LadybugDB for metagraph and agentic memory

**For two years, I searched for a graph database that could run on a user's device.** I tried everything. SQLite with graph schemas. DuckDB. Custom solutions. The result of that search became my first book — building knowledge graphs in pure relational databases, with vector extensions, because nothing better existed for the embedded case. Then Kuzu appeared. A full-featured embedded graph database, native Cypher, columnar storage, real performance. I was following it closely. Then, weeks before I finished writing, the team announced they were discontinuing the project. The chapters I had written about Kuzu never made it into the book. The community picked it up. The most serious fork is **LadybugDB** — and it is not just maintaining Kuzu's codebase. It is actively extending it toward one specific purpose: agentic memory. So I wrote a second book. **The Memory Graph** covers what it actually takes to give an AI agent a proper world model: → Strongly typed property graphs with La