Meta-Memory

Enhanced meta-memory skill for OpenClaw - Local memory management with vector search, predictive wakeup, multi-agent sharing.

Features

• Local Storage: All data stored in C:\Users\Administrator\.meta-memory
• Vector Search: Using Ollama with local embedding model
• Three-Layer Memory: Short-term / Mid-term / Long-term retrieval
• Auto Inference: Automatic context analysis without trigger words
• Predictive Wakeup: Intelligent memory activation
• Multi-Agent Sharing: Share memories between agents

Storage Structure

C:\Users\Administrator\.meta-memory\
├── index\              # Vector index (builtin memories)
├── backups\           # Auto backups
├── memory.db          # SQLite database
└── vector_db\        # ChromaDB (optional)

Usage

from meta_memory_enhancer import deep_recall, auto_infer, search

# Deep recall (with timeout protection)
result = deep_recall("what did I do recently")

# Auto inference
result = auto_infer("from meta-memory get my preferences")

# Basic search
results = search("user preferences")

Requirements

• Python 3.13+
• Ollama with locusai/all-minilm-l6-v2 model

Installation

# Install Ollama model
ollama pull locusai/all-minilm-l6-v2