Aurabase Logo
aurabasedocs
Aurabase Logo
aurabasedocs
API/AI Gateway/Vector Embeddings

Vector Embeddings

POST/v1/ai/{project_id}/embed

Computes text embeddings and stores them in dimension-isolated PostgreSQL columns (768, 1536, 3072).

#
Request

Parameters

ParameterTypeRequiredDescription
textstringyesText to embed and persist.
namespacestringyesTarget partition namespace isolating document spaces.
metadataobjectnoArbitrary JSON metadata attached to the vector record.
#
Architecture

Multi-Dimension Storage

Info

Aurabase provides multi-dimension vector routing with dedicated columns:

  • 768: Gemini text-embedding-004 (stored in embedding_768).
  • 1536: OpenAI text-embedding-3-small (stored in embedding_1536).
  • 3072: OpenAI text-embedding-3-large (stored in embedding_3072 using halfvec for HNSW indexing).

Vectors are tagged with the active embedding_model to prevent cross-model semantic drift.

v1.0 · revised Oct 31, 2026