anaslimem/CortexaDB ? reverse-engineered prompt

Build me a small embedded database for AI agent memory that feels as easy to use as SQLite, but is made for semantic search and long term recall.

I want to be able to store short facts, connect them with simple relationships, and search them later by meaning, by graph links, and by recency. It should run locally with no server, be fast, durable, and work well as a single file on disk. A Python friendly API would be great, and if it also has a Rust version that would be even better.

Please make the basic flow simple, open a database, add memories, connect related items, and run a fluent query to get the best matches back. If you need to, look up current docs online for the embedding and search parts. Focus on something that feels polished, lightweight, and practical for building AI agents that need reliable memory.