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