openclaw-auto-dream

What problem does it solve?

AI agents suffer from amnesia after each session ends, losing valuable context accumulated over weeks of interaction. Daily logs pile up unsorted and unscored, preventing the agent from learning from history or connecting disparate pieces of information. Auto-Dream fixes this by giving agents a cognitive memory system that periodically consolidates raw logs into structured long-term memory, complete with importance scoring, knowledge graphs, and health dashboards.

Core Features & Use Cases

• Cognitive Dream Cycles: Automatically scans daily logs, extracts key insights, and consolidates them into five memory layers on a configurable cron schedule.
• Intelligent Memory Management: Applies importance scoring with forgetting curves, semantic deduplication, and graceful archival of stale entries to keep memory fresh and relevant.
• Knowledge Graph & Health Monitoring: Builds a linked knowledge graph with reachability metrics and provides an interactive HTML dashboard showing memory health scores, trends, and insights.
• Cross-Instance Migration: Export and import memory bundles between MyClaw instances to clone or merge agent personas.

Quick Start

Tell your OpenClaw agent to install Auto-Dream and run a dream cycle to consolidate your daily logs into structured long-term memory.