AMD Strix Halo local LLM guide: direct 100.0 t/s 30B Qwen MoE on Ryzen AI MAX+ 395 / Radeon 8060S. Setup, benchmarks, raw evidence.

A production-grade Python SDK for the Lemonade LLM backend. Featuring auto-discovery, port scanning, and a low-overhead client architecture. Powering the Sorana AI workspace.

Demo to showcase Strix Halo memory QoS under concurrent AI + CPU workloads (Ollama + CMake + Python).

Hyperspace Vision API, interfaced with ion CI build system (CMake)

Local-first AI PC Agent Skill for private daily memory from configured notes, docs, schedules, DingTalk, Feishu, and Obsidian sources.

Menu-first CLI for inspecting and benchmarking local AI accelerators across CPU, GPU, and NPU

Sample MCP Server for weather queries - Reference implementation for ASUS and OEM partners

Source-available view of Cubecloud OS — a sovereign-intelligence, local-first AI operating layer for private work on your own machines.