Thriving Thriving Python PY
โ†‘ 34 ?
astrbot_plugin_memora_connect qa296/astrbot_plugin_memora_connect

A plugin that mimics the biological hippocampus, building a memory network from "concept nodes" and "relationship links". It supports memory formation, retrieval, forgetting, and consolidation, reviews chats using a "bimodal" time distribution, and can save memories periodically.

Translated from Chinese by Claude ยท original on GitHub โ†—

HEALTH ? 82 / Healthy
// solo builder ?

// readme

AstrBot Memora Connect

A hippocampus-mimicking AI memory plugin that gives chatbots persistent memory


Project Overview

Memora Connect gives AstrBot human-like memory. The LLM automatically extracts memories, people, places, and relationships from conversations and writes them into a single SQLite database (partitioned by chat). Before each LLM request, it recalls relevant memories by relevance and injects them.

Conversation โ†’ idle/threshold-triggered LLM analysis โ†’ memories/entities/relations/impressions UPSERT into database
                                            โ†“
LLM request โ†’ keyword + graph association (+ semantic vectors) three-way recall fusion โ†’ context injection

Core Capabilities

๐Ÿง  Memory formationAfter 60s of idle time or once a threshold number of messages accumulates, a single LLM call batch-analyzes the session, producing memories, entities, relations, and impressions
๐Ÿ” Three-way recallKeyword (FTS5 trigram) ยท graph association (entity inverted index + adjacency) ยท semantic vectors (optional, requires an embedding provider)
๐ŸŒ Knowledge graphEntity nodes + edges with semantic relations, each edgeโ€ฆ