PointSav Orchestration SLM Chassis

Bootable seL4/Microkit appliance — Yo-Yo broker chassis connecting Totebox Archives to a shared GPU fleet; ships generic, with Tier B brokering disabled until you supply your own license and endpoints

Turns a machine into your fleet's AI broker.

Plataforma

Arranque

This ships as a bootable seL4/Microkit/libvmm appliance image (loader.img), not
a regular Linux binary — boot it under QEMU rather than running it directly.

This build is deliberately generic: no orchestration/Yo-Yo endpoints or license
are baked in. It boots to safe standalone defaults — health, fleet, and
discovery routes are live; Tier B brokering stays disabled until you supply
your own license and endpoint configuration.

Real qemu invocation (adapt paths to wherever you place the downloaded image):

  qemu-system-aarch64 \
    -machine virt,virtualization=on,secure=off \
    -cpu cortex-a53 \
    -serial mon:stdio \
    -device loader,file=./aarch64-unknown-linux-gnu,addr=0x70000000,cpu-num=0 \
    -m size=2G \
    -nographic \
    -global virtio-mmio.force-legacy=false \
    -drive file=./blk_storage,format=raw,if=none,id=hd \
    -device virtio-blk-device,drive=hd,bus=virtio-mmio-bus.1 \
    -device virtio-net-device,netdev=netdev0,bus=virtio-mmio-bus.0 \
    -netdev user,id=netdev0,hostfwd=tcp::9180-:9180

This runs under TCG software emulation (no KVM/hardware-virt required — works
on any host, including non-aarch64 hosts, but slower than native).

Before first boot, create the data disk file referenced above — QEMU will not
start without it:

  qemu-img create -f raw blk_storage 2G

(2 GiB is a real operational minimum, not a placeholder — the upstream example
tooling defaults to 16 MiB, which is a scratch/demo size that fills up almost
immediately under real use.) The guest formats this disk to ext4 itself
automatically on first boot if it's blank, then mounts it directly on every
later boot — your data genuinely persists across restarts as long as you keep
reusing the same blk_storage file. Do not delete or replace it between boots if
you want to keep what's on it.

Once booted, the chassis is reachable at http://127.0.0.1:9180 on the host
(forwarded from the guest).

Known limitation: there is no supported host-side graceful shutdown for this
image today — stopping the VM (Ctrl-C, kill, or equivalent) hard-kills the
emulator process; there is no in-guest ACPI/power-button path to receive a
shutdown request from the host side. Avoid stopping the VM while writes are
actively in flight if you can avoid it.

Versión y suma de verificación

v0.1.0

SHA256: verificando… — verificar via MANIFEST

← Todos los productos