pplx-embed

Perplexity Engineering detailed its GPU embedding serving stack for pplx-embed, revealing that it reuses prefill and decode kernels from its LLM engine rather than building a separate embedding engine…