Nity Memory — Overview

The memory subsystem provides persistent context across sessions. Located at extensions/nity-memory/.

Components

Session Brain Episode Recording Ralph Universal SimpleMem Bridge

Session Brain — Loads/persists brain.md context from .voyager/ directory
Episode Recording — Records task outcomes with task, approach, outcome, adapter, timestamp
Ralph Universal — Tracks adapter effectiveness per task category, generates recommendations
SimpleMem Bridge — Python CLI bridge for semantic memory operations
Memory Recall — Similarity-based episode search

Data Flow

session_start
  → load brain (.voyager/brain.md)
  → recall episodes (similarity search)
  → inject context into system prompt
  → task execute
  → record episode (task, approach, outcome, adapter, timestamp)
  → update adapter effectiveness (ralph-universal)
  → persist brain

Memory is identity. Every episode recorded shapes how Nity approaches future tasks. The brain file is the single source of truth for session-level context.

Architecture

The memory subsystem is designed as a layered stack:

Layer	Component	Responsibility
Persistence	SimpleMem Bridge	Python CLI for storage operations
Context	Session Brain	Session-scoped context management
Learning	Ralph Universal	Cross-adapter effectiveness tracking
Recording	Episode Recorder	Task outcome capture and retrieval
Recall	Memory Recall	Similarity-based search across episodes

Each layer is independent and can be tested in isolation. The coordinator.ts orchestrates the flow between layers during session lifecycle events.

Bridge SDK Session Brain