Configuring pgvector in Supabase for Semantic Searches

Semantic search relies on representing textual data as high-dimensional vector embeddings. In this article, we enable the pgvector extension inside a managed Supabase database and write Cosine Similarity queries using Prisma.

Setup Extension

Enable the extension inside your database migrations:

CREATE EXTENSION IF NOT EXISTS vector;

Next, define the vector column on your research documents table:

ALTER TABLE "ResearchDocument" ADD COLUMN embedding vector(1536);

We perform cosine similarity searches using custom database functions triggered via RPC.