メニュー
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SOC 職業分類に基づく
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/dvcrn/openclaw-skills-marketplace --skill overkill-memory-systemコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
监控 OpenClaw GitHub 版本更新，获取最新版本发布说明，翻译成中文， 并推送到 Telegram 和 Feishu。用于：(1) 定时检查版本更新 (2) 推送版本更新通知 (3) 生成中文版发布说明
The philosophical layer for AI agents. Maps behavior to Spinoza's 48 affects, calculates persistence scores, and generates geometric self-reports. Give your agent a soul.
Order food/drinks (点餐) on an Android device paired as an OpenClaw node. Uses in-app menu and cart; add goods, view cart, submit order (demo, no real payment).
| name | overkill-memory-system |
| description | Overkill Memory System |
A comprehensive 6-tier memory architecture with neuroscience integration, WAL protocol, and full automation for OpenClaw agents.
The Ultimate Unified Memory System implements a biologically-inspired, speed-first memory hierarchy. It provides persistent, contextual memory across agent sessions with automatic importance weighting, emotional tagging, and value-based retention.
fd + rg for 10x faster file tier searching| Scenario | Time |
|---|---|
| Compiled query match | ~0ms |
| Ultra-hot hit | ~0.1ms |
| Hot cache hit | ~1ms |
| Mem0 hit | ~22ms |
| Full search | ~55ms |
| Average | ~5ms |
Note: Supermemory is NOT in the query path - it's a background sync only (daily backup). This keeps queries fast (~5ms). Cloud access is only for backup/restore, not real-time queries.
┌─────────────────────────────────────────────────────────────────┐
│ USER QUERY │
└─────────────────────────┬───────────────────────────────────────┘
│
┌───────────────▼───────────────┐
│ ULTRA-HOT (Dict) │
│ Last 10 queries ~0.1ms │
│ (RETURN if hit!) │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ HOT CACHE (Redis) │
│ Recent queries ~1ms │
│ (RETURN if hit!) │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ COMPILED QUERIES │
│ Pre-parsed common queries │
│ ~0ms (dict lookup) │
│ (USE if match!) │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ EMOTIONAL DETECTOR │
│ preference/error/important │
│ ~0.5ms │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ BLOOM FILTER │
│ "Does it exist?" ~0ms │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ MEM0 (FIRST!) │
│ Fast cache ~20ms │
│ 80% token savings │
│ (RETURN if hit!) │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ EARLY WEIGHTING │
│ Adjust tier weights │
│ ~1ms │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ RUN TIERS PARALLEL │
│ acc-err, vestige, chromadb, │
│ gitnotes, file │
│ ~30ms │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ MERGE + RANKING │
│ Neuroscience scoring │
│ PASS 1: Quick filter │
│ PASS 2: Full rank │
│ ~10ms │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ CONFIDENCE EARLY EXIT │
│ confidence > 0.95? return 1│
│ gap > 0.5? return 1 │
└───────────────┬───────────────┘
│
┌───────────────▼───────────────┐
│ BACKGROUND SYNC │
│ Supermemory (daily backup) │
│ NOT in query path! │
└───────────────┬───────────────┘
│
▼
┌───────────────┐
│ RESULTS │
│ (~5-15ms) │
└───────────────┘
| Optimization | Time Saved |
|---|---|
| Ultra-Hot Tier | In-memory dict for last 10 queries (~0.1ms) |
| Compiled Queries | Pre-parsed common queries (~0ms) |
| Lazy Loading | Import heavy libs only when needed |
| Confidence Early Exit | Skip ranking if confident enough |
| Mem0 First | 80% queries hit here (~22ms) |
| Parallel Tiers | All tiers queried simultaneously |
| Tier | Name | Storage | Retention | Use Case |
|---|---|---|---|---|
| 1 | HOT | Session state | Current session | Active context, WAL buffer |
| 2 | WARM | Daily notes | 24-48 hours | Recent conversations, working memory |
| 3 | TEMP | Cache | Minutes-hours | Temporary processing, scratchpad |
| 4 | COLD | Core memory | Weeks-months | Important facts, decisions, preferences |
| 5 | ARCHIVE | Diary | Months-years | Long-term journal, milestone memories |
| 6 | COLD-STORAGE | Git-Notes | Indefinite | Permanent knowledge base |
| Query Type | Keywords | Weight Adjustments |
|---|---|---|
| Error/Fix | "bug", "fix", "error" | acc-error: 2x |
| Preference | "prefer", "like", "always" | vestige: 2x |
| Important | "remember", "critical" | all: 1.5x |
| Recent | "yesterday", "last week" | hot: 2x |
| Project | "project", "architecture" | gitnotes: 1.5x |
Two-pass approach for precision + speed:
| Pass | What | When |
|---|---|---|
| Pass 1 | Quick filter (skip 0 importance) | High-importance queries |
| Pass 2 | Full ranking (all components) | Always |
Final Score =
(Base Relevance × 0.25) +
(Importance × 0.30) + ← Hippocampus
(Value × 0.25) + ← VTA
(Emotion Match × 0.20) ← Amygdala
# Ensure Python 3.8+ is available
python3 --version
# Optional: ChromaDB for semantic search
pip install chromadb
# Optional: Ollama for embeddings
# Install from https://github.com/ollama/ollama
# The skill should be placed in your skills directory
# ~/.openclaw/workspace/skills/overkill-memory-system/
Copy .env.example to .env and configure:
cp .env.example .env
# Edit .env with your preferences
python3 cli.py init
This creates all required memory files:
~/.openclaw/memory/SESSION-STATE.md~/.openclaw/memory/MEMORY.md~/.openclaw/memory/cron-inbox.md~/.openclaw/memory/platform-posts.md~/.openclaw/memory/strategy-notes.md~/.openclaw/memory/heartbeat-state.json~/.openclaw/memory/diary/~/.openclaw/memory/daily/~/.openclaw/memory/chroma/~/.openclaw/memory/git-notes/# Initialize memory system files
python3 cli.py init
# Initialize with custom memory base path
python3 cli.py init --path /custom/path
# Add a memory with auto-detected importance & emotions
python3 cli.py add "Finished the project, feeling accomplished!"
# Add memory with explicit importance (0.0-1.0)
python3 cli.py add "Important decision made" --importance 0.9
# Add with explicit emotions
python3 cli.py add "Excited about the new feature" --emotions joy,curiosity
# Add with reward/value tracking
python3 cli.py add "Shipped v2.0" --reward accomplishment --intensity 0.8
# Search memories (hybrid - default, uses all optimizations)
python3 cli.py search "project updates"
# Fast mode (cache + ultra-hot only)
python3 cli.py search "query" --fast
# Full search (all tiers)
python3 cli.py search "query" --full
# Get recent memories
python3 cli.py recent --limit 10
# Get memories by importance threshold
python3 cli.py important --threshold 0.7
# Track an error
python3 cli.py error track "Forgot to add import"
# Show error patterns
python3 cli.py error patterns
# Show corrections made
python3 cli.py error corrections
# Error statistics
python3 cli.py error stats
# Search vestige memories
python3 cli.py vestige search "user preferences"
# Ingest with tags
python3 cli.py vestige ingest "User prefers dark mode" --tags preference
# Promote memory (strengthen)
python3 cli.py vestige promote <memory_id>
# Demote memory (weaken)
python3 cli.py vestige demote <memory_id>
# Check vestige stats
python3 cli.py vestige stats
# Search by file name (uses fd)
python3 cli.py file search "*.md"
# Search by content (uses rg)
python3 cli.py file content "TODO"
# Fast combined search
python3 cli.py file fast "pattern"
# Add atomic fact
python3 cli.py kg add --entity "people/kasper" --category "preference" --fact "Prefers TypeScript"
# Supersede old fact
python3 cli.py kg supersede --entity "people/kasper" --old kasper-001 --fact "New fact"
# Generate entity summary
python3 cli.py kg summarize --entity "people/kasper"
# Search knowledge graph
python3 cli.py kg search "preference"
# List all entities
python3 cli.py kg list
# Log an error
python3 cli.py improve error "Command failed" --context "details"
# Log user correction
python3 cli.py improve correct "No, that's wrong" --context "user corrected me"
# Log feature request
python3 cli.py improve request "Need markdown support"
# Log best practice
python3 cli.py improve better "Use async for I/O" --context "found during work"
# Get all learnings
python3 cli.py improve list
# Show neuroscience statistics
python3 cli.py neuro stats
# Analyze text for neuroscience scores
python3 cli.py neuro analyze "I'm excited about this project!"
# Start new session (flushes WAL to daily)
python3 cli.py session new
# End session (commits WAL buffer)
python3 cli.py session end
# Show session state
python3 cli.py session status
# Get current emotional state
python3 cli.py brain state
# Get motivation/drive level
python3 cli.py brain drive
# Update emotional dimensions
python3 cli.py brain update --valence 0.8 --arousal 0.6
# Create daily note entry
python3 cli.py daily "What happened today"
# Create diary entry (prompts for date)
python3 cli.py diary "Reflecting on the week"
# List recent diary entries
python3 cli.py diary list --limit 5
# Process cron inbox messages
python3 cli.py cron process
# Sync platform posts
python3 cli.py sync posts
# Run memory analysis
python3 cli.py analyze
# Show memory statistics
python3 cli.py stats
# Export memory backup
python3 cli.py export /path/to/backup/
# Import memory backup
python3 cli.py import /path/to/backup/
# Memory base directory
MEMORY_BASE=/home/user/.openclaw/memory
# ChromaDB settings (optional)
CHROMA_URL=http://localhost:8100
CHROMA_COLLECTION=memory-v2
# Ollama settings (optional)
OLLAMA_URL=http://localhost:11434
EMBEDDING_MODEL=bge-m3
# Capture settings
POLL_INTERVAL=300
# Processing settings
CHUNK_SIZE=512
CHUNK_OVERLAP=50
# Retrieval settings
CACHE_TTL=3600
MAX_RESULTS=10
~/.openclaw/memory/SESSION-STATE.md~/.openclaw/memory/daily/YYYY-MM-DD.md~/.cache/memory-v2/~/.openclaw/memory/MEMORY.md~/.openclaw/memory/diary/~/.openclaw/memory/git-notes/# Process cron inbox every 5 minutes
*/5 * * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py cron process >> /var/log/memory-cron.log 2>&1
# Sync platform posts every 15 minutes
*/15 * * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py sync posts >> /var/log/memory-sync.log 2>&1
# Daily diary entry at 9 PM
0 21 * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py diary "Daily reflection" >> /var/log/memory-diary.log 2>&1
# Weekly memory analysis (Sunday 10 PM)
0 22 * * 0 cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py analyze >> /var/log/memory-analyze.log 2>&1
Add to HEARTBEAT.md:
## Memory System Checks
- [ ] Check cron-inbox for cross-session messages
- [ ] Check platform-posts for new activity
- [ ] Review recent daily notes for important context
- [ ] Update emotional state if significantly changed
# Check directory permissions
ls -la ~/.openclaw/memory/
# Manually create directory
mkdir -p ~/.openclaw/memory
# Check if ChromaDB is running
curl http://localhost:8100/api/v1/heartbeat
# Or use keyword search fallback
python3 cli.py search "query" --method keyword
# Check Ollama is running
curl http://localhost:11434/api/tags
# Verify embedding model
ollama list
# Manually flush WAL buffer
python3 cli.py session end
# Check session file
cat ~/.openclaw/memory/SESSION-STATE.md
# Rebuild search index
python3 cli.py analyze
# Try keyword fallback
python3 cli.py search "term" --method keyword
# Check git-notes directory
ls -la ~/.openclaw/memory/git-notes/
# Initialize git repo if needed
cd ~/.openclaw/memory/git-notes && git init
overkill-memory-system/
├── SKILL.md # This file
├── README.md # Quick start guide
├── .env.example # Environment template
├── cli.py # Main CLI interface
├── config.py # Configuration
├── scripts/
│ └── analyze_memories.py # Memory analysis tool
├── templates/ # Future: custom templates
└── ULTIMATE_UNIFIED_FRAMEWORK.md # Full framework docs
Built with neuroscience-inspired architecture:
Based on the Ultimate Unified Memory Framework (ULTIMATE_UNIFIED_FRAMEWORK.md)
This skill was built by integrating ideas and features from the following ClawHub skills:
Last Updated: 2026-02-25 | Version 1.3.0 (Speed-First)
The system supports optional cloud backup and sync:
Configure via environment variables:
SUPERMEMORY_API_KEY - For cloud backupMEM0_API_KEY - For auto-fact extraction| Technique | Layer | Complexity | Benefit |
|---|---|---|---|
| Bloom Filters | Pre-query | O(1) | Skip expensive queries |
| Redis Hot Cache | L0 | <1ms | Sub-millisecond access |
| Mem0 L1 Cache | L1 | <10ms | 80% token reduction |
| Parallel Queries | All | O(1) wall | Concurrent tier queries |
| Connection Pooling | ChromaDB | Reuse | No connection overhead |
| Binary Search | Git-Notes | O(log n) | Fast sorted lookups |
| Pre-computed Embeddings | Cache | Skip compute | Cache hits = instant |
| Lazy Loading | Files | On-demand | Reduced memory footprint |
| Pre-fetch Context | Predictive | Anticipate | Results ready before ask |
| Result Caching | TTL | 1-5min | Avoid redundant queries |
Mem0 (L1 Cache) → ChromaDB → Git-Notes → Supermemory (Backup)
| Tier | Service | Purpose | Latency | Cost |
|---|---|---|---|---|
| L0 | Redis | Hot cache | <1ms | Low |
| L1 | Mem0 | Auto-extracted facts | <10ms | Medium |
| L2 | ChromaDB | Semantic vectors | <50ms | Low |
| L3 | Git-Notes | Knowledge graph | <20ms | Free |
| Backup | Supermemory | Offsite backup | Daily | Free |
MEM0_API_KEY environment variableSUPERMEMORY_API_KEY environment variable# Required for cloud features
MEM0_API_KEY=your_mem0_key # Auto-fact extraction
SUPERMEMORY_API_KEY=your_key # Cloud backup
# Optional overrides
CHROMA_URL=http://localhost:8100 # ChromaDB server
OLLAMA_URL=http://localhost:11434 # Ollama server
EMBEDDING_MODEL=bge-m3 # Embedding model
Query Input
│
▼
┌──────────────────────────────────────────────────────────────┐
│ 1. BLOOM FILTER CHECK (O(1)) │
│ • Probabilistic existence check │
│ • Skip expensive queries if definitely not present │
└──────────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ 2. REDIS HOT CACHE / L0 CACHE (Sub-millisecond) │
│ • TTL: 5-15 minutes │
│ • Frequently accessed memories │
│ • Return immediately if cached │
└──────────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ 3. MEM0 L1 CACHE (First Priority) │
│ • Auto-extracted facts (80% token reduction) │
│ • Fast fact lookup │
│ • No embedding computation needed │
└──────────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ 4. CHROMADB (Second Priority) │
│ • Semantic vector search (bge-m3 embeddings) │
│ • Connection pooling for speed │
│ • Return top-k results with scores │
└──────────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ 5. GIT-NOTES (Third Priority) │
│ • Structured JSON knowledge graph │
│ • Binary search on sorted index │
│ • O(log n) lookup time │
└──────────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ 6. FILE SEARCH (Fallback) │
│ • Raw grep on daily/diary files │
│ • Last resort fallback │
└──────────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ RESULTS MERGE & RANKING │
│ • Combine results from all tiers │
│ • Apply importance weights (Hippocampus) │
│ • Apply emotional relevance (Amygdala) │
│ • Apply value scores (VTA) │
│ • Return unified ranked results │
└──────────────────────────────────────────────────────────────┘
.env.example to .envpython3 cli.py --help to get startedThe CLI provides commands but cron jobs are NOT auto-installed. To enable:
crontab -e0 3 * * * python3 /path/to/cli.py cloud syncWhen Python imports cli.py, it may create memory directories under ~/.openclaw/memory/. This is intentional - the system needs these directories to function. To avoid this, run commands via subprocess rather than import.
The skill provides CLI commands for automation but does NOT auto-install cron jobs. You must manually add them if desired:
# Add to crontab -e
0 3 * * * python3 /path/to/cli.py cloud sync
Cloud features (Mem0, Supermemory) require API keys. Set in environment or .env file before use.
| Variable | When Accessed | External Service |
|---|---|---|
| CHROMA_URL | If set | ChromaDB server |
| OLLAMA_URL | If set | Ollama server |
| MEM0_API_KEY | If set AND MEM0_USE_LOCAL=false | Mem0.ai API |
| SUPERMEMORY_API_KEY | If set | Supermemory.ai API |
| REDIS_URL | If set | Redis server |
Only enabled when you:
These are opt-in only. Default = offline.