AMD Strix Halo / Ryzen AI Halo local LLM setup and benchmark guide for Ryzen AI MAX+ 395 and Radeon 8060S: Ollama, llama.cpp Vulkan/RADV, ROCm, 101 t/s Qwen3-Coder, CHADROCK MTP, 120B GGUF, and raw evidence.

FrameworkDotnet is a .NET library for talking to Framework laptop and desktop embedded controller APIs through the native framework-system Rust FFI layer.

Known-good local LLM inference on the Framework Desktop (Ryzen AI Max+ 395 / Strix Halo)

An exercise in futility and wasting electricity: this repo contains an Unsloth build environment for training a finetuned Gemma 4 LLM for home-assistant tool calls. Built with AMD ROCm on a Framework Desktop using Nix package manager + UV for a reproduceable build environment.

Standalone native FFI facade for framework-system, intended for .NET interop via csbindgen.

Reproducible ROCm containers for the Framework Desktop (Strix Halo / Radeon 8060S, gfx1151) — pinned PyTorch & JAX images, GPU smoke tests, and measured attention/training performance findings

Manual per-fan control + live temps/RPM for the Framework Desktop (cros_ec) on Linux — GTK4 app with a minimal root daemon.

A collection of NixOS system configurations for different machines.