Overview

Learn the core fundamentals of Actian VectorAI DB.

This section covers the core fundamentals for working with Actian VectorAI DB.

Overview

VectorAI DB organizes data in a hierarchical structure. Collections contain points, and each point consists of a vector with optional payload metadata. This structure enables semantic search: you query with a vector, and VectorAI DB finds points with similar vectors in your collection.

Category	Component	Description
Data structure	Collections	Named containers that store points, similar to tables in relational databases. Each collection has a fixed vector dimension and distance metric.
Data structure	Points	Individual data units within a collection. Each point has a unique ID and contains a vector with optional metadata.
Data structure	Vectors	Numerical embeddings that represent your data semantically. Generated by embedding models from text, images, or other content.
Data structure	Payload	Optional JSON metadata attached to points. Use payloads for filtering and storing contextual information.
Operations	Search	Vector similarity search using distance metrics to find semantically similar content.
Operations	Hybrid search	Combine dense and sparse vector search to improve recall and relevance.
Operations	Filtering	Combine vector similarity with metadata conditions using must, should, and must-not filters.
Operations	Hybrid search	Combine dense and sparse vector search with fusion strategies such as RRF and DBSF for improved retrieval quality.
Configuration	Indexing	HNSW algorithm for efficient approximate nearest neighbor search at scale.
Configuration	Distance metrics	Cosine similarity, Euclidean distance, and dot product for measuring vector similarity.

Next steps

• Start with Collections to understand how data is organized.
• Learn about Points to manage your vector data.
• Explore Search to query your collections.