yserver

Four jobs on your own machine: ask a question, write a test, fix a bug, add one small function. Runs a local 8B model, one action at a time, and only changes files inside the folder you choose. Apache-2.0.

Run JAX and MuJoCo MJX on AMD Ryzen AI MAX (Strix Halo, gfx1151) from pip wheels - no system ROCm - with a hang-safe check that tells you which wheel set works

Find the failures that stay silent: a hardware watchdog that never loads, a nightly job that fails every night, firmware months behind upstream, a box that froze and nobody was told. Small read-only checks for Linux servers and home labs.

Will this mixture-of-experts model run on my machine, and how fast? Answered from the model's index, before downloading it.

See every hop of a request through your local LLM system: a tracing server, a live flow map, a waterfall per request, and an MCP server for Claude Code and Cursor. One static Go binary.

Run your private repositories' CI on a machine you already own. One command per repo, and it refuses to register against a public repo - where anyone's pull request would run code on your machine.

Can this server lose data? Checks the data, the backup and the restore: backup on the same disk, snapshot missing a database, backup never restore-tested. Consistent SQLite copies, restore test. One static Go binary and two small libraries.

Compare new open-weights AI models with the machine you have: which fit its GPU and memory? Watches publishers on Hugging Face, sizes every GGUF build. Downloads nothing.

yserver: one small computer at home serves open LLMs to Claude Code and my coding tools. The architecture, the flows, the measured numbers and the lessons. AMD Ryzen AI MAX+ 395 (Strix Halo), 128 GB shared memory, ROCm.