{ "schemaVersion": "1.0", "item": { "slug": "overkill-memory-system", "name": "Overkill Memory System", "source": "tencent", "type": "skill", "category": "开发工具", "sourceUrl": "https://clawhub.ai/Broedkrummen/overkill-memory-system", "canonicalUrl": "https://clawhub.ai/Broedkrummen/overkill-memory-system", "targetPlatform": "OpenClaw" }, "install": { "downloadMode": "manual_only", "downloadUrl": "/downloads/overkill-memory-system", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=overkill-memory-system", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "installMethod": "Manual import", "extraction": "Extract archive", "prerequisites": [ "OpenClaw" ], "packageFormat": "ZIP package", "includedAssets": [ "BRAIN_INTEGRATION.md", "FILE_SEARCH_INTEGRATION.md", "FINAL_ARCHITECTURE.md", "KG_INTEGRATION.md", "MULTI_AGENT_CHROMA.md", "NEURAL_MEMORY_ANALYSIS.md" ], "primaryDoc": "SKILL.md", "quickSetup": [ "Wait for the source to recover or retry later.", "Review SKILL.md only after the source returns a real package.", "Do not rely on this source for automated install yet." ], "agentAssist": { "summary": "Use the source page and any available docs to guide the install because the item is currently unstable or timing out.", "steps": [ "Open the source page via Review source status.", "If you can obtain the package, extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the source page and extracted files." ], "prompts": [ { "label": "New install", "body": "I tried to install a skill package from Yavira, but the item is currently unstable or timing out. Inspect the source page and any extracted docs, then tell me what you can confirm and any manual steps still required." }, { "label": "Upgrade existing", "body": "I tried to upgrade a skill package from Yavira, but the item is currently unstable or timing out. Compare the source page and any extracted docs with my current installation, then summarize what changed and what manual follow-up I still need." } ] }, "sourceHealth": { "source": "tencent", "slug": "overkill-memory-system", "status": "unstable", "reason": "timeout", "recommendedAction": "retry_later", "checkedAt": "2026-05-06T19:23:29.241Z", "expiresAt": "2026-05-07T07:23:29.241Z", "httpStatus": null, "finalUrl": null, "contentType": null, "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=overkill-memory-system", "error": "Timed out after 5000ms", "slug": "overkill-memory-system" }, "scope": "item", "summary": "Item is unstable.", "detail": "This item is timing out or returning errors right now. Review the source page and try again later.", "primaryActionLabel": "Review source status", "primaryActionHref": "https://clawhub.ai/Broedkrummen/overkill-memory-system" }, "validation": { "installChecklist": [ "Wait for the source to recover or retry later.", "Review SKILL.md only after the download returns a real package.", "Treat this source as transient until the upstream errors clear." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] }, "downloadPageUrl": "https://openagent3.xyz/downloads/overkill-memory-system", "agentPageUrl": "https://openagent3.xyz/skills/overkill-memory-system/agent", "manifestUrl": "https://openagent3.xyz/skills/overkill-memory-system/agent.json", "briefUrl": "https://openagent3.xyz/skills/overkill-memory-system/agent.md" }, "agentAssist": { "summary": "Use the source page and any available docs to guide the install because the item is currently unstable or timing out.", "steps": [ "Open the source page via Review source status.", "If you can obtain the package, extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the source page and extracted files." ], "prompts": [ { "label": "New install", "body": "I tried to install a skill package from Yavira, but the item is currently unstable or timing out. Inspect the source page and any extracted docs, then tell me what you can confirm and any manual steps still required." }, { "label": "Upgrade existing", "body": "I tried to upgrade a skill package from Yavira, but the item is currently unstable or timing out. Compare the source page and any extracted docs with my current installation, then summarize what changed and what manual follow-up I still need." } ] }, "documentation": { "source": "clawhub", "primaryDoc": "SKILL.md", "sections": [ { "title": "VERSION 1.9.3 (SPEED-FIRST)", "body": "A comprehensive 6-tier memory architecture with neuroscience integration, WAL protocol, and full automation for OpenClaw agents." }, { "title": "Overview", "body": "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." }, { "title": "What It Does", "body": "Brain-Full Architecture: 6 brain regions (Hippocampus, Amygdala, VTA, Basal Ganglia, Insula, ACC)\nSpeed-First Architecture: Optimized for ~5ms average query time\nFast File Search: Uses fd + rg for 10x faster file tier searching\nKnowledge Graph: Structured atomic facts with versioning\nSelf-Improving: Continuous learning from errors and corrections\nSelf-Reflection: Periodic self-assessment and performance review\nMulti-Agent Support: Shared + private ChromaDB areas per agent\n6-Tier Memory Architecture: From instant recall (HOT) to archival (COLD/GIT-NOTES)\nHybrid Neuroscience: Filter + Ranker approach for precision + speed\nWAL (Write-Ahead Log) Protocol: Ensures no memory is ever lost\nNeuroscience Integration: Hippocampus (importance), Amygdala (emotions), VTA (rewards/motivation)\nError Learning: Tracks and learns from user corrections\nSpaced Repetition: FSRS-6 via Vestige for natural memory decay\nSemantic Search: ChromaDB-powered vector storage for contextual retrieval\nCloud Backup: Supermemory integration for cross-device backup (NOT in query path)\nFull Automation: Cron jobs for cross-session messages, platform posts, diary entries, and proactive memory maintenance" }, { "title": "Speed Targets", "body": "ScenarioTimeCompiled query match~0msUltra-hot hit~0.1msHot cache hit~1msMem0 hit~22msFull search~55msAverage~5ms\n\nNote: 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." }, { "title": "Speed-First Architecture Diagram", "body": "┌─────────────────────────────────────────────────────────────────┐\n│ USER QUERY │\n└─────────────────────────┬───────────────────────────────────────┘\n │\n ┌───────────────▼───────────────┐\n │ ULTRA-HOT (Dict) │\n │ Last 10 queries ~0.1ms │\n │ (RETURN if hit!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ HOT CACHE (Redis) │\n │ Recent queries ~1ms │\n │ (RETURN if hit!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ COMPILED QUERIES │\n │ Pre-parsed common queries │\n │ ~0ms (dict lookup) │\n │ (USE if match!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ EMOTIONAL DETECTOR │\n │ preference/error/important │\n │ ~0.5ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ BLOOM FILTER │\n │ \"Does it exist?\" ~0ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ MEM0 (FIRST!) │\n │ Fast cache ~20ms │\n │ 80% token savings │\n │ (RETURN if hit!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ EARLY WEIGHTING │\n │ Adjust tier weights │\n │ ~1ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ RUN TIERS PARALLEL │\n │ acc-err, vestige, chromadb, │\n │ gitnotes, file │\n │ ~30ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ MERGE + RANKING │\n │ Neuroscience scoring │\n │ PASS 1: Quick filter │\n │ PASS 2: Full rank │\n │ ~10ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ CONFIDENCE EARLY EXIT │\n │ confidence > 0.95? return 1│\n │ gap > 0.5? return 1 │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ BACKGROUND SYNC │\n │ Supermemory (daily backup) │\n │ NOT in query path! │\n └───────────────┬───────────────┘\n │\n ▼\n ┌───────────────┐\n │ RESULTS │\n │ (~5-15ms) │\n └───────────────┘" }, { "title": "1. Speed Optimizations (NEW in v1.3.0)", "body": "OptimizationTime SavedUltra-Hot TierIn-memory dict for last 10 queries (~0.1ms)Compiled QueriesPre-parsed common queries (~0ms)Lazy LoadingImport heavy libs only when neededConfidence Early ExitSkip ranking if confident enoughMem0 First80% queries hit here (~22ms)Parallel TiersAll tiers queried simultaneously" }, { "title": "2. Six-Tier Memory Architecture", "body": "TierNameStorageRetentionUse Case1HOTSession stateCurrent sessionActive context, WAL buffer2WARMDaily notes24-48 hoursRecent conversations, working memory3TEMPCacheMinutes-hoursTemporary processing, scratchpad4COLDCore memoryWeeks-monthsImportant facts, decisions, preferences5ARCHIVEDiaryMonths-yearsLong-term journal, milestone memories6COLD-STORAGEGit-NotesIndefinitePermanent knowledge base" }, { "title": "2. Neuroscience Components", "body": "Hippocampus (Importance Scoring)\n\nAnalyzes content for importance signals\nMaintains index.json with memory importance scores\nAuto-weights memories based on repetition and context\n\nAmygdala (Emotional Tagging)\n\nDetects 8 emotions: joy, sadness, anger, fear, curiosity, connection, accomplishment, fatigue\nTracks emotional dimensions: valence, arousal, connection, curiosity, energy\nStores state in emotional-state.json\n\nVTA (Value/Reward System)\n\nComputes motivation scores based on reward types\nReward categories: accomplishment, social, curiosity, connection, creative, competence\nDrives attention toward high-value memories" }, { "title": "3. Hybrid Search (NEW in v1.3.0)", "body": "Emotional Detector\n\nDetects query intent: preference, error, important, recent, project, general\nAdjusts tier weights based on detected intent\nRuns AFTER cache checks (only when needed)\n\nEarly Weighting\n\nQuery TypeKeywordsWeight AdjustmentsError/Fix\"bug\", \"fix\", \"error\"acc-error: 2xPreference\"prefer\", \"like\", \"always\"vestige: 2xImportant\"remember\", \"critical\"all: 1.5xRecent\"yesterday\", \"last week\"hot: 2xProject\"project\", \"architecture\"gitnotes: 1.5x" }, { "title": "4. Hybrid Neuroscience (NEW in v1.3.0)", "body": "Two-pass approach for precision + speed:\n\nPassWhatWhenPass 1Quick filter (skip 0 importance)High-importance queriesPass 2Full ranking (all components)Always\n\nScoring Formula\n\nFinal Score = \n (Base Relevance × 0.25) +\n (Importance × 0.30) + ← Hippocampus\n (Value × 0.25) + ← VTA\n (Emotion Match × 0.20) ← Amygdala" }, { "title": "5. Error Learning (NEW in v1.3.0)", "body": "acc-error-memory integration\nTracks error patterns over time\nRecords user corrections\nLearns from mistakes\nHigh priority in search results" }, { "title": "6. Spaced Repetition (NEW in v1.3.0)", "body": "vestige integration (FSRS-6)\nMemories fade naturally like human memory\nPreferences strengthen with use\nSolutions decay if unused" }, { "title": "7. Write-Ahead Log (WAL) Protocol", "body": "Session state maintained in SESSION-STATE.md\nWAL buffer ensures atomic commits\nCrash recovery from uncommitted state" }, { "title": "4. Automation Features", "body": "Cron Inbox: Cross-session messages via cron-inbox.md\nPlatform Posts: Tracks Discord/Telegram posts in platform-posts.md\nDiary Entry: Daily journal entries in diary/ directory\nDaily Notes: Session logs in daily/ directory\nHeartbeat State: Tracks periodic check timestamps" }, { "title": "Prerequisites", "body": "# Ensure Python 3.8+ is available\npython3 --version\n\n# Optional: ChromaDB for semantic search\npip install chromadb\n\n# Optional: Ollama for embeddings\n# Install from https://github.com/ollama/ollama" }, { "title": "Step 1: Install the Skill", "body": "# The skill should be placed in your skills directory\n# ~/.openclaw/workspace/skills/overkill-memory-system/" }, { "title": "Step 2: Configure Environment", "body": "Copy .env.example to .env and configure:\n\ncp .env.example .env\n# Edit .env with your preferences" }, { "title": "Step 3: Initialize Memory System", "body": "python3 cli.py init\n\nThis creates all required memory files:\n\n~/.openclaw/memory/SESSION-STATE.md\n~/.openclaw/memory/MEMORY.md\n~/.openclaw/memory/cron-inbox.md\n~/.openclaw/memory/platform-posts.md\n~/.openclaw/memory/strategy-notes.md\n~/.openclaw/memory/heartbeat-state.json\n~/.openclaw/memory/diary/\n~/.openclaw/memory/daily/\n~/.openclaw/memory/chroma/\n~/.openclaw/memory/git-notes/" }, { "title": "Initialization", "body": "# Initialize memory system files\npython3 cli.py init\n\n# Initialize with custom memory base path\npython3 cli.py init --path /custom/path" }, { "title": "Memory Operations", "body": "# Add a memory with auto-detected importance & emotions\npython3 cli.py add \"Finished the project, feeling accomplished!\"\n\n# Add memory with explicit importance (0.0-1.0)\npython3 cli.py add \"Important decision made\" --importance 0.9\n\n# Add with explicit emotions\npython3 cli.py add \"Excited about the new feature\" --emotions joy,curiosity\n\n# Add with reward/value tracking\npython3 cli.py add \"Shipped v2.0\" --reward accomplishment --intensity 0.8" }, { "title": "Retrieval", "body": "# Search memories (hybrid - default, uses all optimizations)\npython3 cli.py search \"project updates\"\n\n# Fast mode (cache + ultra-hot only)\npython3 cli.py search \"query\" --fast\n\n# Full search (all tiers)\npython3 cli.py search \"query\" --full\n\n# Get recent memories\npython3 cli.py recent --limit 10\n\n# Get memories by importance threshold\npython3 cli.py important --threshold 0.7" }, { "title": "Error Tracking (NEW)", "body": "# Track an error\npython3 cli.py error track \"Forgot to add import\"\n\n# Show error patterns\npython3 cli.py error patterns\n\n# Show corrections made\npython3 cli.py error corrections\n\n# Error statistics\npython3 cli.py error stats" }, { "title": "Vestige Integration (NEW)", "body": "# Search vestige memories\npython3 cli.py vestige search \"user preferences\"\n\n# Ingest with tags\npython3 cli.py vestige ingest \"User prefers dark mode\" --tags preference\n\n# Promote memory (strengthen)\npython3 cli.py vestige promote \n\n# Demote memory (weaken)\npython3 cli.py vestige demote \n\n# Check vestige stats\npython3 cli.py vestige stats" }, { "title": "File Search (NEW)", "body": "# Search by file name (uses fd)\npython3 cli.py file search \"*.md\"\n\n# Search by content (uses rg)\npython3 cli.py file content \"TODO\"\n\n# Fast combined search\npython3 cli.py file fast \"pattern\"" }, { "title": "Knowledge Graph (NEW)", "body": "# Add atomic fact\npython3 cli.py kg add --entity \"people/kasper\" --category \"preference\" --fact \"Prefers TypeScript\"\n\n# Supersede old fact\npython3 cli.py kg supersede --entity \"people/kasper\" --old kasper-001 --fact \"New fact\"\n\n# Generate entity summary\npython3 cli.py kg summarize --entity \"people/kasper\"\n\n# Search knowledge graph\npython3 cli.py kg search \"preference\"\n\n# List all entities\npython3 cli.py kg list" }, { "title": "Self-Improving (NEW)", "body": "# Log an error\npython3 cli.py improve error \"Command failed\" --context \"details\"\n\n# Log user correction\npython3 cli.py improve correct \"No, that's wrong\" --context \"user corrected me\"\n\n# Log feature request\npython3 cli.py improve request \"Need markdown support\"\n\n# Log best practice\npython3 cli.py improve better \"Use async for I/O\" --context \"found during work\"\n\n# Get all learnings\npython3 cli.py improve list" }, { "title": "Neuroscience (NEW)", "body": "# Show neuroscience statistics\npython3 cli.py neuro stats\n\n# Analyze text for neuroscience scores\npython3 cli.py neuro analyze \"I'm excited about this project!\"" }, { "title": "Session Management", "body": "# Start new session (flushes WAL to daily)\npython3 cli.py session new\n\n# End session (commits WAL buffer)\npython3 cli.py session end\n\n# Show session state\npython3 cli.py session status" }, { "title": "Neuroscience Queries", "body": "# Get current emotional state\npython3 cli.py brain state\n\n# Get motivation/drive level\npython3 cli.py brain drive\n\n# Update emotional dimensions\npython3 cli.py brain update --valence 0.8 --arousal 0.6" }, { "title": "Daily & Diary", "body": "# Create daily note entry\npython3 cli.py daily \"What happened today\"\n\n# Create diary entry (prompts for date)\npython3 cli.py diary \"Reflecting on the week\"\n\n# List recent diary entries\npython3 cli.py diary list --limit 5" }, { "title": "Automation", "body": "# Process cron inbox messages\npython3 cli.py cron process\n\n# Sync platform posts\npython3 cli.py sync posts\n\n# Run memory analysis\npython3 cli.py analyze" }, { "title": "Utilities", "body": "# Show memory statistics\npython3 cli.py stats\n\n# Export memory backup\npython3 cli.py export /path/to/backup/\n\n# Import memory backup\npython3 cli.py import /path/to/backup/" }, { "title": "Configuration (.env)", "body": "# Memory base directory\nMEMORY_BASE=/home/user/.openclaw/memory\n\n# ChromaDB settings (optional)\nCHROMA_URL=http://localhost:8100\nCHROMA_COLLECTION=memory-v2\n\n# Ollama settings (optional)\nOLLAMA_URL=http://localhost:11434\nEMBEDDING_MODEL=bge-m3\n\n# Capture settings\nPOLL_INTERVAL=300\n\n# Processing settings\nCHUNK_SIZE=512\nCHUNK_OVERLAP=50\n\n# Retrieval settings\nCACHE_TTL=3600\nMAX_RESULTS=10" }, { "title": "Tier 1: HOT (Session State)", "body": "Location: ~/.openclaw/memory/SESSION-STATE.md\nSize: Keep under 50KB\nContent: Active context, current task, recent messages" }, { "title": "Tier 2: WARM (Daily)", "body": "Location: ~/.openclaw/memory/daily/YYYY-MM-DD.md\nSize: Up to 100KB per day\nContent: Daily logs, conversation summaries" }, { "title": "Tier 3: TEMP (Cache)", "body": "Location: ~/.cache/memory-v2/\nSize: Auto-cleaned after 24h\nContent: Processing scratchpad, temporary embeddings" }, { "title": "Tier 4: COLD (Core)", "body": "Location: ~/.openclaw/memory/MEMORY.md\nSize: Keep under 500KB\nContent: Key facts, decisions, preferences, lessons learned" }, { "title": "Tier 5: ARCHIVE (Diary)", "body": "Location: ~/.openclaw/memory/diary/\nSize: Unlimited\nContent: Personal journal, milestone reflections" }, { "title": "Tier 6: COLD-STORAGE (Git-Notes)", "body": "Location: ~/.openclaw/memory/git-notes/\nSize: Unlimited\nContent: Knowledge base, permanent reference" }, { "title": "Recommended Cron Setup", "body": "# Process cron inbox every 5 minutes\n*/5 * * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py cron process >> /var/log/memory-cron.log 2>&1\n\n# Sync platform posts every 15 minutes\n*/15 * * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py sync posts >> /var/log/memory-sync.log 2>&1\n\n# Daily diary entry at 9 PM\n0 21 * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py diary \"Daily reflection\" >> /var/log/memory-diary.log 2>&1\n\n# Weekly memory analysis (Sunday 10 PM)\n0 22 * * 0 cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py analyze >> /var/log/memory-analyze.log 2>&1" }, { "title": "Heartbeat Integration", "body": "Add to HEARTBEAT.md:\n\n## Memory System Checks\n\n- [ ] Check cron-inbox for cross-session messages\n- [ ] Check platform-posts for new activity\n- [ ] Review recent daily notes for important context\n- [ ] Update emotional state if significantly changed" }, { "title": "Memory System Won't Initialize", "body": "# Check directory permissions\nls -la ~/.openclaw/memory/\n\n# Manually create directory\nmkdir -p ~/.openclaw/memory" }, { "title": "ChromaDB Connection Failed", "body": "# Check if ChromaDB is running\ncurl http://localhost:8100/api/v1/heartbeat\n\n# Or use keyword search fallback\npython3 cli.py search \"query\" --method keyword" }, { "title": "Ollama Embeddings Not Working", "body": "# Check Ollama is running\ncurl http://localhost:11434/api/tags\n\n# Verify embedding model\nollama list" }, { "title": "Session State Not Persisting", "body": "# Manually flush WAL buffer\npython3 cli.py session end\n\n# Check session file\ncat ~/.openclaw/memory/SESSION-STATE.md" }, { "title": "Memory Search Returns No Results", "body": "# Rebuild search index\npython3 cli.py analyze\n\n# Try keyword fallback\npython3 cli.py search \"term\" --method keyword" }, { "title": "Git-Notes Sync Issues", "body": "# Check git-notes directory\nls -la ~/.openclaw/memory/git-notes/\n\n# Initialize git repo if needed\ncd ~/.openclaw/memory/git-notes && git init" }, { "title": "File Structure", "body": "overkill-memory-system/\n├── SKILL.md # This file\n├── README.md # Quick start guide\n├── .env.example # Environment template\n├── cli.py # Main CLI interface\n├── config.py # Configuration\n├── scripts/\n│ └── analyze_memories.py # Memory analysis tool\n├── templates/ # Future: custom templates\n└── ULTIMATE_UNIFIED_FRAMEWORK.md # Full framework docs" }, { "title": "Credits & Sources", "body": "vestige - FSRS-6 spaced repetition for natural memory decay and preferences\nacc-error-memory - Error pattern tracking and correction learning\n\nBuilt with neuroscience-inspired architecture:\n\nHippocampus: Importance-based memory consolidation\nAmygdala: Emotional tagging and valence processing\nVTA: Reward-driven attention and motivation\n\nBased on the Ultimate Unified Memory Framework (ULTIMATE_UNIFIED_FRAMEWORK.md)" }, { "title": "Credits & Sources", "body": "vestige - FSRS-6 spaced repetition for natural memory decay and preferences\nacc-error-memory - Error pattern tracking and correction learning\n\nThis skill was built by integrating ideas and features from the following ClawHub skills:" }, { "title": "Core Architecture", "body": "elite-longterm-memory - WAL Protocol, Git-Notes knowledge graph, SESSION-STATE.md concept\njarvis-memory-architecture - Cron inbox, diary, daily logs, platform post tracking, adaptive learning\nmemory-hygiene - Auto-cleanup, storage guidelines" }, { "title": "Neuroscience Components", "body": "hippocampus-memory - Importance-weighted recall and memory encoding\namygdala-memory - Emotional tagging and processing\nvta-memory - Value scoring and motivation tracking" }, { "title": "Storage & Integration", "body": "chromadb-memory - Vector storage integration (ChromaDB + Ollama bge-m3)\nsupermemory-free - Optional cloud backup integration\nmem0 - Auto-fact extraction (80% token reduction)\nmemory-system-v2 - Core unified memory framework" }, { "title": "Created By", "body": "Initial implementation by Cody (AI coding specialist)\nFramework designed by Broedkrummen\nBuilt with OpenClaw agent-orchestrator\n\nLast Updated: 2026-02-25 | Version 1.3.0 (Speed-First)" }, { "title": "Cloud Integration (Requires Setup)", "body": "The system supports optional cloud backup and sync:\n\nSupermemory Integration: Push memories to cloud for cross-device access\nMem0 Auto-Fact Extraction: Automatic fact extraction from conversations (80% token reduction)\n\nConfigure via environment variables:\n\nSUPERMEMORY_API_KEY - For cloud backup\nMEM0_API_KEY - For auto-fact extraction" }, { "title": "Optimization Techniques Implemented", "body": "TechniqueLayerComplexityBenefitBloom FiltersPre-queryO(1)Skip expensive queriesRedis Hot CacheL0<1msSub-millisecond accessMem0 L1 CacheL1<10ms80% token reductionParallel QueriesAllO(1) wallConcurrent tier queriesConnection PoolingChromaDBReuseNo connection overheadBinary SearchGit-NotesO(log n)Fast sorted lookupsPre-computed EmbeddingsCacheSkip computeCache hits = instantLazy LoadingFilesOn-demandReduced memory footprintPre-fetch ContextPredictiveAnticipateResults ready before askResult CachingTTL1-5minAvoid redundant queries" }, { "title": "L1 Cache (Mem0)", "body": "Purpose: First-layer cache for 80% token reduction\nHow: Mem0 extracts facts from conversations automatically\nBenefit: Reduces context window usage while preserving key information" }, { "title": "Parallel Tier Query", "body": "Purpose: Query all memory tiers simultaneously\nHow: Async queries to Mem0, ChromaDB, Git-Notes, and file search\nBenefit: O(1) wall-clock time instead of sequential O(n) tier traversal" }, { "title": "Redis Hot Cache (L0)", "body": "Purpose: Ultra-fast L0 cache for frequently accessed memories\nTTL: 5-15 minutes for hot data\nBenefit: Sub-millisecond access for top results" }, { "title": "Result Caching with TTL", "body": "Purpose: Cache search results to avoid redundant queries\nTTL: 1-5 minutes depending on tier\nBenefit: Dramatically reduces API calls and computation" }, { "title": "Binary Search (Git-Notes)", "body": "Purpose: O(log n) lookup in sorted memory index\nHow: Maintain sorted timestamp/index files\nBenefit: Fast retrieval from large Git-Notes collections" }, { "title": "Connection Pooling", "body": "Purpose: Reuse ChromaDB and Ollama connections\nHow: Persistent connection pools with health checks\nBenefit: Eliminates connection overhead on each query" }, { "title": "Bloom Filters", "body": "Purpose: Quick existence checks before expensive queries\nHow: Probabilistic filter for memory presence\nBenefit: Skip unnecessary tier searches when result is definitely not present" }, { "title": "Pre-fetch Context", "body": "Purpose: Predictive memory loading based on context\nHow: Anticipate likely queries based on current session\nBenefit: Results ready before user asks" }, { "title": "Lazy Loading", "body": "Purpose: Load files only when needed\nHow: On-demand loading of large files\nBenefit: Reduced memory footprint and faster initial response" }, { "title": "Pre-computed Embeddings", "body": "Purpose: Cache embeddings for frequently queried content\nHow: Store embeddings alongside source data\nBenefit: Skip embedding computation on cache hit\nHow: Store embeddings alongside source data\nBenefit: Skip embedding computation on cache hit" }, { "title": "Priority Order", "body": "Mem0 (L1 Cache) → ChromaDB → Git-Notes → Supermemory (Backup)\n\nTierServicePurposeLatencyCostL0RedisHot cache<1msLowL1Mem0Auto-extracted facts<10msMediumL2ChromaDBSemantic vectors<50msLowL3Git-NotesKnowledge graph<20msFreeBackupSupermemoryOffsite backupDailyFree" }, { "title": "Cloud Services Integration", "body": "Mem0 (L1 Cache)\n\nPurpose: First-layer cache for 80% token reduction\nHow: Auto-extracts facts from conversations\nAPI: MEM0_API_KEY environment variable\nBenefit: Reduces context window usage while preserving key information\n\nChromaDB (Vector Storage)\n\nPurpose: Semantic similarity search\nEmbeddings: bge-m3 via Ollama\nConnection: Pooled connections for speed\nFallback: Keyword search if unavailable\n\nGit-Notes (Knowledge Graph)\n\nPurpose: Structured JSON storage\nLookup: Binary search O(log n)\nSync: Git-based versioning\n\nSupermemory (Cloud Backup)\n\nPurpose: Daily backup only (not real-time sync)\nFrequency: Once per day\nAPI: SUPERMEMORY_API_KEY environment variable\nBenefit: Reduces API calls while maintaining offsite backup" }, { "title": "Environment Variables", "body": "# Required for cloud features\nMEM0_API_KEY=your_mem0_key # Auto-fact extraction\nSUPERMEMORY_API_KEY=your_key # Cloud backup\n\n# Optional overrides\nCHROMA_URL=http://localhost:8100 # ChromaDB server\nOLLAMA_URL=http://localhost:11434 # Ollama server\nEMBEDDING_MODEL=bge-m3 # Embedding model" }, { "title": "Search Priority Flow (v1.0.5)", "body": "Query Input\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 1. BLOOM FILTER CHECK (O(1)) │\n│ • Probabilistic existence check │\n│ • Skip expensive queries if definitely not present │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 2. REDIS HOT CACHE / L0 CACHE (Sub-millisecond) │\n│ • TTL: 5-15 minutes │\n│ • Frequently accessed memories │\n│ • Return immediately if cached │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 3. MEM0 L1 CACHE (First Priority) │\n│ • Auto-extracted facts (80% token reduction) │\n│ • Fast fact lookup │\n│ • No embedding computation needed │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 4. CHROMADB (Second Priority) │\n│ • Semantic vector search (bge-m3 embeddings) │\n│ • Connection pooling for speed │\n│ • Return top-k results with scores │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 5. GIT-NOTES (Third Priority) │\n│ • Structured JSON knowledge graph │\n│ • Binary search on sorted index │\n│ • O(log n) lookup time │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 6. FILE SEARCH (Fallback) │\n│ • Raw grep on daily/diary files │\n│ • Last resort fallback │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ RESULTS MERGE & RANKING │\n│ • Combine results from all tiers │\n│ • Apply importance weights (Hippocampus) │\n│ • Apply emotional relevance (Amygdala) │\n│ • Apply value scores (VTA) │\n│ • Return unified ranked results │\n└──────────────────────────────────────────────────────────────┘" }, { "title": "Cache Strategy Details", "body": "Cache Hit: Return cached result immediately (sub-ms)\nCache Miss: Query next tier, cache result with TTL\nNegative Cache: Optionally cache \"not found\" results (shorter TTL)\nCache Invalidation: On session end, new memory add, or manual trigger" }, { "title": "Required Services (must be running)", "body": "ChromaDB on http://localhost:8100\nOllama on http://localhost:11434 with bge-m3 model" }, { "title": "Optional Services (require API keys)", "body": "Mem0.ai account (for cloud fact extraction)\nSupermemory.ai account (for cloud backup)\nRedis (optional, falls back to in-memory)" }, { "title": "Environment Setup", "body": "Copy .env.example to .env\nFill in optional API keys if using cloud features\nRun python3 cli.py --help to get started" }, { "title": "Manual Setup for Automation", "body": "The CLI provides commands but cron jobs are NOT auto-installed. To enable:\n\nAdd cron jobs manually via crontab -e\nExample: 0 3 * * * python3 /path/to/cli.py cloud sync" }, { "title": "On-Import Side Effects", "body": "When 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." }, { "title": "No Auto-Installed Cron Jobs", "body": "The skill provides CLI commands for automation but does NOT auto-install cron jobs. You must manually add them if desired:\n\n# Add to crontab -e\n0 3 * * * python3 /path/to/cli.py cloud sync" }, { "title": "Cloud Features", "body": "Cloud features (Mem0, Supermemory) require API keys. Set in environment or .env file before use." }, { "title": "When Network Access Occurs", "body": "VariableWhen AccessedExternal ServiceCHROMA_URLIf setChromaDB serverOLLAMA_URLIf setOllama serverMEM0_API_KEYIf set AND MEM0_USE_LOCAL=falseMem0.ai APISUPERMEMORY_API_KEYIf setSupermemory.ai APIREDIS_URLIf setRedis server" }, { "title": "Default Behavior (No Network)", "body": "Without API keys, system runs fully offline\nUses local ChromaDB + local Ollama (if available)\nAll data stored locally in ~/.openclaw/memory/" }, { "title": "Cloud Features", "body": "Only enabled when you:\n\nSet MEM0_API_KEY and set MEM0_USE_LOCAL=false\nSet SUPERMEMORY_API_KEY\n\nThese are opt-in only. Default = offline." } ], "body": "Ultimate Unified Memory System (Overkill Memory System)\nVERSION 1.9.3 (SPEED-FIRST)\n\nA comprehensive 6-tier memory architecture with neuroscience integration, WAL protocol, and full automation for OpenClaw agents.\n\nOverview\n\nThe 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.\n\nWhat It Does\nBrain-Full Architecture: 6 brain regions (Hippocampus, Amygdala, VTA, Basal Ganglia, Insula, ACC)\nSpeed-First Architecture: Optimized for ~5ms average query time\nFast File Search: Uses fd + rg for 10x faster file tier searching\nKnowledge Graph: Structured atomic facts with versioning\nSelf-Improving: Continuous learning from errors and corrections\nSelf-Reflection: Periodic self-assessment and performance review\nMulti-Agent Support: Shared + private ChromaDB areas per agent\n6-Tier Memory Architecture: From instant recall (HOT) to archival (COLD/GIT-NOTES)\nHybrid Neuroscience: Filter + Ranker approach for precision + speed\nWAL (Write-Ahead Log) Protocol: Ensures no memory is ever lost\nNeuroscience Integration: Hippocampus (importance), Amygdala (emotions), VTA (rewards/motivation)\nError Learning: Tracks and learns from user corrections\nSpaced Repetition: FSRS-6 via Vestige for natural memory decay\nSemantic Search: ChromaDB-powered vector storage for contextual retrieval\nCloud Backup: Supermemory integration for cross-device backup (NOT in query path)\nFull Automation: Cron jobs for cross-session messages, platform posts, diary entries, and proactive memory maintenance\nSpeed Targets\nScenario\tTime\nCompiled query match\t~0ms\nUltra-hot hit\t~0.1ms\nHot cache hit\t~1ms\nMem0 hit\t~22ms\nFull search\t~55ms\nAverage\t~5ms\n\nNote: 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.\n\nSpeed-First Architecture Diagram\n┌─────────────────────────────────────────────────────────────────┐\n│ USER QUERY │\n└─────────────────────────┬───────────────────────────────────────┘\n │\n ┌───────────────▼───────────────┐\n │ ULTRA-HOT (Dict) │\n │ Last 10 queries ~0.1ms │\n │ (RETURN if hit!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ HOT CACHE (Redis) │\n │ Recent queries ~1ms │\n │ (RETURN if hit!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ COMPILED QUERIES │\n │ Pre-parsed common queries │\n │ ~0ms (dict lookup) │\n │ (USE if match!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ EMOTIONAL DETECTOR │\n │ preference/error/important │\n │ ~0.5ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ BLOOM FILTER │\n │ \"Does it exist?\" ~0ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ MEM0 (FIRST!) │\n │ Fast cache ~20ms │\n │ 80% token savings │\n │ (RETURN if hit!) │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ EARLY WEIGHTING │\n │ Adjust tier weights │\n │ ~1ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ RUN TIERS PARALLEL │\n │ acc-err, vestige, chromadb, │\n │ gitnotes, file │\n │ ~30ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ MERGE + RANKING │\n │ Neuroscience scoring │\n │ PASS 1: Quick filter │\n │ PASS 2: Full rank │\n │ ~10ms │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ CONFIDENCE EARLY EXIT │\n │ confidence > 0.95? return 1│\n │ gap > 0.5? return 1 │\n └───────────────┬───────────────┘\n │\n ┌───────────────▼───────────────┐\n │ BACKGROUND SYNC │\n │ Supermemory (daily backup) │\n │ NOT in query path! │\n └───────────────┬───────────────┘\n │\n ▼\n ┌───────────────┐\n │ RESULTS │\n │ (~5-15ms) │\n └───────────────┘\n\nFeatures\n1. Speed Optimizations (NEW in v1.3.0)\nOptimization\tTime Saved\nUltra-Hot Tier\tIn-memory dict for last 10 queries (~0.1ms)\nCompiled Queries\tPre-parsed common queries (~0ms)\nLazy Loading\tImport heavy libs only when needed\nConfidence Early Exit\tSkip ranking if confident enough\nMem0 First\t80% queries hit here (~22ms)\nParallel Tiers\tAll tiers queried simultaneously\n2. Six-Tier Memory Architecture\nTier\tName\tStorage\tRetention\tUse Case\n1\tHOT\tSession state\tCurrent session\tActive context, WAL buffer\n2\tWARM\tDaily notes\t24-48 hours\tRecent conversations, working memory\n3\tTEMP\tCache\tMinutes-hours\tTemporary processing, scratchpad\n4\tCOLD\tCore memory\tWeeks-months\tImportant facts, decisions, preferences\n5\tARCHIVE\tDiary\tMonths-years\tLong-term journal, milestone memories\n6\tCOLD-STORAGE\tGit-Notes\tIndefinite\tPermanent knowledge base\n2. Neuroscience Components\nHippocampus (Importance Scoring)\nAnalyzes content for importance signals\nMaintains index.json with memory importance scores\nAuto-weights memories based on repetition and context\nAmygdala (Emotional Tagging)\nDetects 8 emotions: joy, sadness, anger, fear, curiosity, connection, accomplishment, fatigue\nTracks emotional dimensions: valence, arousal, connection, curiosity, energy\nStores state in emotional-state.json\nVTA (Value/Reward System)\nComputes motivation scores based on reward types\nReward categories: accomplishment, social, curiosity, connection, creative, competence\nDrives attention toward high-value memories\n3. Hybrid Search (NEW in v1.3.0)\nEmotional Detector\nDetects query intent: preference, error, important, recent, project, general\nAdjusts tier weights based on detected intent\nRuns AFTER cache checks (only when needed)\nEarly Weighting\nQuery Type\tKeywords\tWeight Adjustments\nError/Fix\t\"bug\", \"fix\", \"error\"\tacc-error: 2x\nPreference\t\"prefer\", \"like\", \"always\"\tvestige: 2x\nImportant\t\"remember\", \"critical\"\tall: 1.5x\nRecent\t\"yesterday\", \"last week\"\thot: 2x\nProject\t\"project\", \"architecture\"\tgitnotes: 1.5x\n4. Hybrid Neuroscience (NEW in v1.3.0)\n\nTwo-pass approach for precision + speed:\n\nPass\tWhat\tWhen\nPass 1\tQuick filter (skip 0 importance)\tHigh-importance queries\nPass 2\tFull ranking (all components)\tAlways\nScoring Formula\nFinal Score = \n (Base Relevance × 0.25) +\n (Importance × 0.30) + ← Hippocampus\n (Value × 0.25) + ← VTA\n (Emotion Match × 0.20) ← Amygdala\n\n5. Error Learning (NEW in v1.3.0)\nacc-error-memory integration\nTracks error patterns over time\nRecords user corrections\nLearns from mistakes\nHigh priority in search results\n6. Spaced Repetition (NEW in v1.3.0)\nvestige integration (FSRS-6)\nMemories fade naturally like human memory\nPreferences strengthen with use\nSolutions decay if unused\n7. Write-Ahead Log (WAL) Protocol\nSession state maintained in SESSION-STATE.md\nWAL buffer ensures atomic commits\nCrash recovery from uncommitted state\n4. Automation Features\nCron Inbox: Cross-session messages via cron-inbox.md\nPlatform Posts: Tracks Discord/Telegram posts in platform-posts.md\nDiary Entry: Daily journal entries in diary/ directory\nDaily Notes: Session logs in daily/ directory\nHeartbeat State: Tracks periodic check timestamps\nInstallation & Setup\nPrerequisites\n# Ensure Python 3.8+ is available\npython3 --version\n\n# Optional: ChromaDB for semantic search\npip install chromadb\n\n# Optional: Ollama for embeddings\n# Install from https://github.com/ollama/ollama\n\nStep 1: Install the Skill\n# The skill should be placed in your skills directory\n# ~/.openclaw/workspace/skills/overkill-memory-system/\n\nStep 2: Configure Environment\n\nCopy .env.example to .env and configure:\n\ncp .env.example .env\n# Edit .env with your preferences\n\nStep 3: Initialize Memory System\npython3 cli.py init\n\n\nThis creates all required memory files:\n\n~/.openclaw/memory/SESSION-STATE.md\n~/.openclaw/memory/MEMORY.md\n~/.openclaw/memory/cron-inbox.md\n~/.openclaw/memory/platform-posts.md\n~/.openclaw/memory/strategy-notes.md\n~/.openclaw/memory/heartbeat-state.json\n~/.openclaw/memory/diary/\n~/.openclaw/memory/daily/\n~/.openclaw/memory/chroma/\n~/.openclaw/memory/git-notes/\nCLI Commands\nInitialization\n# Initialize memory system files\npython3 cli.py init\n\n# Initialize with custom memory base path\npython3 cli.py init --path /custom/path\n\nMemory Operations\n# Add a memory with auto-detected importance & emotions\npython3 cli.py add \"Finished the project, feeling accomplished!\"\n\n# Add memory with explicit importance (0.0-1.0)\npython3 cli.py add \"Important decision made\" --importance 0.9\n\n# Add with explicit emotions\npython3 cli.py add \"Excited about the new feature\" --emotions joy,curiosity\n\n# Add with reward/value tracking\npython3 cli.py add \"Shipped v2.0\" --reward accomplishment --intensity 0.8\n\nRetrieval\n# Search memories (hybrid - default, uses all optimizations)\npython3 cli.py search \"project updates\"\n\n# Fast mode (cache + ultra-hot only)\npython3 cli.py search \"query\" --fast\n\n# Full search (all tiers)\npython3 cli.py search \"query\" --full\n\n# Get recent memories\npython3 cli.py recent --limit 10\n\n# Get memories by importance threshold\npython3 cli.py important --threshold 0.7\n\nError Tracking (NEW)\n# Track an error\npython3 cli.py error track \"Forgot to add import\"\n\n# Show error patterns\npython3 cli.py error patterns\n\n# Show corrections made\npython3 cli.py error corrections\n\n# Error statistics\npython3 cli.py error stats\n\nVestige Integration (NEW)\n# Search vestige memories\npython3 cli.py vestige search \"user preferences\"\n\n# Ingest with tags\npython3 cli.py vestige ingest \"User prefers dark mode\" --tags preference\n\n# Promote memory (strengthen)\npython3 cli.py vestige promote \n\n# Demote memory (weaken)\npython3 cli.py vestige demote \n\n# Check vestige stats\npython3 cli.py vestige stats\n\nFile Search (NEW)\n# Search by file name (uses fd)\npython3 cli.py file search \"*.md\"\n\n# Search by content (uses rg)\npython3 cli.py file content \"TODO\"\n\n# Fast combined search\npython3 cli.py file fast \"pattern\"\n\nKnowledge Graph (NEW)\n# Add atomic fact\npython3 cli.py kg add --entity \"people/kasper\" --category \"preference\" --fact \"Prefers TypeScript\"\n\n# Supersede old fact\npython3 cli.py kg supersede --entity \"people/kasper\" --old kasper-001 --fact \"New fact\"\n\n# Generate entity summary\npython3 cli.py kg summarize --entity \"people/kasper\"\n\n# Search knowledge graph\npython3 cli.py kg search \"preference\"\n\n# List all entities\npython3 cli.py kg list\n\nSelf-Improving (NEW)\n# Log an error\npython3 cli.py improve error \"Command failed\" --context \"details\"\n\n# Log user correction\npython3 cli.py improve correct \"No, that's wrong\" --context \"user corrected me\"\n\n# Log feature request\npython3 cli.py improve request \"Need markdown support\"\n\n# Log best practice\npython3 cli.py improve better \"Use async for I/O\" --context \"found during work\"\n\n# Get all learnings\npython3 cli.py improve list\n\nNeuroscience (NEW)\n# Show neuroscience statistics\npython3 cli.py neuro stats\n\n# Analyze text for neuroscience scores\npython3 cli.py neuro analyze \"I'm excited about this project!\"\n\nSession Management\n# Start new session (flushes WAL to daily)\npython3 cli.py session new\n\n# End session (commits WAL buffer)\npython3 cli.py session end\n\n# Show session state\npython3 cli.py session status\n\nNeuroscience Queries\n# Get current emotional state\npython3 cli.py brain state\n\n# Get motivation/drive level\npython3 cli.py brain drive\n\n# Update emotional dimensions\npython3 cli.py brain update --valence 0.8 --arousal 0.6\n\nDaily & Diary\n# Create daily note entry\npython3 cli.py daily \"What happened today\"\n\n# Create diary entry (prompts for date)\npython3 cli.py diary \"Reflecting on the week\"\n\n# List recent diary entries\npython3 cli.py diary list --limit 5\n\nAutomation\n# Process cron inbox messages\npython3 cli.py cron process\n\n# Sync platform posts\npython3 cli.py sync posts\n\n# Run memory analysis\npython3 cli.py analyze\n\nUtilities\n# Show memory statistics\npython3 cli.py stats\n\n# Export memory backup\npython3 cli.py export /path/to/backup/\n\n# Import memory backup\npython3 cli.py import /path/to/backup/\n\nConfiguration (.env)\n# Memory base directory\nMEMORY_BASE=/home/user/.openclaw/memory\n\n# ChromaDB settings (optional)\nCHROMA_URL=http://localhost:8100\nCHROMA_COLLECTION=memory-v2\n\n# Ollama settings (optional)\nOLLAMA_URL=http://localhost:11434\nEMBEDDING_MODEL=bge-m3\n\n# Capture settings\nPOLL_INTERVAL=300\n\n# Processing settings\nCHUNK_SIZE=512\nCHUNK_OVERLAP=50\n\n# Retrieval settings\nCACHE_TTL=3600\nMAX_RESULTS=10\n\nStorage Guidelines\nTier 1: HOT (Session State)\nLocation: ~/.openclaw/memory/SESSION-STATE.md\nSize: Keep under 50KB\nContent: Active context, current task, recent messages\nTier 2: WARM (Daily)\nLocation: ~/.openclaw/memory/daily/YYYY-MM-DD.md\nSize: Up to 100KB per day\nContent: Daily logs, conversation summaries\nTier 3: TEMP (Cache)\nLocation: ~/.cache/memory-v2/\nSize: Auto-cleaned after 24h\nContent: Processing scratchpad, temporary embeddings\nTier 4: COLD (Core)\nLocation: ~/.openclaw/memory/MEMORY.md\nSize: Keep under 500KB\nContent: Key facts, decisions, preferences, lessons learned\nTier 5: ARCHIVE (Diary)\nLocation: ~/.openclaw/memory/diary/\nSize: Unlimited\nContent: Personal journal, milestone reflections\nTier 6: COLD-STORAGE (Git-Notes)\nLocation: ~/.openclaw/memory/git-notes/\nSize: Unlimited\nContent: Knowledge base, permanent reference\nCron Jobs\nRecommended Cron Setup\n# Process cron inbox every 5 minutes\n*/5 * * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py cron process >> /var/log/memory-cron.log 2>&1\n\n# Sync platform posts every 15 minutes\n*/15 * * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py sync posts >> /var/log/memory-sync.log 2>&1\n\n# Daily diary entry at 9 PM\n0 21 * * * cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py diary \"Daily reflection\" >> /var/log/memory-diary.log 2>&1\n\n# Weekly memory analysis (Sunday 10 PM)\n0 22 * * 0 cd ~/.openclaw/workspace-cody/skills/overkill-memory-system && python3 cli.py analyze >> /var/log/memory-analyze.log 2>&1\n\nHeartbeat Integration\n\nAdd to HEARTBEAT.md:\n\n## Memory System Checks\n\n- [ ] Check cron-inbox for cross-session messages\n- [ ] Check platform-posts for new activity\n- [ ] Review recent daily notes for important context\n- [ ] Update emotional state if significantly changed\n\nTroubleshooting\nMemory System Won't Initialize\n# Check directory permissions\nls -la ~/.openclaw/memory/\n\n# Manually create directory\nmkdir -p ~/.openclaw/memory\n\nChromaDB Connection Failed\n# Check if ChromaDB is running\ncurl http://localhost:8100/api/v1/heartbeat\n\n# Or use keyword search fallback\npython3 cli.py search \"query\" --method keyword\n\nOllama Embeddings Not Working\n# Check Ollama is running\ncurl http://localhost:11434/api/tags\n\n# Verify embedding model\nollama list\n\nSession State Not Persisting\n# Manually flush WAL buffer\npython3 cli.py session end\n\n# Check session file\ncat ~/.openclaw/memory/SESSION-STATE.md\n\nMemory Search Returns No Results\n# Rebuild search index\npython3 cli.py analyze\n\n# Try keyword fallback\npython3 cli.py search \"term\" --method keyword\n\nGit-Notes Sync Issues\n# Check git-notes directory\nls -la ~/.openclaw/memory/git-notes/\n\n# Initialize git repo if needed\ncd ~/.openclaw/memory/git-notes && git init\n\nFile Structure\noverkill-memory-system/\n├── SKILL.md # This file\n├── README.md # Quick start guide\n├── .env.example # Environment template\n├── cli.py # Main CLI interface\n├── config.py # Configuration\n├── scripts/\n│ └── analyze_memories.py # Memory analysis tool\n├── templates/ # Future: custom templates\n└── ULTIMATE_UNIFIED_FRAMEWORK.md # Full framework docs\n\nCredits & Sources\nvestige - FSRS-6 spaced repetition for natural memory decay and preferences\nacc-error-memory - Error pattern tracking and correction learning\n\nBuilt with neuroscience-inspired architecture:\n\nHippocampus: Importance-based memory consolidation\nAmygdala: Emotional tagging and valence processing\nVTA: Reward-driven attention and motivation\n\nBased on the Ultimate Unified Memory Framework (ULTIMATE_UNIFIED_FRAMEWORK.md)\n\nCredits & Sources\nvestige - FSRS-6 spaced repetition for natural memory decay and preferences\nacc-error-memory - Error pattern tracking and correction learning\n\nThis skill was built by integrating ideas and features from the following ClawHub skills:\n\nCore Architecture\nelite-longterm-memory - WAL Protocol, Git-Notes knowledge graph, SESSION-STATE.md concept\njarvis-memory-architecture - Cron inbox, diary, daily logs, platform post tracking, adaptive learning\nmemory-hygiene - Auto-cleanup, storage guidelines\nNeuroscience Components\nhippocampus-memory - Importance-weighted recall and memory encoding\namygdala-memory - Emotional tagging and processing\nvta-memory - Value scoring and motivation tracking\nStorage & Integration\nchromadb-memory - Vector storage integration (ChromaDB + Ollama bge-m3)\nsupermemory-free - Optional cloud backup integration\nmem0 - Auto-fact extraction (80% token reduction)\nmemory-system-v2 - Core unified memory framework\nCreated By\nInitial implementation by Cody (AI coding specialist)\nFramework designed by Broedkrummen\nBuilt with OpenClaw agent-orchestrator\n\nLast Updated: 2026-02-25 | Version 1.3.0 (Speed-First)\n\nCloud Integration (Requires Setup)\n\nThe system supports optional cloud backup and sync:\n\nSupermemory Integration: Push memories to cloud for cross-device access\nMem0 Auto-Fact Extraction: Automatic fact extraction from conversations (80% token reduction)\n\nConfigure via environment variables:\n\nSUPERMEMORY_API_KEY - For cloud backup\nMEM0_API_KEY - For auto-fact extraction\nSpeed Optimizations (v1.0.5)\nOptimization Techniques Implemented\nTechnique\tLayer\tComplexity\tBenefit\nBloom Filters\tPre-query\tO(1)\tSkip expensive queries\nRedis Hot Cache\tL0\t<1ms\tSub-millisecond access\nMem0 L1 Cache\tL1\t<10ms\t80% token reduction\nParallel Queries\tAll\tO(1) wall\tConcurrent tier queries\nConnection Pooling\tChromaDB\tReuse\tNo connection overhead\nBinary Search\tGit-Notes\tO(log n)\tFast sorted lookups\nPre-computed Embeddings\tCache\tSkip compute\tCache hits = instant\nLazy Loading\tFiles\tOn-demand\tReduced memory footprint\nPre-fetch Context\tPredictive\tAnticipate\tResults ready before ask\nResult Caching\tTTL\t1-5min\tAvoid redundant queries\nL1 Cache (Mem0)\nPurpose: First-layer cache for 80% token reduction\nHow: Mem0 extracts facts from conversations automatically\nBenefit: Reduces context window usage while preserving key information\nParallel Tier Query\nPurpose: Query all memory tiers simultaneously\nHow: Async queries to Mem0, ChromaDB, Git-Notes, and file search\nBenefit: O(1) wall-clock time instead of sequential O(n) tier traversal\nRedis Hot Cache (L0)\nPurpose: Ultra-fast L0 cache for frequently accessed memories\nTTL: 5-15 minutes for hot data\nBenefit: Sub-millisecond access for top results\nResult Caching with TTL\nPurpose: Cache search results to avoid redundant queries\nTTL: 1-5 minutes depending on tier\nBenefit: Dramatically reduces API calls and computation\nBinary Search (Git-Notes)\nPurpose: O(log n) lookup in sorted memory index\nHow: Maintain sorted timestamp/index files\nBenefit: Fast retrieval from large Git-Notes collections\nConnection Pooling\nPurpose: Reuse ChromaDB and Ollama connections\nHow: Persistent connection pools with health checks\nBenefit: Eliminates connection overhead on each query\nBloom Filters\nPurpose: Quick existence checks before expensive queries\nHow: Probabilistic filter for memory presence\nBenefit: Skip unnecessary tier searches when result is definitely not present\nPre-fetch Context\nPurpose: Predictive memory loading based on context\nHow: Anticipate likely queries based on current session\nBenefit: Results ready before user asks\nLazy Loading\nPurpose: Load files only when needed\nHow: On-demand loading of large files\nBenefit: Reduced memory footprint and faster initial response\nPre-computed Embeddings\nPurpose: Cache embeddings for frequently queried content\nHow: Store embeddings alongside source data\nBenefit: Skip embedding computation on cache hit\nHow: Store embeddings alongside source data\nBenefit: Skip embedding computation on cache hit\nCloud Architecture (v1.0.5)\nPriority Order\nMem0 (L1 Cache) → ChromaDB → Git-Notes → Supermemory (Backup)\n\nTier\tService\tPurpose\tLatency\tCost\nL0\tRedis\tHot cache\t<1ms\tLow\nL1\tMem0\tAuto-extracted facts\t<10ms\tMedium\nL2\tChromaDB\tSemantic vectors\t<50ms\tLow\nL3\tGit-Notes\tKnowledge graph\t<20ms\tFree\nBackup\tSupermemory\tOffsite backup\tDaily\tFree\nCloud Services Integration\nMem0 (L1 Cache)\nPurpose: First-layer cache for 80% token reduction\nHow: Auto-extracts facts from conversations\nAPI: MEM0_API_KEY environment variable\nBenefit: Reduces context window usage while preserving key information\nChromaDB (Vector Storage)\nPurpose: Semantic similarity search\nEmbeddings: bge-m3 via Ollama\nConnection: Pooled connections for speed\nFallback: Keyword search if unavailable\nGit-Notes (Knowledge Graph)\nPurpose: Structured JSON storage\nLookup: Binary search O(log n)\nSync: Git-based versioning\nSupermemory (Cloud Backup)\nPurpose: Daily backup only (not real-time sync)\nFrequency: Once per day\nAPI: SUPERMEMORY_API_KEY environment variable\nBenefit: Reduces API calls while maintaining offsite backup\nEnvironment Variables\n# Required for cloud features\nMEM0_API_KEY=your_mem0_key # Auto-fact extraction\nSUPERMEMORY_API_KEY=your_key # Cloud backup\n\n# Optional overrides\nCHROMA_URL=http://localhost:8100 # ChromaDB server\nOLLAMA_URL=http://localhost:11434 # Ollama server\nEMBEDDING_MODEL=bge-m3 # Embedding model\n\nSearch Priority Flow (v1.0.5)\nQuery Input\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 1. BLOOM FILTER CHECK (O(1)) │\n│ • Probabilistic existence check │\n│ • Skip expensive queries if definitely not present │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 2. REDIS HOT CACHE / L0 CACHE (Sub-millisecond) │\n│ • TTL: 5-15 minutes │\n│ • Frequently accessed memories │\n│ • Return immediately if cached │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 3. MEM0 L1 CACHE (First Priority) │\n│ • Auto-extracted facts (80% token reduction) │\n│ • Fast fact lookup │\n│ • No embedding computation needed │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 4. CHROMADB (Second Priority) │\n│ • Semantic vector search (bge-m3 embeddings) │\n│ • Connection pooling for speed │\n│ • Return top-k results with scores │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 5. GIT-NOTES (Third Priority) │\n│ • Structured JSON knowledge graph │\n│ • Binary search on sorted index │\n│ • O(log n) lookup time │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ 6. FILE SEARCH (Fallback) │\n│ • Raw grep on daily/diary files │\n│ • Last resort fallback │\n└──────────────────────────────────────────────────────────────┘\n │\n ▼\n┌──────────────────────────────────────────────────────────────┐\n│ RESULTS MERGE & RANKING │\n│ • Combine results from all tiers │\n│ • Apply importance weights (Hippocampus) │\n│ • Apply emotional relevance (Amygdala) │\n│ • Apply value scores (VTA) │\n│ • Return unified ranked results │\n└──────────────────────────────────────────────────────────────┘\n\nCache Strategy Details\nCache Hit: Return cached result immediately (sub-ms)\nCache Miss: Query next tier, cache result with TTL\nNegative Cache: Optionally cache \"not found\" results (shorter TTL)\nCache Invalidation: On session end, new memory add, or manual trigger\n⚠️ Prerequisites & Setup\nRequired Services (must be running)\nChromaDB on http://localhost:8100\nOllama on http://localhost:11434 with bge-m3 model\nOptional Services (require API keys)\nMem0.ai account (for cloud fact extraction)\nSupermemory.ai account (for cloud backup)\nRedis (optional, falls back to in-memory)\nEnvironment Setup\nCopy .env.example to .env\nFill in optional API keys if using cloud features\nRun python3 cli.py --help to get started\nManual Setup for Automation\n\nThe CLI provides commands but cron jobs are NOT auto-installed. To enable:\n\nAdd cron jobs manually via crontab -e\nExample: 0 3 * * * python3 /path/to/cli.py cloud sync\n⚠️ Important Notes\nOn-Import Side Effects\n\nWhen 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.\n\nNo Auto-Installed Cron Jobs\n\nThe skill provides CLI commands for automation but does NOT auto-install cron jobs. You must manually add them if desired:\n\n# Add to crontab -e\n0 3 * * * python3 /path/to/cli.py cloud sync\n\nCloud Features\n\nCloud features (Mem0, Supermemory) require API keys. Set in environment or .env file before use.\n\n🔐 Security & Network Access\nWhen Network Access Occurs\nVariable\tWhen Accessed\tExternal Service\nCHROMA_URL\tIf set\tChromaDB server\nOLLAMA_URL\tIf set\tOllama server\nMEM0_API_KEY\tIf set AND MEM0_USE_LOCAL=false\tMem0.ai API\nSUPERMEMORY_API_KEY\tIf set\tSupermemory.ai API\nREDIS_URL\tIf set\tRedis server\nDefault Behavior (No Network)\nWithout API keys, system runs fully offline\nUses local ChromaDB + local Ollama (if available)\nAll data stored locally in ~/.openclaw/memory/\nCloud Features\n\nOnly enabled when you:\n\nSet MEM0_API_KEY and set MEM0_USE_LOCAL=false\nSet SUPERMEMORY_API_KEY\n\nThese are opt-in only. Default = offline." }, "trust": { "sourceLabel": "tencent", "provenanceUrl": "https://clawhub.ai/Broedkrummen/overkill-memory-system", "publisherUrl": "https://clawhub.ai/Broedkrummen/overkill-memory-system", "owner": "Broedkrummen", "version": "1.9.5", "license": null, "verificationStatus": "Indexed source record" }, "links": { "detailUrl": "https://openagent3.xyz/skills/overkill-memory-system", "downloadUrl": "https://openagent3.xyz/downloads/overkill-memory-system", "agentUrl": "https://openagent3.xyz/skills/overkill-memory-system/agent", "manifestUrl": "https://openagent3.xyz/skills/overkill-memory-system/agent.json", "briefUrl": "https://openagent3.xyz/skills/overkill-memory-system/agent.md" } }