core-brain-mula — Zero-Decay Neural Memory for OpenClaw

A neural memory system where neurons never die. Every decision, preference, and instruction persists forever with intelligent conflict resolution and automatic deduplication.

Overview

core-brain-mula installs and operates a zero-decay neural memory backend for OpenClaw agents. With decay_rate=0.0 and prune_threshold=0.0, no memory is ever pruned. Conflict resolution uses confidence scoring to auto-resolve high-certainty conflicts or ask the user when uncertain. SimHash deduplication prevents redundant entries, and a 4000-token context budget keeps recall fast. Every new decision or preference is written immediately.

Features

• Zero decay — decay_rate=0.0, prune_threshold=0.0, all neurons survive
• Conflict resolution — certainty ≥ 0.8 → auto-resolve + announce; < 0.8 → ask user; both valid → keep with context tags
• Auto-remember — every decision / preference / instruction written immediately
• SimHash dedup — no duplicate entries
• Token budget — max_context_tokens=4000 for fast recall

Usage / Quick Start

pip install neural-memory[server]
npm install -g @neuralmemory/openclaw-plugin
python3 scripts/setup.py
# restart gateway, then:
python3 scripts/import_memory.py

Trigger Keywords (OpenClaw)

install NeuralMemory, import MEMORY.md, conflict resolution, query memory, backup brain, restore brain

Related Skills

• context-budgeting — context window management
• memory-tiering — tiered storage layer

---

Part of the NachaFromMars OpenClaw skill ecosystem.