magicrulerr-art

An Android AI agent that survives long runs on its own silicon — no server, no bill, no mercy for a prompt that won't fit. llama.cpp fork with Adreno OpenCL offload, ~180 on-demand tools, usage budgets, expert sub-agents, cold memory, chained workflows. Kotlin/Compose, Shizuku/Termux/MCP. Built on RikkaHub, ExTV and the Pure line.