AgentDB Performance Optimization

What it does

This skill documents concrete performance optimizations for AgentDB vector databases. It explains quantization strategies (binary, scalar, product), HNSW index tuning, in-memory caching, batch insert/retrieve patterns, memory consolidation and pruning, and monitoring recipes to achieve much faster search and smaller storage at acceptable accuracy trade-offs.

When to use it

Use these techniques when scaling vector stores from thousands to millions of vectors, when memory is constrained (edge/mobile), or when search latency must be reduced for real-time agent reasoning. Also apply before large migrations or when monitoring shows high search latency or memory pressure.

What's included

• Scripts: none bundled; includes code snippets and configuration examples (TypeScript/Node)
• References: pointers to quantitative benchmarks and papers (docs/quantization-techniques.pdf, HNSW references)
• Instructions: stepwise recipes for quantization selection, HNSW parameter tuning, cache sizing, batch operations, and troubleshooting common issues.

Compatible agents

Useful for agents that can run Node.js tooling, edit code/config, and deploy database changes. Targets developer-facing automation (Copilot, Codex, CLI-capable agents).