This repository was archived by the owner on Oct 28, 2026. It is now read-only.
Repository navigation
Settings and Data
speedyrulz edited this page Oct 28, 2026
·
3 revisions
Open it with Ctrl+5.
| Section | What it holds |
|---|---|
| Engine | The folder with llama-server.exe, a Test button, and the detected version and GPUs. See Building the Engine. |
| Models | The models folder, and whether loading a model unloads the previous one (on by default). See Model Library. |
| Default load settings | Load settings used for models that have none of their own. Press Save load defaults to apply. See Loading Models. |
| New chats | The default system prompt, sampling values (temperature, top P, top K, min P, repeat penalty) and thinking settings for new chats, and whether reasoning blocks start expanded. Existing chats keep their own settings. |
| Appearance | Dark, Light or System. The sun/moon button in the top bar switches between dark and light quickly. Also Minimize to system tray and Close to system tray (both off by default); see Running in the system tray. |
| Keyboard shortcuts | See Keyboard Shortcuts. |
| About | The app version and a button that opens the data folder. |
Some settings live on the page they belong to rather than here:
- MCP Tools page — tool use on or off, tool approval, tool loading mode and budget, shortening old tool results. See MCP Tools.
- Developer page — port, access, API key, idle unload. See Local API Server.
- Chat settings panel — system prompt, sampling, presets, structured output and what to do when the context fills up, for the current chat. See Chat.
- Load panel and My Models — per-model load settings. See Loading Models.
Two switches under Settings → Appearance let the app keep running out of the way, with loaded models and the Local API Server still working:
- Minimize to system tray — minimizing hides the window from the taskbar and shows a RulZ LLM Suite icon in the notification area (by the clock).
- Close to system tray — the window's close button (X) hides the window to the tray instead of quitting.
While the window is hidden:
- Click the tray icon to bring the window back. The icon disappears again once the window is showing.
- Right-click it for Show RulZ LLM Suite and Quit. Quit exits the app properly and stops any running models.
- Starting the app again (for example from the Start menu) also brings the hidden window back.
- Signing out of Windows or shutting down still closes the app normally.
If you turn both switches off while the window is hidden, it comes back by itself.
Everything the app saves is in %APPDATA%\RulZ LLM Suite (usually C:\Users\<you>\AppData\Roaming\RulZ LLM Suite). Settings → About → Open takes you there.
| File or folder | Contents |
|---|---|
settings.json |
App settings, including the Developer page's API server settings |
model-defaults.json |
Saved load settings for individual models |
presets.json |
Chat presets |
conversations\ |
One JSON file per chat |
folders.json |
The names of your chat folders |
mcp.json |
MCP server configuration |
gguf-cache.json |
Cached model metadata; safe to delete, it is rebuilt on the next scan |
running-pids.json |
Which llama-server processes the app started, so it can clean up after a crash |
Your models are not stored here; they stay in your models folder, and the app never changes them except when you choose Move to Recycle Bin.
Things to know:
- Back up by copying the folder while the app is closed.
- The API key (Developer page) and any tokens in
mcp.json(envandheaders) are stored as plain text. Don't share these files or post them in an issue. - Deleting a chat moves its file to the Recycle Bin, so a mistaken delete can be undone from Windows.
- To reset the app's settings, close it and delete
settings.json. Chats, presets and models are untouched.
Using the app
Reference
Contributing