Reference · Glossary

Vector

Last updated

A **vector** is an ordered list of numbers. In AI, embeddings are vectors that place meaning in a space where similar items are close together.

#When to use

Similarity search, clustering, recommendations, and RAG retrieval — anywhere “nearness” should track meaning, not exact strings.

#When not to

Exact ID lookups, keyword legal matches, or tiny feature sets where a simple key-value store is enough.

#Practical tips

  • Same embedding model at index and query time
  • Normalize or use cosine similarity consistently
  • Dimensionality and model choice affect recall and cost

#Example

"refund policy" → [0.12, -0.44, ...]
"money-back guarantee" → nearby vector
"shipping rates" → farther away

#Learn next