File size: 1,704 Bytes
ed804aa b022028 ed804aa b022028 ed804aa 395c43d ed804aa 395c43d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 | ---
title: DOD LLM Server
emoji: 🔥
colorFrom: pink
colorTo: gray
sdk: gradio
sdk_version: 6.17.3
python_version: '3.10'
app_file: app.py
pinned: false
license: mit
short_description: LLM Server for DOD Multiplayer UNO Game
tags:
- build-small-hackathon
- thousand-token-wood
- gradio
- nemotron
- llama-cpp
- sponsor:openai
- sponsor:nvidia
- achievement:llama
---
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
# DOD LLM Server
This Space is the lightweight LLM inference companion for **DOD - Deploy or Draw**, the multiplayer UNO-inspired production-incident game.
The main game calls this service whenever it needs small-model reasoning:
- **Nemotron bot turns**: choose whether to play or draw, select a valid card, and pick the next stack after wild cards.
- **IT Director reactions**: generate short crisis-aware comments in English and Portuguese after important card plays.
The server exposes a Gradio API around a GGUF model served through `llama-cpp-python`. In the public DOD demo, it is used as the hosted Nemotron inference endpoint. In local development, the same app can run on `127.0.0.1:7880` and be wired into the game through `LLM_URL`.
The game backend remains authoritative: it computes legal cards, validates model output, and keeps fallback behavior available if an endpoint is unavailable. This server provides the AI decisions and personality layer; it does not own the game rules.
## Related Links
- Main DOD game Space: https://huggingface.co/spaces/build-small-hackathon/dod-uno
- Development repository: https://github.com/DEVAIEXP/doduno
- Field notes: https://huggingface.co/blog/elismasilva/dod-uno-en |