Dunil's Hold
A short, atmospheric overview for anyone curious about the offline hybrid roguelike.
Designed as a focused 1–3 hour descent, Dunil's Hold keeps the math deterministic and the narration lush by pairing a solid engine with local AI personas running entirely on-device.
About the Experience
The game stitches old-school dungeon-crawl rigor with a custom engine that translates combat rolls, quiet exploration, and atmospheric triggers into poetic descriptions. It never imagines the state; the map, monsters, and rules are always grounded in Swift code, while MLX-enabled AI models supply prose.
- Engine: Swift-native determinism for macOS and iOS, with a CLI harness for deterministic runs.
- AI: Local model (Liquid AI LFM2) run through Apple's MLX stack without network calls.
You can expect tight pacing, mechanical clarity, and pockets of narrative warmth. The player is always the source of agency, the AI merely colors the aftermath.
Privacy Policy
In line with Apple’s App Privacy policy, Dunil's Hold does not collect, store, or transmit any user data. There are no analytics, no advertising, and no network calls related to user behavior.
The app runs fully offline. Any AI narration results from models bundled inside the app and managed by the game engine; nothing leaves the device during gameplay.
Apple App Privacy – Data Not Collected
| Data Type | Collection Status |
|---|---|
| Location | Not collected |
| Contact Info | Not collected |
| Identifiers (IDFA, etc.) | Not collected |
| Usage Data | Not collected |
| Diagnostics | Not collected |
Because no data is collected, Dunil's Hold does not share any information with third parties. If you have privacy questions or need further clarification, reach out via the contact information provided on the main project website.
Support & Issues
Found a bug? Visit the project on GitHub to report issues or browse existing reports.