memory-consolidation

# Memory Consolidation Cross-session learning through periodic memory optimization, inspired by how biological sleep consolidates memories. ## When to Run | Frequency | Type | Duration | Purpose | |-----------|------|----------|---------| | Daily | Light | ~2 min | Health check, quick cleanup | | Weekly | Standard | ~10 min | Full consolidation | | Monthly | Deep | ~30 min | System-wide optimization | ## Consolidation Phases ### Phase 1: Pattern Detection Identify recurring patterns across memories: - Similar problems solved multiple ways - Frequently co-accessed memories - Successful strategies repeated ### Phase 2: Schema Formation Create higher-level abstractions: - "When X happens, Y usually works" - Merge similar memories into schemas - Reduce redundancy ### Phase 3: Conflict Resolution Detect and resolve contradictions: - Same problem, different solutions stored - Outdated information conflicting with new - Decide which to keep based on recency + success ### Phase 4: Temporal Decay Apply time-based forgetting: - Unused memories decay in Q-value - Very old, never-accessed memories pruned - Keep core knowledge stable ### Phase 5: Cross-Context Transfer Share useful patterns between contexts: - Debugging insight useful for coding - Research finding applicable to implementation - Generalize specific solutions ### Phase 6: QAVR Sync Update Q-value rankings based on consolidation: - Boost patterns with high success - Demote patterns that didn't work - Rebalance context weights ## Output Report ```markdown ## Memory Consolidation Report **Date**: [timestamp] **Type**: [Light/Standard/Deep] **Duration**: [X minutes] ### Health Metrics - Total memories: X - Active (accessed 30d): X - Dormant (no access 30d): X - Conflicts detected: X - Patterns identified: X ### Actions Taken - Memories merged: X - Memories pruned: X - Schemas created: X - Conflicts resolved: X - Cross-context transfers: X ### Top Patterns Identified 1. [pattern description] 2. [pattern description] 3. [pattern description] ### Recommendations - [suggestion for improvement] ### System Health Score [X/100] ``` ## Usage ### Manual Invocation ``` /memory-consolidation daily # Quick health check /memory-consolidation weekly # Standard consolidation /memory-consolidation monthly # Deep optimization ``` ### Automated Scheduling Configure in clawdbot.json: ```json { "memory": { "consolidation": { "daily": "03:00", "weekly": "sunday 04:00", "monthly": "1st 05:00" } } } ``` ## Benefits 1. **Reduced Noise**: Old, unhelpful memories pruned 2. **Better Retrieval**: Patterns surface useful info 3. **Conflict-Free**: No contradictory advice 4. **Cross-Pollination**: Insights shared across domains 5. **Efficiency**: Less storage, faster queries