Supabase

Supabase is a popular open-source Firebase alternative built on Postgres, and its vector story is simple: every Supabase project can enable pgvector, the Postgres extension that adds a vector column type, similarity operators, and ANN indexes. That means your relational data, auth, storage, and embeddings all live in one Postgres database — vector search becomes just another SQL query.

For a working developer this is often the most pragmatic choice on the list. You skip an entire piece of infrastructure: no separate vector service to provision, no data sync, no second billing relationship. Embeddings sit in the same tables as your users and documents, so joins, foreign keys, transactions, and row-level security all apply to vector data — filtering search results to “only rows this user may see” is a WHERE clause plus RLS, not application logic. For most products whose corpus is thousands to a few million vectors, pgvector with an HNSW index is plenty fast, and you only outgrow it at scales where a dedicated engine earns its complexity.

In practice you’ll run create extension vector, add a column like embedding vector(1536), insert embeddings from your model of choice, and index with create index on docs using hnsw (embedding vector_cosine_ops). Queries order by distance: order by embedding <=> $1 limit 5, typically wrapped in a Postgres function you call from supabase-js via .rpc(). Supabase’s docs include ready-made RAG patterns, and Edge Functions can handle the embedding step.