How does vector search and embeddings work in Data Cloud for Agentforce?

Data Cloud supports storing and querying vector embeddings to power semantic search and AI grounding. Key concepts:

1. Embeddings — numerical vector representations of text (articles, emails, product descriptions) generated by an embedding model (e.g., OpenAI embeddings or Salesforce's own model).
2. Vector Search — instead of exact keyword matching, vector search finds semantically similar content based on cosine similarity between query and stored vectors.
3. Chunking — long documents are split into smaller text chunks before embedding to fit model token limits; each chunk is stored as a separate vector.
4. Grounding for Agentforce — when an Einstein/Agentforce agent needs to answer a question, Data Cloud's vector store is queried to retrieve the most relevant document chunks (RAG — Retrieval-Augmented Generation), which are included in the LLM prompt context for accurate, grounded responses.