skills.rest
kaifashraff
kaifashraffCOMMUNITY
Explore all skills by this author →
GitHub

elite-longterm-memory-local

Manage private on-device memory for AI agents with LanceDB and JavaScript embeddings.

Updated Mar 18, 2026
Search Tags:#memory#ai-agent#lancedb#vector-search#long-term-memory#openclaw#local-embedding
One-click install
npx skills add https://github.com/kaifashraff/jarvis-research --skill elite-longterm-memory-local
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: elite-longterm-memory-local
Source: https://github.com/kaifashraff/jarvis-research/tree/main/skills/elite-longterm-memory-local
Command: npx skills add https://github.com/kaifashraff/jarvis-research --skill elite-longterm-memory-local

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Private, on-device memory management enables AI agents to remember context, preferences, decisions, and key events without sending sensitive data to external services. It provides a LanceDB-backed vector store and pure JavaScript embeddings to keep operations offline and private.

Core Features & Use Cases

  • Local long-term memory with hot RAM (SESSION-STATE.md), warm vector store, and cold archive for structured decisions.
  • Semantic recall via LanceDB vector search using pure-JS embeddings, enabling fast contextual recall.
  • Auto-recall before agent start, manual memory_store/memory_recall/memory_forget tooling, and daily logs for human-readable archival.

Quick Start

Initialize the workspace, store a memory with memory_store, then recall relevant memories with memory_recall.

Frequently Asked Questions about elite-longterm-memory-local

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

FAQPage Schema
How do I add private local memory to an AI agent without external APIs?▼

Local long-term memory for AI agents can be managed privately on-device using a LanceDB-backed vector store with pure JavaScript embeddings. This operates completely offline without external APIs.

How does vector search recall work for on-device agent memory?▼

Vector search recall for agent memory uses LanceDB to match pure-JS embeddings locally. This enables fast contextual retrieval across the warm vector store without external services.

What's the best way to structure long-term memory across different access speeds?▼

Long-term memory is structured across three tiers: hot RAM via SESSION-STATE.md, a warm LanceDB vector store for semantic recall, and a cold archive for structured decisions. This balances fast access and persistent storage.

Can I manually store and forget specific memories in a local AI agent database?▼

Yes, local AI agent memory can be manually managed using memory_store, memory_recall, and memory_forget tooling. These provide explicit control over storing, retrieving, and deleting agent context.

Does this local memory approach require external embedding services to run?▼

No, this local memory approach uses pure JavaScript embeddings instead of external embedding services. This ensures vector search operations remain completely offline and private within the local environment.

When should I use on-device memory management instead of cloud-based agent memory?▼

On-device memory management is ideal when handling sensitive data that cannot be sent to external services. It provides a private local database ensuring agent context and structured decisions remain completely offline.

Related Skills

Explore More Software Engineering Skills→
hiyenwonghiyenwong
2

chat-history-lancedb

Store and search chat histories locally with LanceDB vector embeddings and RAG retrieval.

Community
Advanced
zhcndoczhcndoc

session-logs

Analyze JSONL session logs with jq and ripgrep to locate past conversations.

Official
Intermediate
ruvnetruvnet
4.4k

AgentDB Memory Patterns

Automate AI agent memory management with AgentDB persistent storage and ReasoningBank integration.

Community
Advanced
dglijin-ossdglijin-oss
10

elite-longterm-memory-local

Store and search memories locally using LanceDB and Ollama embeddings.

Community
Advanced