DocumentationNeuronDB Documentation

Vector Types in NeuronDB

What Are Vectors?

A vector is a mathematical object represented as an array of numbers. In AI and machine learning, vectors represent data (text, images, audio) in numerical format that computers can process and compare.

Example Vector

[0.234, -0.891, 0.456, 0.123, -0.678]

This is a 5-dimensional vector where each number represents a feature.

Traditional Database vs Vector Database

Traditional Database: Stores structured data: numbers, text, dates. Searches using exact matches or patterns.

Vector Database: Stores numerical representations of data. Searches by semantic similarity and meaning.

Why Use Vectors?

Semantic Search

Find similar items based on meaning, not just keywords. Search for "laptop" and get results for "notebook computer", "portable PC".

Recommendations

Build recommendation systems that suggest related products, content, or services based on similarity.

Anomaly Detection

Identify outliers and unusual patterns in high-dimensional data.

Clustering

Group similar items together for analysis and organization.

Vector Types

vector(n)

Standard dense vector type (float32). Up to 16,000 dimensions. Storage: 4 bytes per dimension.

vector type

CREATE TABLE documents (
  id SERIAL PRIMARY KEY,
  embedding vector(1536)
);

INSERT INTO documents (embedding) VALUES ('[0.1, 0.2, 0.3]');

vectorp

Packed vector with Product Quantization. Compressed (2x-32x smaller). Best for large-scale search and memory optimization.

halfvec

Half-precision vector (float16). 2 bytes per dimension. Best for memory-constrained environments.

sparsevec

Sparse vector representation. Up to 1,000,000 dimensions. Only non-zero values stored. Best for high-dimensional text and TF-IDF vectors.

binaryvec

Binary vector (bit-packed). 1 bit per dimension. Best for Hamming distance and binary embeddings.

vecmap

Sparse vector map type. Stores key-value pairs for sparse vector representations. Efficient for high-dimensional sparse data like TF-IDF vectors, text features, and recommendation systems. Supports sparse vector operations including L2 distance, cosine distance, inner product, and arithmetic operations.

vgraph

Vector graph type for graph-based vector operations. Supports BFS (Breadth-First Search), DFS (Depth-First Search), PageRank algorithms, and community detection. Ideal for recommendation systems, knowledge graphs, and graph neural network applications.

rtext

Retrieval text type that combines retrieval capabilities with text operations. Optimized for RAG pipelines and semantic text search. Provides seamless integration between vector similarity search and full-text search operations.

Usage Examples

Complete example

-- Create table with different vector types
CREATE TABLE documents (
  id SERIAL PRIMARY KEY,
  dense_embedding vector(1536),
  sparse_features sparsevec,
  binary_fingerprint binaryvec
);

-- Insert vectors
INSERT INTO documents (dense_embedding, sparse_features, binary_fingerprint)
VALUES (
  '[0.1, 0.2, ...]'::vector(1536),
  sparsevec_from_map('{"0": 0.5, "100": 0.3, "500": 0.8}'),
  binaryvec_from_array(ARRAY[1,0,1,1,0]::int[])
);

-- Search with different types
SELECT id FROM documents
WHERE dense_embedding <=> '[0.1, 0.2, ...]'::vector < 0.5;

Next Steps

PreviousFeatures
NextDistance Metrics