This is exciting news, Perplexity launches an API offering. Their custom llama2 and mistral based inference layer is wicked fast at performing web search augmented generation and produces really impressive results.