AgentDB Vector Search

Semantic search uses vector embeddings to find documents by meaning rather than keyword matching. AgentDB provides HNSW indexing for fast similarity matching, enabling you to index documents with OpenAI or custom embeddings and retrieve contextually relevant results in milliseconds.

Yes. Vector search is core to RAG systems—it retrieves relevant context from your knowledge base to ground LLM responses. AgentDB's hybrid search combines vector similarity with metadata filtering, and MMR diversity ensures you get varied, high-quality context for generation.

Keyword search matches exact terms; vector search captures semantic meaning. Vector embeddings let you find related documents even when they use different words, making it superior for knowledge bases and intelligent document systems where intent matters more than exact phrasing.

No. AgentDB integrates with OpenAI embeddings and supports custom models in Node.js. You pass documents and queries; the Skill handles embedding generation, HNSW indexing, and quantization to deliver sub-millisecond latency at scale.

HNSW (Hierarchical Navigable Small World) is a graph-based indexing algorithm that accelerates nearest-neighbor lookup in high-dimensional spaces. It scales to large vector collections while maintaining sub-millisecond query latency, making it ideal for knowledge bases with thousands or millions of documents.