---
title: HDS BlackBox Translator (Chess)
emoji: ♟️
colorFrom: blue
colorTo: yellow
sdk: gradio
sdk_version: 5.33.0
app_file: app.py
pinned: false
tags:
- gradio
- chess
- interpretability
- explainability
- education
- visualization
- open-source
- hds
---

# Chess Trainer (HDS-style Trace)

- Board-only input (click to move)
- English output
- Short verdict + HDS-style trace (state vector, causal deltas, risks)
- Top-3 candidate moves with pros/risks
- No opponent auto-move (you can move both sides manually)
## Overview

HDS BlackBox Translator (Chess) is a lightweight, deterministic demo that turns a chess position into a short causal explanation.

**Important:** this is **not** a chess engine for best-move search, and it does **not** read an engine’s PV/evaluation output.  
Instead, it uses engine-inspired evaluation factors (e.g., material, king safety, piece activity, center reach, initiative, tactical tension), projects them into an **HDS state vector**, and generates an **HDS Trace** on each move:

**State → Delta → Cause → Risk → Next**

### What you get per move
- A concise English comment explaining what changed and why (cause/effect)
- A state-vector table (before/after/Δ) for transparency
- Top-3 candidate moves with brief pros/risks (from the current position)

### How to use
- Click pieces to make moves on the board (white orientation is fixed).
- There is no automatic opponent move — you can move both sides manually.
- Use the trace as a “thinking scaffold” for learning: what matters, why it matters, and what the immediate risks are.

### Design goals
- Deterministic output (same position → same trace)
- Minimal UI (board + trace + reset)
- Clear, community-editable vocabulary and axes (easy to extend)

📌 Full description (design, specs, guardrails, reproducibility): **[DESCRIPTION.md](./DESCRIPTION.md)**

## Links
- GitHub (source code): https://github.com/gatchimuchio/HDS-BlackBox-Translator
- Hugging Face Space: https://huggingface.co/spaces/gatchimuchio/HDS.BlackBoxTranslator

## Author
- Created by: gatchimuchio (GitHub / Hugging Face)

## License
- MIT License (see LICENSE)

## Reuse / Run locally
This Space is fully open-source.
- Clone the GitHub repo, or use the Space files directly.
- Install: `pip install -r requirements.txt`
- Run: `python app.py`