memory-systems

Design multi-layer memory architectures with vector stores and knowledge graphs.

Updated Dec 5, 2025

Search Tags:#agent #retrieval #memory #persistence #knowledge #temporal-knowledge-graph

Ready to run

npx skills add https://github.com/mhismail3/dotfiles --skill memory-systems

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: memory-systems
Source: https://github.com/mhismail3/dotfiles/tree/main/claude/skills/memory-systems
Command: npx skills add https://github.com/mhismail3/dotfiles --skill memory-systems

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Memory systems enable agents to persist state across sessions, maintain entity consistency, and reason over accumulated knowledge.

Core Features & Use Cases

Layered memory architectures (working, short-term, long-term, and permanent)
Knowledge graphs and temporal knowledge graphs for reasoning across facts
Efficient indexing by entity and time to improve retrieval

Quick Start

Outline a memory stack with a working memory, a long-term store, and a simple temporal fact graph for a sample domain.

Frequently Asked Questions about memory-systems

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

FAQPage Schema

How do I design a memory architecture for an agent that persists across sessions?▼

A memory architecture layers working memory, short-term memory, long-term memory, and permanent storage to maintain agent state across sessions. This Skill guides you through building these layers, integrating vector stores and knowledge graphs, and implementing entity and time-based indexing so your agent retrieves relevant facts efficiently.

What's the best way to maintain entity consistency when an agent handles multiple conversations?▼

Entity consistency requires a temporal knowledge graph that tracks facts about entities over time. This Skill shows how to structure multi-layer memory to resolve the same entity across conversations, handle temporal validity, and update entity state as new information arrives.

Can I use a knowledge graph to reason over accumulated agent knowledge?▼

Yes. A knowledge graph organizes facts so agents can reason across relationships and infer new conclusions. This Skill covers temporal knowledge graph design for facts with time validity, entity-based indexing for fast retrieval, and integration with vector stores for semantic search.

How do I index agent memory by entity and time for efficient retrieval?▼

Efficient retrieval requires dual indexing: organize facts by entity identity and temporal validity. This Skill demonstrates scalable indexing strategies within a layered memory stack so agents retrieve relevant historical context without full-scan overhead.

What's the difference between working memory and long-term memory in agent systems?▼

Working memory holds immediate context for the current task; long-term memory stores facts for later retrieval across sessions. This Skill shows how to separate these layers, when to promote facts from working to long-term storage, and how to integrate permanent storage for historical reference.

Do I need a vector store to build a memory system for agents?▼

A vector store enables semantic search across agent memory, but it's optional depending on your retrieval needs. This Skill covers integration patterns with vector stores and knowledge graphs so you can choose architectures suited to your agent's reasoning requirements.