Vector Embedding Pipeline

Store text as vector embeddings in Supabase and retrieve semantically similar results using cosine similarity search.

Embedding Model

text-embedding-3-large

Vector Database

Supabase pgvector

Similarity Search

Cosine via match_documents

Text Input OpenAI Embedding Supabase Vector Similarity Match

Store Embedding

Text → OpenAI Embedding → Supabase pgvector

How it works

  1. Your text is sent to text-embedding-3-large
  2. OpenAI returns a 1536-dimension float vector
  3. Vector + content stored in Supabase pgvector table

Semantic Search

Query → Embedding → Cosine Similarity → Results