Data Infrastructure & Indexing

Equinox AI relies on a high-throughput indexing framework to unify and synchronise diverse data sources, both on-chain and off-chain. Each activity—from a swap on Uniswap to a Telegram message—is indexed with temporal precision and behavioural context.

---

Core Infrastructure Design

[Node Layer] → [Scraper Layer] → [Index Layer] → [Relational + Graph DBs] → [Serve Layer]

---

Indexing Strategy

• On-Chain Indexing: Real-time data pulled via WebSocket, JSON-RPC, and archive nodes. Tracked: token transfers, contract calls, bridge events, mixer interactions.

• Off-Chain Indexing: Rate-limited scrapers collect Telegram chat logs, pinned messages, X post metadata, usernames, and profile history. Snapshots stored at regular intervals.

• Timestamp Normalisation: All events are time-anchored using UTC+0 standard and stored in event trees for sequence querying.

---

Storage Format

{
  "wallet": "0xabc...def",
  "event": "swap",
  "platform": "uniswap",
  "token_in": "ETH",
  "token_out": "USDC",
  "amount": "3.2",
  "timestamp": "2025-06-15T16:45:23Z",
  "linked_social": "@example"
}

---

Database Schema Types

Type	Purpose
SQL Tables	Transaction indexing, swap history
Graph DBs	Wallet ↔ Social ↔ Token mapping
Logs (S3)	Scraper archives, flagged activity logs
KV Stores	Realtime aliases, recent calls cache

---

Performance Benchmarks (Internal)

• Indexed over 1.2B on-chain events across 4 EVM chains.

• Normalised >12M Telegram messages.

• 400ms average query response time on wallet timeline lookups.

• 100K+ behavioural fingerprints stored in graph structure.

---

Our indexing model allows queries like:

SELECT * FROM wallet_activity
WHERE wallet = '0x123...def'
AND timestamp BETWEEN '2025-06-01' AND '2025-06-07';

Or:

{
  "query": "Find all wallets who joined same 3 Telegram groups as @example and made >$1k swaps after X post"
}

Equinox AI turns raw behavioural data into structured knowledge—pipelined, queryable, and scalable for DeFi, security, and market intelligence use cases.