What problem does it solve? Building semantic search and RAG pipelines requires a vector database that stores embeddings, indexes them efficiently, and retrieves relevant context quickly. This Skill provides CLI commands and TypeScript API patterns for implementing vector search with AgentDB, including HNSW indexing, quantization, and hybrid metadata filtering. ## Core Features & Use Cases - Vector Storage & Search: Initialize databases with configurable dimensions (1536, 768, 384), store documents with embeddings, and run similarity queries with cosine, euclidean, or dot-product metrics. - Memory-Efficient Quantization: Apply binary (32x), scalar (4x), or product (8-16x) quantization to reduce memory footprint for large vector collections. - RAG Pipelines: Retrieve top-k relevant chunks with MMR diversity and inject them into LLM prompts for grounded answers. - Use Case: Build a knowledge base where documents are embedded via OpenAI or sentence-transformers, stored in AgentDB, and queried semantically to provide context for a chatbot answering questions about your documentation. ## Quick Start Initialize an AgentDB vector database with npx agentdb@latest init ./vectors.db, then store document embeddings and run a similarity query to retrieve the top matching documents.
