lxfight-s-Astrbot-Plugins/astrbot_plugin_livingmemory

What it solves

LivingMemory provides a long-term memory system for AstrBot, enabling AI agents to remember facts and personality contexts across conversations. It solves the problem of "forgetting" by allowing the bot to capture, retrieve, and evolve information over time rather than relying on a limited short-term context window.

How it works

The system breaks down information into "memory atoms" and uses a dual-path retrieval process combining BM25 (keyword) and vector (semantic) search. These memories are organized into a graph structure where relationship confidence changes based on evidence. The system employs a reinforcement mechanism to strengthen important memories and a decay/TTL (Time-To-Live) mechanism to phase out outdated or low-value information. It also provides native agent tools for recalling and memorizing information.

Who it’s for

Users of the AstrBot framework who want to give their AI bots persistent, evolving memory and the ability to manage that memory through a visual dashboard.

Highlights

• Hybrid Retrieval: Combines keyword and semantic search for more accurate recall.
• Dynamic Context: Memory atoms feature importance levels, TTL, and time-decay mechanisms.
• Visual Memory Graph: A high-performance canvas for exploring and managing memory relationships.
• Granular Access Control: Supports memory scoping by session, user, or global sharing with whitelist and identity alias support.
• Safety Features: Includes automatic backups, transactional deletions, and shadow index switching for safe index reconstruction.