Building a Voice RAG System Under a 200ms Latency Budget

What happens when you combine Voice AI + RAG + multilingual retrieval, but give yourself a strict...