DuckDB Knowledge Memory Server

Example Use Cases

• 1

  A persistent, queryable memory store for entities, relations, and observations using an embedded SQL database (DuckDB).

• 2

  Efficient full- and fuzzy-search across large knowledge-graph-style memories, combining SQL filtering with Fuse.js ranking.

• 3

  Transactional integrity, indexing, and scalable memory performance over a JSON/in-memory approach.

Description

This MCP server replaces the original JSON-backed Knowledge Graph Memory Server with a DuckDB-backed relational store, preserving entities, observations, and relations in a SQL schema. It combines DuckDB queries with Fuse.js fuzzy matching to provide flexible, fast searches, supports transactions and indexing for data integrity and performance, and is intended for use as a persistent memory backend (e.g., with Claude Desktop via Smithery). The project is a community fork that emphasizes scalability and query flexibility for large memory datasets.