Show HN: Nibia Fabric – Turn everyday computers into a shared AI compute cluster

Category: ai-ml

Tags: distributed-inference, llm, llama-cpp, edge-computing, go, local-ai

Score: 6.8/10 (Innovation: 6, Technical: 8, Documentation: 7, Utility: 6)

Nibia Fabric is an experimental Go-based distributed LLM inference system that pools CPU and RAM across LAN-connected Macs, Windows, and Linux machines to run larger GGUF models than any single node could handle. It uses llama.cpp's RPC backend with mTLS-secured worker agents, lazy worker activation, and a local OpenAI-compatible API, making heterogeneous everyday hardware act as a shared compute fabric. It's interesting because it targets a real gap—running larger local models without dedicated GPU clusters—while keeping defaults loopback-only and prebuilt-binary friendly.

Target audience: local-AI enthusiasts, backend developers, homelab and edge-computing users, MLOps engineers exploring distributed inference

Repository: https://github.com/nibia-ai/fabric · Go · Apache-2.0 · 4 stars

View on Hacker News