Vercel eliminates embeddings from knowledge agent architecture, reducing latency and infrastructure costs. Builders can now simplify RAG systems with direct LLM approaches.