Large Language Models
Embeddings
Meaning as coordinates. The representation that makes search, clustering and RAG possible.
Grasp
An embedding maps a piece of text to a fixed-length vector of numbers, positioned so that texts with similar meaning land near each other. "How do I reset my password" and "I forgot my login" end up close together despite sharing almost no words.
That single property is what makes semantic search possible, and semantic search is what makes retrieval-augmented generation possible. Keyword search matches strings; embedding search matches meaning, which is what users actually want most of the time.
The practical decisions are unglamorous and they matter more than the theory. Which embedding model - they differ substantially by domain and language. What dimensionality - larger is not automatically better, and it multiplies your storage and query cost. And critically: the same model must embed both your documents and your queries, or the geometry is meaningless. Changing embedding models means re-embedding your entire corpus.