simple-memory-skill

The zero-dependency memory system for AI agents.

No API keys. No external services. No cloud dependencies. Just pure local storage with intelligent search.

Architecture

┌─────────────────────────────────────────────────┐
│          SIMPLE LOCAL MEMORY                    │
├─────────────────────────────────────────────────┤
│                                                 │
│  ┌─────────────┐  ┌─────────────┐             │
│  │   HOT RAM   │  │  COLD STORE │             │
│  │             │  │             │             │
│  │ SESSION-    │  │  Indexed    │             │
│  │ STATE.json  │  │  Memories   │             │
│  │             │  │  (JSON +    │             │
│  │ (active     │  │   Search)   │             │
│  │  context)   │  │             │             │
│  └─────────────┘  └─────────────┘             │
│         │                │                     │
│         └────────────────┼─────────────────┘   │
│                          ▼                      │
│                  ┌─────────────┐                │
│                  │ MEMORY.md   │ ← Human        │
│                  │ + daily/    │   readable     │
│                  └─────────────┘                │
│                                                 │
└─────────────────────────────────────────────────┘

The 3 Memory Layers

Layer 1: HOT RAM (SESSION-STATE.json)

Fast, active working memory

{
  "current_task": "...",
  "key_context": ["...", "..."],
  "pending_actions": ["...", "..."],
  "recent_decisions": ["..."],
  "last_updated": "2026-03-15T10:30:00Z"
}

Benefits:

• Fast JSON read/write
• Survives compaction
• Easy to parse programmatically

Layer 2: COLD STORE (Indexed Memories)

Persistent, searchable memory

memory-store --type preference --content "User prefers dark mode" --importance 0.9

memory-search "what did user say about CSS"

memory-list --limit 10

Storage: memories/ directory with indexed JSON files

Layer 3: CURATED ARCHIVE (MEMORY.md + daily/)

Human-readable long-term memory

workspace/
├── MEMORY.md              # Curated insights
├── SESSION-STATE.json     # Active context
└── memories/
    ├── 2026-03-15.json    # Daily memory dump
    ├── preferences.json   # User preferences
    ├── decisions.json     # Key decisions
    └── lessons.json       # Lessons learned

Quick Setup

Step 1: Initialize

npm install -g simple-local-memory
cd your-project
memory-init

This creates:

• SESSION-STATE.json - Active working memory
• MEMORY.md - Long-term curated memory
• memories/ - Directory for memory storage

Step 2: Use with Your AI Agent

For Claude Code:

When I give you important information:
1. Write it to SESSION-STATE.json FIRST
2. Then store it using memory-store
3. Then respond to me

When starting a conversation:
1. Read SESSION-STATE.json
2. Search relevant memories with memory-search
3. Check MEMORY.md for context

For ChatGPT/Cursor: Add to your system prompt:

You have access to local memory tools:
- memory-store: Save important information
- memory-search: Find relevant past context
- Read SESSION-STATE.json before responding
- Update SESSION-STATE.json when user shares preferences

Memory CLI Commands

memory-init

memory-store --type preference --content "User loves TypeScript" --importance 0.9

memory-search "TypeScript preferences"

memory-list --limit 10 --type preference

memory-stats

memory-export --format json --output backup.json

memory-import --file backup.json

WAL Protocol (Write-Ahead Logging)

CRITICAL: Write to memory BEFORE responding

Why? If response crashes before saving, context is lost.

Memory Storage Format

memories/YYYY-MM-DD.json

{
  "date": "2026-03-15",
  "memories": [
    {
      "id": "uuid",
      "type": "preference|decision|fact|lesson",
      "content": "User prefers dark mode",
      "importance": 0.9,
      "tags": ["ui", "preferences"],
      "timestamp": "2026-03-15T10:30:00Z",
      "context": "Discussed during UI setup"
    }
  ]
}

memories/preferences.json

{
  "preferences": [
    {
      "key": "css_framework",
      "value": "Tailwind",
      "set_at": "2026-03-15T10:30:00Z",
      "reason": "User prefers over vanilla CSS"
    }
  ]
}

memories/decisions.json

{
  "decisions": [
    {
      "id": "uuid",
      "title": "Use React for frontend",
      "reason": "User requested component-based architecture",
      "made_at": "2026-03-15T10:30:00Z",
      "status": "active"
    }
  ]
}

Search Algorithm

TF-IDF based local search:

1. Tokenize query and memories
2. Calculate term frequency
3. Rank by relevance + importance + recency
4. Return top N results

// Example search logic
function searchMemories(query, limit = 5) {
  const queryTokens = tokenize(query);
  const allMemories = loadAllMemories();

  const scored = allMemories.map(memory => {
    const score = calculateTFIDF(queryTokens, memory.content);
    const recencyBoost = calculateRecencyBoost(memory.timestamp);
    const importanceBoost = memory.importance || 0.5;

    return {
      ...memory,
      totalScore: score + recencyBoost + importanceBoost
    };
  });

  return scored
    .sort((a, b) => b.totalScore - a.totalScore)
    .slice(0, limit);
}

Example Workflow

User: "Let's use Tailwind for this project, not vanilla CSS"

Agent process:
1. Update SESSION-STATE.json with decision
2. Execute: memory-store --type decision --content "Use Tailwind, not vanilla CSS" --importance 0.9
3. Execute: memory-store --type preference --content "User prefers Tailwind over vanilla CSS" --importance 0.95
4. THEN respond: "Got it — Tailwind it is. I've saved this preference."

Memory Categories

Maintenance

Daily

memory-stats

memory-list --date today

Weekly

memory-archive --days 7

memory-deduplicate

Monthly

memory-export --format json --output monthly-backup.json

memory-cleanup --days 30

Memory Hygiene Tips

1. Be specific - "User likes dark mode" > "User has preference"
2. Add context - Why was this decision made?
3. Use importance - Not everything is 1.0
4. Tag properly - Helps with retrieval
5. Archive regularly - Keep SESSION-STATE.json small

Troubleshooting

Search returns nothing: → Check memories/ directory exists → Verify JSON files are valid → Try broader search terms

SESSION-STATE.json grows too large: → Move old items to memory-store → Archive completed tasks → Keep only active context

Memories not being saved: → Check file permissions → Verify disk space → Check JSON syntax

Advanced Features

Memory Relationships

{
  "id": "uuid",
  "content": "Use React for frontend",
  "related_to": ["uuid-of-other-memory"],
  "followed_by": ["uuid-of-decision"]
}

Confidence Scores

{
  "confidence": 0.95,
  "source": "explicit_user_statement",
  "verified_count": 3
}

Expiry Dates

{
  "expires_at": "2026-04-15T00:00:00Z",
  "auto_archive": true
}

Comparison with elite-longterm-memory

Migration from elite-longterm-memory

memory-export > elite-backup.json

node convert-elite-to-simple.js elite-backup.json > simple-backup.json

memory-import --file simple-backup.json

Future Enhancements (Optional)

• Add local embedding models (Transformers.js)
• Add compression for old memories
• Add encryption for sensitive data
• Add sync via GitHub Gist
• Add web UI for memory management

No API keys. No cloud. No tracking. Just pure local memory.

Perfect for:

• Privacy-conscious users
• Offline development
• Learning how memory systems work
• Building custom AI agents
• Projects with strict data policies

依赖说明

运行环境

• Agent平台: 支持SKILL.md的任意AI Agent(Claude Code / Cursor / Codex / Gemini CLI等)
• 操作系统: Windows / macOS / Linux

第三方依赖

API Key 配置

• 本Skill基于Markdown指令,无需额外API Key(除内容中明确标注的外部API)

可用性分类

• 分类: MD+EXEC(纯Markdown指令,部分功能需要exec命令行执行能力)
• 说明: 基于Markdown的AI Skill,通过自然语言指令驱动Agent执行任务