|
Download README.md from nalemi/Trinity_Demo_Win64: direct link, hf CLI and curl.
- Browser
- Download file 1.39 kB
-
https://huggingface.co/nalemi/Trinity_Demo_Win64/resolve/main/README.md
- Command line
-
hf download hf://nalemi/Trinity_Demo_Win64/README.md
-
curl -L -o README.md https://huggingface.co/nalemi/Trinity_Demo_Win64/resolve/main/README.md
1.39 kB
| license: mit | |
| language: | |
| - en | |
| - de | |
| tags: | |
| - game-engine | |
| - local-ai | |
| - tts | |
| - rust | |
| - roleplay | |
|  | |
| # Trinity (Offline Demo Build) | |
| Trinity is a 100% offline-capable narrative game engine and cognitive framework built natively in Rust (`egui`). | |
| This standalone distribution is a playable technical demo showcasing the integration of local language models, deterministic game-state tracking, and low-latency neural TTS. | |
| ## Architecture & Tech Stack | |
| - **Inference:** Local abliterated Gemma 2 9B running via bundled `llama-server`. Completely offline, zero telemetry. | |
| - **State Bridge:** Deterministic game mechanics (vital signs, inventory management, D20 skill checks) integrated via local IPC and logged in SQLite. | |
| - **Voice Pipeline:** Piper neural TTS (ONNX runtime) with sentence-level audio buffer streaming for near-zero response latency. | |
| - **Client:** Native Rust desktop client (`egui`). | |
| ## Getting Started (Windows x64) | |
| 1. Download `Trinity_Demo_Win64.zip` (approx. 6.4 GB, bundled with quantized model weights and Piper runtime). | |
| 2. Extract the archive to a local directory (e.g. `C:\Trinity`). | |
| 3. Run `Trinity.exe`. | |
| ### Hardware Requirements | |
| - **OS:** Windows 10 / 11 (64-bit) | |
| - **GPU:** NVIDIA GPU with at least 6–8 GB VRAM recommended for smooth inference and audio streaming. | |
| - **Storage:** ~8 GB free disk space after extraction. |