Guides › Using this library · Emergency · Index A–Z
The library comes with an AI that runs on the computer itself. It needs no internet, and nothing you type leaves the computer.
⭐ Easiest: use the Library Assistant
Double-click START-Windows, START-Mac or START-Linux.sh in the OfflineLibrary folder (other devices: see Start on Any Device). Type a question. It searches every book, guide and encyclopedia on the drive, and the AI writes an answer with links to the exact pages.
- It picks the right AI size for your computer. To change it, open More → Offline AI settings.
- Search results show up in a second. The written answer takes 30 seconds to a few minutes.
- Search in (under the question box) picks where it looks: Guides, Books, Wikipedia. Guides alone gives short, plain answers from the library's own guides; on a phone it also skips loading every book page, so the first answer comes much sooner. For a quicker written answer, set Sources to 4: the AI then has less to read.
- Keep the black window open while you use it.
Most people can stop reading here. The rest is for chatting with the AI on its own, without the library search.
⚠️ Remember
- The AI can be wrong or make things up. Check anything about health or safety in the books.
- It can't look anything up online. Its knowledge stops when it was made.
| Double-click | When to use it |
|---|---|
| CHAT-3B-light | Almost any computer. Fast |
| CHAT-8B-smart | Computers with 8 GB+ of memory. Smarter, slower |
| CHAT-GPU-Qwen7B | Gaming computers with a graphics card. Much faster |
| WEB-UI | Chat in your web browser: wait until it says server listening, then go to http://127.0.0.1:8080 |
The 8B and GPU ones are only on bigger drives (about 64 GB or more).
Open a terminal in the AI-Offline folder (Mac: open Terminal, type cd , drag the folder in, press Enter. Linux: right-click in the folder → Open in Terminal). Then type:
./chat-3B-light.sh(or ./chat-8B-smart.sh, or ./web-ui.sh for the browser). If it says permission denied, type chmod +x *.sh first. These files are only on drives made on a Mac or Linux computer.
| Problem | Fix |
|---|---|
| Windows protected your PC | Click More info, then Run anyway |
| The black window flashes and closes | A model file may be missing. Run the installer again to fetch it |
| Out of memory, or it crashes | Use CHAT-3B-light and close other programs |
| It's very slow | Normal on older computers. Use the 3B one |
| The web page won't open | Check the WEB-UI window says server listening, and type exactly http://127.0.0.1:8080 |
Models (in AI-Offline\models): Qwen 2.5 1.5B (Quick), Llama 3.2 3B (Standard), Llama 3.1 8B (Smart) and Qwen 2.5 7B (Graphics card). Free open-weight models.
Engine (in AI-Offline\engine): llama.cpp, for Intel/AMD, ARM and graphics cards.
To copy the AI to another computer, copy the whole AI-Offline folder. Nothing needs installing.
To run it by hand (to see an error), in Command Prompt, with your drive letter:
"E:\OfflineLibrary\AI-Offline\engine\x64\llama-cli.exe" -m "E:\OfflineLibrary\AI-Offline\models\Llama-3.2-3B-Instruct-Q4_K_M.gguf" -c 4096 -cnv