README.md

metadata

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

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