Engineering 11 min read· 13 Aug 2026· By Engineering
Embeddings and semantic search — a builder's guide
Model choice, dimensions, and vector DB pairing for practical RAG.
Embeddings turn text into vectors so you can search by meaning. The stack is small but choice-dense.
Model choice#
- text-embedding-3-small — 1536 dims, cheap default
- text-embedding-3-large — 3072 dims, best quality
- gemini-embedding-001 — 768 dims, cheapest per token
Vector store#
For under 1M rows, pgvector in Postgres is more than enough. Over that, look at Pinecone, Qdrant, or Milvus.
Always store the raw text alongside the vector. You'll need it for reranking and display.