Streaming Requests & Realtime API in vLLM

Large language model inference has traditionally operated on a simple premise: the user submits a complete prompt (request), the model processes it, and returns a response (either streaming or at…