevance data with trading activity. See Relevance — From Relevance to Oracle Price for the full breakdown.