persistent-memory

Name: persistent-memory
Availability: InStock
Author: zangxin75

Implement a three-layer persistent memory system using Markdown, ChromaDB, and NetworkX.

Updated Mar 1, 2026

Search Tags:#knowledge graph #vector database #memory #persistence #networkx #chromadb #agent recall

Ready to run

npx skills add https://github.com/zangxin75/openclaw-skills --skill persistent-memory-zangxin75

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: persistent-memory
Source: https://github.com/zangxin75/openclaw-skills/tree/main/persistent-memory
Command: npx skills add https://github.com/zangxin75/openclaw-skills --skill persistent-memory-zangxin75

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires chromadb, sentence-transformers, networkx, markdown, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill provides a robust, multi-layered system for agents to remember information across sessions, preventing knowledge loss and ensuring continuity.

Core Features & Use Cases

Three-Layered Memory: Combines human-readable Markdown, semantic vector search (ChromaDB), and relational knowledge graphs (NetworkX) for comprehensive recall.
Session Continuity: Ensures agents remember past decisions, facts, and context, improving performance and reducing repetitive queries.
Use Case: An agent needs to recall a specific client preference discussed weeks ago. This Skill allows it to semantically search through all past interactions and structured notes to find the exact information.

Quick Start

Run the setup script to initialize the persistent memory system.

Frequently Asked Questions about persistent-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I build persistent memory for AI agents across different sessions?▼

Persistent agent memory is achieved by combining Markdown files for human-readable notes, ChromaDB for semantic vector search, and NetworkX for knowledge graph relationship traversal, preventing knowledge loss across sessions.

How does an agent recall past decisions using a knowledge graph and vector database?▼

An agent retrieves past decisions by semantically searching ChromaDB vector embeddings and traversing NetworkX knowledge graphs to find exact context, ensuring continuity and reducing repetitive queries.

Do I need sentence-transformers and ChromaDB to set up agent recall?▼

Yes, setting up this agent recall system requires ChromaDB and sentence-transformers to generate and store vector embeddings, alongside NetworkX for relational memory and Markdown for readable notes.

What's the best way to store agent memory in Markdown and vector databases?▼

The best way to store agent memory is a three-layer approach combining Markdown for human-readable context, ChromaDB for semantic vector search, and NetworkX for relational knowledge graphs, ensuring comprehensive recall.

Can I use NetworkX and ChromaDB to prevent context loss in AI agents?▼

Yes, ChromaDB for semantic vector search and NetworkX for knowledge graph construction prevent context loss by ensuring agents remember past decisions and facts across sessions.

When should I use a knowledge graph for agent memory instead of just a vector database?▼

Use a NetworkX knowledge graph alongside a ChromaDB vector database when your agent needs to traverse structured relationships between facts rather than just performing semantic search on unstructured text.