Settings¶
Open Settings with the gear in the top bar, the model chip, or Ctrl+, in the desktop app. Esc or close closes it. The theme, navigation and dialogue sections apply at once; the provider and the keys apply when you press save.
Theme¶
| Choice | |
|---|---|
| Follow the sim's daylight (the default) | The app blends from light to dark with the province's own sunrise and sunset; dusk and dawn sit in between. |
| Always light, Always dark | A steady theme. |
| Follow this device | The system's light or dark setting. |
Navigation¶
Unreal Engine's viewport controls, for the 3D close-up.
| Setting | Range | Default |
|---|---|---|
| Camera speed | 0.33 to 32 (the wheel with a mouse button held goes further) | 1 |
| Scroll speed | 1 to 8: how far a wheel notch dollies (96 cm at 5, at 10 m from what the camera looks at) | 5 |
| Mouse sensitivity | 0.01° to 1° a pixel | 0.20° |
| Fly with W A S D | only with a mouse button held (Unreal's default), always, never | only with a button held |
| Distance-scaled speed | Every movement scales with the distance to what the camera looks at | on |
| Invert mouse look, Invert orbit, Invert middle-mouse pan, Invert right-mouse dolly, Invert the right stick | off |
Unreal defaults puts them all back.
Dialogue¶
| Mode | |
|---|---|
| Off (procedural chatter) | No model is asked. |
| On demand (the default) | A conversation is scripted when you press Script with AI. |
| Automatic when zoomed in | The conversation in view is scripted, one at a time, while the clock runs at an animated speed and you are zoomed in close. |
See Conversations.
Provider¶
| Setting | |
|---|---|
| Serve dialogue with | auto (Ollama if it has a model, otherwise the first hosted provider with a key), Ollama (local), Anthropic Claude, OpenAI, Google Gemini, OpenAI-compatible endpoint. The line under it says which provider and model are in use now. |
| Ollama model | The models installed in Ollama; refresh looks again (needed if Ollama was started after the app). Without a choice, the first of the gpt-oss, Gemma, Qwen and Llama families. |
| gpt-oss reasoning | low (the fastest, and recommended for dialogue), medium, high. gpt-oss cannot switch its reasoning off. |
| Compatible base URL | For an OpenAI-compatible endpoint, such as http://localhost:1234/v1 (LM Studio, vLLM, Groq…). |
| Model for each hosted provider | Left empty: claude-sonnet-4-6, gpt-4o-mini, gemini-2.0-flash, local-model. |
Keys¶
API keys for Anthropic, OpenAI, Google Gemini and an OpenAI-compatible endpoint. Each shows configured once set. They are kept in the app's storage on this machine and handed to the engine, which holds them in memory and sends them only to their provider.
The cache¶
clear dialogue cache empties the cache of scripted conversations and says how many it removed.
What the app remembers¶
| Remembered | |
|---|---|
| The basis you built in the Scenario panel | Yes; the province starts on it next time, at day 0 |
| The view, the side panels, the dialogue mode, the 3D close-up on or off | Yes |
| Navigation, theme, provider, keys | Yes |
| The workbench's folder, open files, expanded folders and pane sizes | Yes |
| The running province, the speed, the legend | No: each launch starts the province afresh, paused |