MACHETE DGX DGX Spark operations PL Appendix Page 001
Job 1.0.0Sheet 1 of 5Printed on demand

Your Spark answers the phone.

Machete DGX opens a direct SSH connection from your phone to the NVIDIA DGX Spark on your desk. Telemetry, terminal, containers and the model running on your own hardware, with nothing of ours in between.

Version 1.0 is in review at Apple and Google. The links open the listings as soon as each store approves it.

GPUutilisation, temperature, power draw and the cost of that power0 % · 44 °C · 10.7 W
Memoryunified across CPU and GPU, with what is resident in VRAM right now122 GiB
Diskfree space before the next model download fails halfway1.8 / 3.5 TiB
CPUload across all cores, sampled continuously20 cores
Uptimehow long the machine has been up, which answers most questions on its own7 weeks

Actual readout Values printed above were taken from a running DGX Spark, not composed for this page.

Section 2CapabilitiesSheet 2 of 5

Everything the machine does, legible from a phone.

One connection carries all of it. There is no separate agent to babysit for the basics. The optional on-server agent exists only for the work that must continue while your phone is asleep.

ALive dashboard. GPU load, temperature, power and its cost, unified memory, disk, cores, network, uptime. What is resident in VRAM, with an action to unload it.
BTerminal. A real interactive SSH shell, with the keyboard row a phone is missing.
CDocker. List, inspect, start and stop containers.
DChat. Talk to the model already serving on your Spark. Benchmark it: time to first token, tokens per second.
EModels. Search Hugging Face, download to the box, serve it. Community recipes that are ready to run.
FImages. Install ComfyUI in a container, generate, then shelve it so it stops holding GPU memory.
GTraining. Watch a run and get told when it finishes, stalls, or the loss stops falling.
HAlerts and backups. Thresholds by push and e-mail, daily reports, scheduled snapshots to S3, Google Drive, Dropbox, Box, FTP or SCP.
Dashboard showing GPU utilisation, temperature and power draw.
Plate A Dashboard
Interactive SSH terminal session running on the Spark.
Plate B Terminal
Docker container list with status and resource use.
Plate C Docker
Chat with the language model served locally on the Spark.
Plate D Chat
Section 3The connectionSheet 3 of 5

Nothing of ours sits in the path.

Most tools that show you a machine put a relay in the middle, and your credentials with it. This one dials your machine directly, over your own network or VPN.

What we never receive

SSH credentials, private keys, terminal output, container data, GPU metrics, and every prompt and answer exchanged with the model on your Spark. None of it is transmitted to us, because none of it passes through us.

What an account holds

Only if you create one: your e-mail, and the settings worth having on a second device: server addresses and usernames, themes, alert rules, backup plans. Stored in Supabase in the EU. No analytics, no advertising, no trackers.

Read the privacy policy for the full accounting, field by field.

Section 4ProSheet 4 of 5

Free to run your machine. Pro for the heavy end.

The dashboard, terminal, Docker and chat are the product, not a trial. Pro is a subscription that opens the parts that do sustained work.

Model downloadspull weights from Hugging Face straight onto the Spark and serve themPro
Image generationinstall, run and shelve ComfyUI from the appPro
Automationalert rules, scheduled backups, daily and weekly reportsPro
Multiple serversmore than one machine on the same accountPro
Training panelrun monitoring with finish, stall and loss alertsPro

The App Store and Google Play show the current price in your own currency, including any regional difference. Subscriptions renew until cancelled; cancel any time in your store account. See the Terms of Use.

Hugging Face model search inside the app.
Plate E Models
Alert rules with thresholds and notification channels.
Plate H Alerts
Scheduled backup plans to cloud storage targets.
Plate H Backups
Section 5AppendixSheet 5 of 5

Appendix

Documents

Privacy policy
Terms of Use
Support
Delete your account

Get the app

App Store
Google Play
Ta strona po polsku

Requirements

An NVIDIA DGX Spark, or any Linux host you can reach over SSH. GPU metrics need nvidia-smi; container features need the login user in the docker group.

Published by

SLOVE Sp. z o.o.
ul. Kwiska 194, 54-210 Wrocław, Poland
NIP 8943264082 · KRS 0001182349
[email protected]