← Glossary/Architecture

Defined term

Embeddings

Numerical vectors that represent the meaning of a text, image, or other piece of content.

Embeddings are dense vector representations produced by a model trained to map semantically similar inputs to nearby points in vector space. They power semantic search, retrieval, clustering, classification, and deduplication. Embedding choice (model family, dimension, normalization) and indexing strategy directly drive retrieval quality in RAG systems.

When it matters

When you need to find semantically similar content from a large corpus. Required for any RAG system, recommendation engine, or duplicate detection workflow at scale.

Real example

A support knowledge base with 12,000 articles. We embed each article into a 1536-dim vector (OpenAI text-embedding-3) and store in pgvector. Each customer query is embedded and matched against the corpus using cosine similarity, returning top-5 in under 50ms.

KPIs to watch

Embedding precision@5 on labelled queries (>0.75 target), embedding refresh cycle (weekly for active corpus), corpus coverage (>95% of expected query topics).

Related terms

See it in action

We use this every week

Book a 30-min call and we'll walk you through how Embeddings shows up in a real engagement we're running.

Book a 30-min call