Add Space README
Browse files
README.md
CHANGED
|
@@ -1,10 +1,61 @@
|
|
| 1 |
---
|
| 2 |
title: Phantom Grid
|
| 3 |
-
emoji:
|
| 4 |
-
colorFrom:
|
| 5 |
-
colorTo:
|
| 6 |
sdk: docker
|
|
|
|
| 7 |
pinned: false
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
---
|
| 9 |
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
title: Phantom Grid
|
| 3 |
+
emoji: 🕵️
|
| 4 |
+
colorFrom: indigo
|
| 5 |
+
colorTo: gray
|
| 6 |
sdk: docker
|
| 7 |
+
app_port: 7860
|
| 8 |
pinned: false
|
| 9 |
+
license: mit
|
| 10 |
+
tags:
|
| 11 |
+
- thousand-token-wood
|
| 12 |
+
- delightful
|
| 13 |
+
- game
|
| 14 |
+
- agent
|
| 15 |
+
- minicpm
|
| 16 |
---
|
| 17 |
|
| 18 |
+
# 🕵️ Phantom Grid
|
| 19 |
+
|
| 20 |
+
**An AI-driven noir detective game.** You are a detective hunting a phantom suspect
|
| 21 |
+
across a stylized London grid. Issue notices, raise lookouts, set blockades, and
|
| 22 |
+
**interview AI-roleplayed witnesses** whose memories decay over time — all rendered
|
| 23 |
+
in a custom `gr.Server` HTML/JS board interface (no stock Gradio components).
|
| 24 |
+
|
| 25 |
+
> Track: **Delightful — Thousand Token Wood** (an AI-driven game).
|
| 26 |
+
|
| 27 |
+
## 🏅 Prize-category badges
|
| 28 |
+
|
| 29 |
+
- 🎮 **Thousand Token Wood / Delightful** — a playable, AI-driven detective game.
|
| 30 |
+
- 🎨 **Off Brand** — fully custom `gr.Server` HTML/JS frontend, well beyond the stock components.
|
| 31 |
+
- 🪶 **Small & Mighty** — runs entirely on a single under-32B model (MiniCPM4.1-8B).
|
| 32 |
+
|
| 33 |
+
## 🤖 Model & inference
|
| 34 |
+
|
| 35 |
+
- **Model:** [`openbmb/MiniCPM4.1-8B`](https://huggingface.co/openbmb/MiniCPM4.1-8B)
|
| 36 |
+
(text, Q4_K_M GGUF) — ~8B params, well under the 32B cap.
|
| 37 |
+
- **Inference:** the blessed llama.cpp path — an OpenAI-compatible
|
| 38 |
+
`llama.cpp` server (`llama-cpp-python`) on `:8080/v1`, with the app pointing at it via
|
| 39 |
+
`PHANTOM_GRID_LLAMACPP_BASE_URL`. The model is **baked into the image at build time**, so the
|
| 40 |
+
Space boots without any multi-GB download or build-from-source.
|
| 41 |
+
|
| 42 |
+
## 🖥️ Hardware
|
| 43 |
+
|
| 44 |
+
Runs on **CPU** (the model is Q4_K_M, ~5 GB RAM). Witness replies are slow on the free
|
| 45 |
+
CPU tier; an upgraded CPU or a GPU Space makes interviews snappy. The text backend has no
|
| 46 |
+
voice path (`PHANTOM_GRID_WITNESS_CHAT_TTS=0`).
|
| 47 |
+
|
| 48 |
+
## 🎥 Demo video
|
| 49 |
+
|
| 50 |
+
<!-- SHIP: replace with the published demo video URL -->
|
| 51 |
+
_Demo video: TODO — add link before final submission._
|
| 52 |
+
|
| 53 |
+
## 📣 Social post
|
| 54 |
+
|
| 55 |
+
<!-- SHIP: replace with the published social-media post URL -->
|
| 56 |
+
_Social post: TODO — add link before final submission._
|
| 57 |
+
|
| 58 |
+
## ▶️ How to play
|
| 59 |
+
|
| 60 |
+
Start a new case, read the briefing, then use the board tools (notices, lookouts,
|
| 61 |
+
blockades, searches) and question witnesses to corner the suspect before the turn limit.
|