The code · github.com/fthrvi/nakshatra

nakshatra

Distributed LLM inference across heterogeneous workers (NVIDIA / AMD / Apple Silicon / CPU). Splits one model by layer ranges; patched llama.cpp + gRPC chain protocol.