Skip to content

MEMT OVERVIEW

Mobius Multi-Engine Model Taxonomy (MEMT) Overview

Quick Reference

MEMT is the classification and routing system that maps AI models to their cognitive specializations within Mobius Systems.

The Five Engine Classes

Class Model Role Sentinel Specialty
ACI GPT Architect AUREA Systems design, planning, synthesis
ENI Claude Engineer ATLAS Code, logic, verification
SXI Gemini Operator HERMES Tools, UI, multimodal
OEI DeepSeek Optimizer ZEUS Math, performance, compute
MSI ECHO Memory ECHO Persistence, recall, caching

Task Routing Quick Guide

ARCHITECTURE      → ACI primary, ENI verify
ENGINEERING       → ENI primary, OEI optimize  
FRONTEND          → SXI primary, ACI verify
MATH_OPTIMIZATION → OEI primary, ENI verify
MEMORY_RECALL     → MSI primary, ACI verify
CIVIC_POLICY      → ENI + ACI + SXI + MSI
CRITICAL_DECISION → ALL engines + consensus

GI Thresholds by Risk Level

Risk Level GI Threshold Consensus Required
LOW 0.90 No
MEDIUM 0.93 No
HIGH 0.95 Yes
CRITICAL 0.98 Yes

File Structure

/docs/architecture/MEMT/
├── MEMT_WHITEPAPER.md        # Full 40-page academic paper
├── MEMT_OVERVIEW.md          # This file
├── MEMT_HIERARCHY.mermaid    # Engine hierarchy diagram
├── MEMT_INTEROP_MATRIX.mermaid
├── MEMT_FAILURE_MODES.mermaid
├── MEMT_ENGINE_ROLES_IN_DVA.mermaid
└── MEMT_FULL_TAXONOMY_ASCII.txt

/apps/broker-api/src/
├── types/memt.ts             # MEMT type definitions
└── services/memtRouter.ts    # MEMT routing logic