All projects
1 repositories · updated 2026-11-07
Loading…
1defilantech / LLMKubeKubernetes operator for self-hosted LLM inference across a heterogeneous GPU fleet: NVIDIA CUDA, AMD Vulkan, and Apple Silicon Metal. Runtimes: llama.cpp, vLLM, TGI, mlx-server. Multi-GPU sharding, model caching, OpenAI-compatible endpoints. Apache-2.0, run across homelab and on-prem fleets, actively developed.GoApache-2.0 · PermissiveInsufficient rank history251
