Embeddings
Numeric vector representations of text that capture semantic meaning, the foundation of semantic search, RAG, and most modern NLP applications.
Quick answer
What is Embeddings?
Numeric vector representations of text that capture semantic meaning, the foundation of semantic search, RAG, and most modern NLP applications.
Last updated:
An embedding is a list of numbers (typically 384, 768, 1536, or 3072 dimensions) that represents a piece of text in a high-dimensional space. Texts with similar meanings end up near each other in this space, regardless of exact word overlap, "purchased shoes online" and "ordered sneakers from a website" produce nearby embeddings.
Embeddings power semantic search: instead of matching keywords, search becomes "find vectors near this query vector." In Salesforce contexts, this is how RAG systems retrieve relevant Knowledge Articles, prior Case resolutions, or related Opportunities, by comparing the embedding of a new question to a vector database of embedded company content.
Common embedding models: OpenAI's text-embedding-3-small/large, Cohere's embed, open-source alternatives like sentence-transformers. The choice affects search quality, cost, and infrastructure (some embeddings can run on-device, others require API calls).
Related terms
Browse all terms- Vector DatabaseA database optimized for storing and querying high-dimensional vectors (embeddings), the storage layer that makes semantic search and RAG fast at scale.
- RAG (Retrieval-Augmented Generation)An LLM is given relevant retrieved documents as context before generating a response, grounding outputs in your specific data, not just the model's training.