Tools & retrieval
What is Embeddings?
Embeddings are numeric vectors that represent text or other objects for similarity search.
They power retrieval. They do not certify that a passage answers the question.
Why it matters
Embeddings are the numeric representation that makes semantic search possible. They convert text into vectors where similar meanings are close together in space. This is what allows a search system to find "error handling" when the query says "failure recovery."
Embedding quality directly affects retrieval quality, which directly affects RAG quality. The embedding model is a critical choice.
Key takeaways
- 1Embeddings convert text to vectors for similarity computation.
- 2Embedding model choice directly affects retrieval quality.
- 3Similar vectors may not be semantically equivalent — validate results.
Common mistakes
- ✕Using a general-purpose embedding model for domain-specific content without evaluation.
- ✕Assuming embedding similarity equals semantic equivalence.
Related terms
Concept neighborhood
Terms linked from Embeddings in the glossary graph.
- Embeddings
- Vector Database
- Retrieval