defilantech / LLMKube
Kubernetes 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.
- Repository stars
- 252All-time #21,712
- Fork
- 43
- Last push
- updated 3 days ago
- Repository created
- since 2025
About
Summarised by AI from the README; the original prevails
Author description: Kubernetes 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.
Basic listing based on official public information; functionality has not been independently evaluated.
- Platform
- Not verified
- Pricing
- Pricing not verifiedCheck current official software, hosting, model and infrastructure pricing.
- Maintenance notes
- Not verified