LIGHTPAPERLIGHTPAPERLIGHTPAPER

The Deterministic Audit • Version 1.0 (Alpha Kernel)

A technical overview of Yurei's architecture — from bare-metal Helius RPC ingestion to LLM-powered risk translation.

Deep Dive Documentation

YUREI AGENT

The Kernel

The Kernel • Real-Time Detection Layer

  • C11 + ASM Core
  • Helius RPC Streaming
  • <20ms Target Latency
Read Documentation
PROTOCOL NAVI

The Interface

The Interface • LLM Reasoning Layer

  • Semantic Translation
  • Vector Risk Analysis
  • Deterministic Scoring
Read Documentation

01The Latency of Trust

The fundamental premise of "Do Your Own Research" (DYOR) in the Solana ecosystem is broken. It fails not due to a lack of tools, but due to latency. By the time a user begins their "research," the outcome has often already been determined by actors with microsecond-level visibility.

Within 200-400ms of a token launch, sniper bots have already executed. YUREI exists to collapse this asymmetry.

02System Architecture

[SOLANA L1] │ ▼ +----------------------+ | YUREI AGENT KERNEL | <-- (Autonomous Engine) | [C11 / ASM / SQL] | * Runs Independently +----------------------+ * Ingests & Calculates │ │ (Push Risk Vectors) ▼ +----------------------+ +-----------------------+ | DATA LAYER | <--- | PROTOCOL NAVI | | [Postgres / Redis] | | [Semantic Logic] | +----------------------+ +-----------------------+ │ │ (Read Context) ▼ [USER DASHBOARD]

Fig 1. Simplified data flow from validator to user interface.

03Risk Vector Methodology

The Agent's risk engine uses weighted Vectors to determine the final safety score.

VectorTypeMax Penalty
V_ACode Risk (Auths)+40.0
V_BLiquidity RiskUNBOUNDED
V_CBehavioral (Bots)+25.0

Don't trust. Verify with the Kernel.

Agent DocumentationNAVI Documentation