System Architecture

Everything runs on Rust. No Python, no JIT experiments.

Stack

Component
What It Does

Dexstream

WebSocket feeds from DexScreener, 9 chains

Aegis

Security checks (blacklist, holders, coordination)

Seraphim

HTTP client for GoPlus, Honeypot, The Graph

Grimoire

SQLite database for signals, blacklist, cooldowns

Mitch

AI report generator (Claude API)

Data Flow

  1. Dexstream pulls token pairs via WebSocket

  2. Filters apply: $5K-$40K mcap, liquidity, spam checks

  3. Aegis runs security modules in parallel

  4. Passed signals go to Telegram

Security Stack

Aegis runs these checks on every signal:

  • Contract scanning (GoPlus, Honeypot.is, AI)

  • Blacklist lookup (50K+ wallets)

  • Holder concentration analysis

  • Leiden-based coordination detection

  • Paper hands / new wallet heuristics

Each chain has tuned thresholds (e.g., Solana 60%/65% block concentration, Ethereum 80%/90%).

Performance

  • ~40% faster than the old Python stack

  • 75% test coverage and climbing

  • Hot reload for thresholds — no downtime for config changes

  • Retry/backoff on all external APIs

Last updated