llama.cpp

0d842cb5 - server: support multiple generations from one prompt (OAI "n" option)