natuleadan pgvector Embeddings
When to use
When implementing semantic search, recommendation systems, or any feature requiring vector similarity in PostgreSQL.
References
| Topic | File |
|---|---|
| Embedding generation | references/embedding-generation.md |
| HNSW index setup | references/hnsw-index.md |
| Cosine similarity search | references/similarity-search.md |
| Hybrid search patterns | references/hybrid-search.md |
Quick checklist
- Install pgvector extension:
CREATE EXTENSION vector; - Define columns with
vector({ dimensions: 384 })or 1536 for OpenAI - Create HNSW index for fast approximate search
- Use cosine similarity (
<=>) for text embeddings - Generate embeddings with OpenAI or deterministic hash fallback
- Search with
WHERE 1 - (embedding <=> $vector) > threshold - JOIN with entities for enriched results