Memory Management

This is AgentDB's vector memory system with HNSW search that claims 150x to 12,500x faster pattern retrieval than whatever baseline they're comparing against. You get persistent storage for successful patterns, semantic search across past work, and the ability to share knowledge between agents. The workflow is straightforward: check memory before starting tasks, store what worked, consolidate periodically. It's overkill for one-off scripts but makes sense if you're running agents that should learn from previous sessions or need to avoid solving the same problem twice. The CLI covers the basics with store, search, and export commands, plus you can initialize HNSW indexing separately if you want the speed boost.