Qdrant (vector specialist)

Qdrant is a specialized vector database designed exclusively for storing, indexing, and searching embeddings. Unlike pgvector (a PostgreSQL extension), Qdrant is a dedicated tool optimized for vector operations—faster with millions of vectors, with built-in quantization, filtering, and scaling.

In the semantic audit pipeline, Qdrant is recommended for large datasets (100k+ vectors) where pgvector becomes slow. Qdrant offers a free cloud layer (Qdrant Cloud) and a local version (Docker).

Creating a collection in Qdrant requires defining the vector size (e.g., 768 for Gemini) and distance metric (cosine, euclidean).

For example, an e-commerce service with 500k products → embeddings in Qdrant → nearest neighbors in <100ms. In SEO, Qdrant is excessive for most projects (thousands, not millions of URLs), but becomes necessary for large e-commerce sites and aggregators.

In practice, start with pgvector in Supabase—migration to Qdrant is simple when performance needs actually arise.