Three modes
Chat only; Copilot, which keeps watch and suggests fixes; or Autonomous, which fixes minor things itself and tells you.
The LGM-OS AI
Ask it as you would ask a person. Cloudy AI really checks your disks, your backups and your apps before it answers, explains what is going on and, if something is wrong, suggests the fix. It thinks in your home, with no cloud required.
Included in both editions of LGM-OS, with no extra fee.
You choose how much it does on its own. Anything important always waits for your yes.
Chat only; Copilot, which keeps watch and suggests fixes; or Autonomous, which fixes minor things itself and tells you.
For each group of tasks you choose “do it yourself”, “ask me” or “never”. Serious things always ask.
Before changing anything it shows you exactly what it will do. You approve or reject it with one tap.
Everything goes into the History, with who asked for it, a verifiable receipt and, where possible, a button to undo it.
One conversation for the whole NAS, and for things that have nothing to do with it too.
Disks, space, backups, apps, cameras and virtual machines: it checks before answering and tells you what it saw.
“What errors were there today?” gets the warnings from the whole NAS, grouped, and a single question diagnoses an app or service with the fix ready.
Have it check something every day, every week or when an alert fires. With “only if…” it doesn’t start unless needed, and it can also start when a file lands in a folder.
It remembers what you tell it and the fixes that worked, and finds them even if you put it another way.
Ask about your paperwork and every answer says which document and which page it comes from. Group folders into knowledge spaces.
Tables and comparisons next to the chat, and live notes such as the shopping list or a NAS logbook.
It opens a page, looks at it, clicks and fills things in. When it meets a captcha or a code, it asks you to take over.
Telegram, WhatsApp, Signal, Discord, Matrix and Slack: ask it from there and approve its proposals without opening the panel.
Dictate instead of typing, send it voice notes or say “Cloudy” and your question: it answers out loud.
Add reviewed tools and skills from its store, or teach it how things are done in your home.
Other programs at home can talk to Cloudy AI with their own key: with an expiry date, allowed networks and no power to approve anything.
Open it full screen at your NAS’s /cloudy address, install it as an app on your phone or computer, or use it inside LGM Connect.
Like someone you trust: it looks, acts, checks and tells you what it did.
Everyday requests like “is everything OK?” or “turn off the living room light” are answered without waiting for the model.
Once you approve, it moves on to the next step of what you asked; you’ll see it when you come back, even if you closed the screen.
It breaks long jobs into steps that depend on each other: if one fails, only the ones that relied on it are skipped.
If something fails, it tries another way up to three times and then tells you what it tried.
That the installed app responds, that the virtual machine ends up as you asked and that the recorder really records.
Under every answer you see how long it took and what it did, and in a side column, what it is looking at live.
The same question to the NAS and to your PC or the outside provider, side by side: you keep the best one.
If what you ask could mean two things, it asks you with buttons instead of assuming.
Photos, cameras, lights, machines and accounts: ask in your own words.
Search them by what’s in them or by who appears, or send it a photo on WhatsApp or Telegram and it tells you what’s there.
“Did anything happen in the yard last night?”: it looks at the recordings and tells you in plain words.
Connected to your Home Assistant, it tells you how the house is doing and turns things on or off with a yes.
It finds out why one won’t start or runs badly and proposes how to fix it.
It sets up, changes, repairs and removes containers, always as a proposal you approve.
“I deleted it by mistake…”: it looks in the recycle bin, in previous versions and in snapshots.
It puts what can be deleted into one plan, with how much space each item frees.
“Create an account for my daughter that only sees Photos and Music”: it sets it up and she picks her password through her link.
It reads, changes and runs the code of one of your projects, with your permissions and without leaving that folder.
Shortcuts, versions and search to get to the point and lose nothing.
Type /disks or /errors to get straight to it, and @ to name a file, a camera, a container or a disk.
Edit any earlier question and switch between the answers with “‹ 2/3 ›”.
Find what was said in any conversation and jump to the exact spot.
Save a conversation as Markdown or PDF, or open a temporary one that isn’t kept and teaches it nothing.
It understands “dad”, “the living room TV” or “the old disk” and knows which account, device or disk you mean.
It suggests up to three questions based on how your NAS is doing right then.
It keeps watch without being asked and gets better with everything that works.
Important alerts reach your chats already explained, with the fix ready to approve.
Tell it what to watch (“that backups run”, “that Photos doesn’t fill up”) and it checks without waking the model.
The weekly report brings what keeps happening, what’s urgent and what it did on its own.
Every week, what you should know about the security of your apps and devices. Off by default.
It goes over what you talked about, suggests what to remember and tidies what it knows about you: you decide what stays.
A multi-step plan that worked is kept for next time, only if you say yes.
What it did right on a website it repeats without the AI, on a schedule too; if the page changes, it stops and says so.
Each popular plugin tells you which NAS feature replaces it, which app to install or why it can’t be done.
You choose where each question is handled: easy ones on the NAS, hard ones wherever there is more power.
With open models you download once. The panel recommends the one that fits your machine, and with a graphics card it answers several times faster.
With LGM Connect, Cloudy AI can use your computer’s graphics card for hard questions, without anything leaving your home.
Optional and off by default: a compatible provider, Anthropic’s Claude included, with your own key. Before anything goes out, your personal data is masked.
Not to think: with the model on the NAS it answers offline. You only need the internet to download the model the first time and for what you turn on: web search, your chats or an outside provider.
Your NAS. When you turn it on, the panel recommends the model that fits its memory; with a graphics card it answers several times faster, and with LGM Connect it can use your computer’s.
No, as long as it thinks on the NAS or your PC: your questions, your files and what it learns about you stay at home. Only if you choose an outside provider does the question go out, with your personal data masked.
It never does serious things without asking you, and almost everything it does can be undone from the History. If you’d rather take it slowly, leave it on Chat only or Copilot.
No. It comes with LGM-OS, in both editions. An outside provider, if you choose one, bills you itself through your own key.
Install LGM-OS, turn on Cloudy AI and ask it how your disks are doing.
Hello. What do you need?
What you write is kept to answer you and deleted after 90 days without messages. Controller: Luis García Menéndez. Privacy