# Rune > Persistent AI memory system with intelligent context injection and adaptive learning - Skill: `thebobloblaw/rune` (Agent Skill, multi-file: 29 files) - Install (CLI): `npx skillmds@latest add thebobloblaw/rune` - Raw SKILL.md: https://api.skillmd.com/api/skills/thebobloblaw/rune/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: TheBobLoblaw (https://skillmd.com/u/thebobloblaw) - Updated: 2026-09-10 - Page: https://skillmd.com/skills/thebobloblaw/rune --- # Rune - Persistent AI Memory System **Rune** provides persistent memory for OpenClaw agents using SQLite storage, intelligent fact retrieval, and adaptive context injection. ## Features ### Memory Management - **Dynamic Context Injection**: Selects relevant facts for each conversation - **Access Pattern Learning**: Prioritizes frequently used facts - **Memory Consolidation**: Merges similar facts and compresses verbose ones - **Temporal Queries**: Time-based fact retrieval ### Intelligence Features - **Interaction Style Detection**: Learns conversation preferences over time - **Behavioral Pattern Analysis**: Tracks work patterns and preferences - **Proactive Memory**: Surfaces relevant context when appropriate ### Project Management - **Task Recommendations**: Suggests next steps based on project state - **Blocker Detection**: Identifies projects that need attention - **Project Health Scoring**: Quantifies project progress and health ### Notification System - **Priority Classification**: Categorizes notifications by importance - **Smart Timing**: Respects user preferences for notification timing - **Channel Routing**: Routes notifications to appropriate channels ## Naming Convention - **Skill name**: `rune` (ClawHub package) - **CLI command**: `rune` (terminal interface) - **Project name**: Rune (overall system) ## Installation Requirements ### System Dependencies - Node.js v18+ with npm - SQLite3 (included via better-sqlite3 npm package) - Optional: Ollama for local LLM features ### Installation Process The installation process will: - Create directories under `~/.openclaw/` - Install the `rune` CLI globally via npm - Create SQLite database at `~/.openclaw/memory.db` - Modify existing `HEARTBEAT.md` if present ### Before Installing - Back up important configuration files - Review package.json dependencies if security is a concern - Consider using local models (Ollama) instead of cloud APIs ## Installation Methods ```bash # Via ClawHub (recommended) clawhub install rune # Manual installation git clone https://github.com/TheBobLoblaw/rune cd rune npm install --production npm install -g . ``` ## Basic Usage ```bash # Check system status rune stats # Add facts rune add person cory.name "Cory - my human user" rune add project website.status "In progress" # Search facts rune search "Cory" # Generate context rune context "Let's work on the website" # Task management rune next-task rune project-state website # Maintenance rune consolidate rune expire ``` ## Core Commands ### Fact Management - `rune add ` - Store a fact - `rune get [key]` - Retrieve facts - `rune search ` - Full-text search - `rune remove ` - Delete facts ### Context Generation - `rune context ` - Generate relevant context - `rune inject` - Create context file - `rune score ` - Score fact relevance - `rune budget ` - Context within token limits ### Project Management - `rune project-state ` - Track project status - `rune next-task` - Get task recommendations - `rune stuck-projects` - Find blocked projects ### Intelligence Features - `rune session-style ` - Detect interaction style - `rune pattern-analysis` - Analyze behavioral patterns - `rune self-review` - Generate improvement insights ### Maintenance - `rune consolidate` - Optimize memory storage - `rune expire` - Remove expired facts - `rune stats` - Show database statistics ## Architecture - **Database**: SQLite with FTS5 full-text search - **LLM Integration**: Supports Ollama (local), Anthropic, OpenAI - **Storage Location**: `~/.openclaw/memory.db` - **Context Output**: Generates markdown for LLM consumption ## Memory Categories - **person**: Individual information (names, roles, preferences) - **project**: Project status, phases, decisions - **tool**: Tool usage patterns and configurations - **lesson**: Learning from past experiences - **decision**: Record of choices and reasoning - **preference**: User settings and preferences - **environment**: System configuration (non-sensitive only) ## Integration with OpenClaw ### Heartbeat Integration Add these lines to `HEARTBEAT.md` for automated maintenance: ```markdown ## Memory Maintenance - Memory expiration: `rune expire` - Context regeneration: `rune inject --output ~/.openclaw/workspace/FACTS.md` - Consolidation: `rune consolidate --auto-prioritize` ``` ### Session Integration The skill provides session hooks that integrate with OpenClaw's session lifecycle. For manual integration, refer to the provided session handler scripts in the installation. ## Security & Privacy ### Data Storage **What Rune Stores:** - Facts explicitly added via commands - Session interaction patterns (style, not content) - Project states and task information **What Rune Does NOT Store:** - Full conversation transcripts (unless explicitly extracted) - API keys or credentials - Sensitive personal information (unless explicitly added) ### Security Measures - Input sanitization for all user-provided data - Parameterized database queries to prevent injection - No execution of user-provided shell commands - Session handlers validate and limit input ### Privacy Options - **Local-only operation**: Use Ollama for complete offline functionality - **Cloud API optional**: OpenAI/Anthropic APIs are optional for advanced features - **Fact review**: Regular review of stored facts recommended - **No credential storage**: Use environment variables for sensitive data ### Installation Security - Review package.json dependencies before installation - Installation runs standard npm lifecycle scripts - Consider isolated environment for high-security deployments - Session hooks process metadata only, not full content ## Automated Maintenance ### Recommended Cron Jobs ```bash # Daily maintenance (3 AM) 0 3 * * * /usr/local/bin/rune expire && /usr/local/bin/rune inject --output ~/.openclaw/workspace/FACTS.md # Weekly optimization (Sunday 2 AM) 0 2 * * 0 /usr/local/bin/rune consolidate --auto-prioritize && /usr/local/bin/rune self-review --days 7 # Monthly deep clean (1st of month, 1 AM) 0 1 1 * * /usr/local/bin/rune pattern-analysis --days 30 && sqlite3 ~/.openclaw/memory.db "VACUUM; ANALYZE;" # Database backup (daily 4 AM) 0 4 * * * cp ~/.openclaw/memory.db ~/.openclaw/memory.db.backup.$(date +\%Y\%m\%d) 5 4 * * * find ~/.openclaw -name "memory.db.backup.*" -mtime +7 -delete ``` ### Benefits of Automation - Memory optimization prevents database bloat - Regular fact consolidation improves query performance - Pattern analysis enables continuous improvement - Automated backups protect against data loss ## Performance Monitoring ```bash # Check database health rune stats # Review consolidation opportunities rune consolidate --dry-run # Analyze patterns rune pattern-analysis --days 7 ``` ## Troubleshooting **Large database size:** - Run `rune consolidate` to merge similar facts - Use `rune expire` to remove old working memory - Check `rune stats` for database metrics **LLM integration issues:** - Verify Ollama status: `systemctl status ollama` - Test connectivity: `rune score "test" --engine ollama` - Check API key configuration for cloud providers **Context generation problems:** - Adjust relevance threshold: `--threshold 0.6` - Use token budgeting: `rune budget "query" --tokens 300` - Review fact relevance with `rune score` ## Contributing Rune welcomes contributions: - **Algorithm improvements**: Better consolidation and scoring - **LLM integrations**: Support for additional providers - **Performance optimization**: Query and storage efficiency - **Documentation**: Usage examples and guides ## License MIT License --- *Persistent memory for intelligent AI assistants.*