shinkuan/Akagi

支持雀魂、天鳳、麻雀一番街、天月麻將，能夠使用自定義的AI模型實時分析對局並給出建議，內建Mortal AI作為示例。 Supports Majsoul, Tenhou, Riichi City, Amatsuki, with the ability to use custom AI models to analyze games in real time and provide suggestions. Comes with Mortal AI as a built-in example.

What it solves

Akagi is a real-time AI assistant for Riichi Mahjong games (specifically Mahjong Soul, Tenhou, and Riichi City). It helps players understand their performance during a match and learn from it by providing live analysis and move suggestions.

How it works

The application mirrors the game state by intercepting game data using either a system-wide MITM proxy or a controlled Chromium browser via the Chrome DevTools Protocol. This data is then processed through a mahjong engine and passed to an AI bot. The tool provides a draggable HUD that displays critical metrics like shanten, waits, agari rate, and tenpai rate, as well as recommended discards.

Who it’s for

It is designed for Mahjong players who want to improve their game through real-time AI analysis and performance tracking.

Highlights

  • Dual AI Backends: Includes a built-in pure-Rust neural network for immediate use and an optional cloud-inference API for access to stronger models.
  • Live HUD: Displays real-time game metrics and suggested attack/defense discards.
  • Flexible Capture Modes: Offers a MITM proxy for system-wide capture or a Chromium-based mode that requires no certificate trust.
  • Comprehensive History: Automatically records matches to track rank distributions and detailed statistics (win rate, riichi rate, etc.) via interactive charts.
  • AutoPlay: Supports autonomous bot play for Mahjong Soul and Tenhou.

Related

  • Project
    tontinton/maki

    A high-performance, token-efficient AI coding agent with a Rust-based TUI, designed to minimize context window pollution and costs.

  • Project
    future-agi/future-agi

    An open-source platform for building reliable AI agents that integrates simulation, evaluation, tracing, and guardrails into a single self-improving feedback loop.

  • Project
    Dwinovo/minecraft-numen

    An AI companion mod for Minecraft that acts as a real player, allowing users to automate mining, building, and combat using natural language instructions.

  • Project
    sanderland/katrain

    KaTrain is an AI-powered tool for analyzing Go games and playing against AI with real-time feedback and calibrated skill levels based on the KataGo engine.

  • Dispatch
    Hugging Face AI vs. AI: Multi-Agent Reinforcement Learning Competition System

    Hugging Face has introduced AI vs. AI, an open-source system for ranking deep reinforcement learning models through continuous multi-agent competitions using an ELO rating system.