AI docs · Foundations
Embeddings
Numeric representations of text (or images) that place similar meanings close together, enabling search and retrieval by meaning.
What it is
- An embedding is a vector (a list of numbers) that represents the meaning of a piece of content.
- Similar content has similar vectors, so you can measure semantic similarity with simple math.
How it works
- An embedding model converts text into a vector.
- Vectors are stored in a vector database; a query is embedded and the nearest vectors are retrieved.
- This powers semantic search and is the retrieval half of RAG.
Trade-offs
- Captures meaning better than keyword search, but quality depends on the embedding model and the data.
- Adds infrastructure (a vector store) and a pipeline to keep embeddings fresh.
When to use it
- Semantic search, recommendations, deduplication, and grounding LLMs in your own data.
Common pitfalls
- Stale embeddings when the source data changes.
- Assuming semantic similarity always equals relevance for the user's intent.