GoLinuxCloud AI Bot

There is an Ask AI button in the bottom-right corner of every page on this site. It opens the GoLinuxCloud AI Bot, which answers questions using the tutorials already published here, and links the exact section each part of its answer came from.

This page explains where those answers come from, how to get a good one, the limits the bot works within, and what happens when it gets something wrong.

Last reviewed: 20 August 2026.

1. Where the answers come from

Only from GoLinuxCloud. There are more than 2,600 tutorials on this site covering Linux, Kubernetes, Ansible, Python, shell scripting, SQL, networking and security. When you ask a question, the bot searches those guides for the passages that match, then writes an answer using nothing but what it found.

That single design decision is what makes the bot useful, and it has two consequences worth knowing about.

  • Every answer cites its source. The numbered markers in an answer point at the guides underneath it, so you can open the original section and check the bot's work in one click. Please do.
  • It will tell you when the site has nothing. Ask about a topic these tutorials do not cover and the bot says so instead of guessing. A general-purpose chatbot will happily invent a plausible answer about a tool it has never seen. This one is built to decline, because an invented firewalld flag costs you an afternoon.

The bot also runs its search in your browser rather than on a server, against a static index the site ships. Nothing is sent anywhere until you actually ask a question.

2. How to ask it

Open the panel in whichever way suits you:

  • Click Ask AI in the bottom-right corner of any page.
  • Press the / key anywhere on the site.
  • Type into the ask box in the site header, and your question opens the panel already asked.

Then treat it like a conversation rather than a search box. Full questions work better than keywords, because the bot is matching meaning and not just words. "Why does my pod stay in CrashLoopBackOff" gets you further than "crashloopbackoff".

A few things about the panel that are not obvious:

  • Follow-up questions work. The bot remembers the conversation, so you can ask "and how do I make that permanent?" without repeating the context.
  • The conversation follows you. Open a different guide, or a second tab, and the same conversation is still there. It is kept for 12 hours or 14 exchanges, whichever runs out first.
  • The broom clears it. The clipboard beside it copies the whole conversation as text, which is handy for pasting into notes or a ticket.
  • Esc clears what you typed when the box has something in it, and closes the panel when it does not.

3. What it is good at, and what it is not

It is genuinely good at three things: finding the right guide out of thousands when you cannot remember its title, quoting a command that has actually been tested on a real host, and stitching together steps that are spread across two or three separate tutorials.

It is not good at anything outside this site, and it does not know your environment. It has never seen your distribution, your versions, your firewall rules or your cluster. So treat every command it gives you the way you would treat any command from the internet: read it, understand it, and run it in a lab before you run it in production.

4. It can make mistakes

It can, and it will. This is worth being specific about rather than hiding behind a disclaimer.

The bot does not copy the guides word for word. It reads the relevant passages and writes a fresh answer, and something written fresh can go wrong in ways the original does not. It can drop a caveat that mattered. It can flatten a "on RHEL 9 only" into general advice. When a question is ambiguous it can pick the wrong guide and answer confidently from it.

That is why the citations are there, and why they are the most important part of any answer. The linked section is the reviewed, tested version. The answer above it is a summary of that section written seconds ago by a machine.

When something is wrong, use Report a wrong answer at the bottom of the panel. It sends the conversation so I can see exactly what happened. I read every report. If the guide behind the answer is wrong, I fix the guide, which helps every future reader and not just the bot. If the guide was right and the bot misread it, that tells me something about how answers are being assembled, which is just as useful.

Reporting an error in a tutorial itself, rather than in a bot answer, is covered in the Corrections Policy.

5. Daily limits

Generating an answer costs real compute, and this site runs the bot inside a limited allowance rather than unlimited budget. Keeping it free for readers means keeping it inside a cap, so there are two: one for the whole site each day, and a smaller one per visitor so a single heavy user cannot spend everyone's share before lunch.

In practice you are unlikely to meet either. If you do, here is what you will see:

  • A popular question comes back instantly. Answers are cached, so a question someone already asked costs nothing and returns immediately.
  • When the day's allowance is gone, you still get an answer. The bot falls back to assembling the relevant passages directly from the guides without rewriting them. It reads more plainly, but the content and the links are the same material.
  • Everything resets at midnight UTC.

6. What happens to what you type

  • The conversation is kept in your browser, not on a server. It lives in localStorage on your own device, and clearing it with the broom removes it.
  • Questions are logged. I keep the question, the answer it produced, and which guides it drew on. That log is how I find out what readers want that this site has not covered yet, which makes it one of the more useful things on the site for deciding what to write next.
  • There is no account and no name attached. A random identifier in a glc_ai cookie counts your questions against the per-visitor daily cap. It is not linked to an email address, because none is collected.
  • Do not paste anything confidential. No passwords, private keys, tokens, licence keys, internal hostnames or customer data. Redact them the same way you would before pasting a log into a public forum.
  • A report sends the conversation. If you use Report a wrong answer, the exchange you chose to report comes with it so I can reproduce the problem.

How the site handles data generally is set out in the Privacy Policy, and how cookies are used in the Cookie Policy.

7. Why it says Beta

Because it is, and the label is not decoration. The way questions are matched to guides, the wording of the answers and the limits above are all still changing, sometimes weekly. Answers you get today may be phrased differently next month, and a question that gets declined this week may be answerable once the guide behind it exists.

The part that will not change is the rule underneath it. Answers come from these tutorials and they say where they came from, so you are never more than one click from the tested original.