SuperAPI AI Directory

diffron

Git commit message and PR description generator using AMD Lemonade via lemonade-python-sdk.

Open website

About this tool

Diffron is a Git commit message and PR description generator that uses the lemonade-python-sdk to call a local AMD Lemonade LLM server to perform generation. It is a production-grade reference implementation submitted to the AMD Lemonade Developer Challenge 2026, part of the Tetramatrix ecosystem, alongside Aicono, TabNeuron, Sorana, RyzenPilot, and Lemonade-Python-SDK.

Main capabilities

  • Automatic commit messages: Generate commit messages in Conventional Commits format based on staged changes.
  • PR descriptions: Generate detailed PR titles and descriptions based on branch diffs.
  • Lemonade integration: Works with a local Lemonade LLM server, no cloud required.
  • Cross-platform: Supports Windows, Linux, and macOS, and is compatible with GitHub Desktop 3.5.5+.
  • Auto-detection: Automatically discovers running Lemonade instances.
  • Curated models: Provides recommended models for different tasks for easy selection.
  • AI-aware: Automatically skips when AI coding agents (Claude, Copilot, Cursor, Aider, MiMo, Kilo, Qwen, etc.) are making the commit.

How to use

  1. Install AMD Lemonade Server: Download the installer from AMD Lemonade Releases, run Lemonade_Server_Installer.exe, launch Lemonade Server from the desktop shortcut, and download a model (e.g., qwen3.5-0.8b-gguf) via the Lemonade UI.
  2. Install lemonade-python-sdk: The source code is at github.com/Tetramatrix/lemonade-python-sdk.
  3. Configure the environment: Set the Lemonade Server URL (permanently or temporarily).
  4. Install Diffron.
  5. Set the model: Diffron comes with curated models, and the default model works out of the box (default model: qwen2.5-it-3b-FLM, provided with Lemonade). Other models can be specified manually via the Python API or environment variables.
  6. Install Git Hooks.

GitHub Desktop workflow

  • Modify files.
  • Open GitHub Desktop.
  • Enter any commit message (e.g., "auto").
  • Click "Commit to main".
  • Diffron will replace your message with an AI-generated one.

AI agent detection Diffron automatically detects whether an AI coding agent (Claude, Copilot, Cursor, Aider, Codex, Kilo, Mimo, Hermes, OpenCode, FreeBuff, etc.) is making the commit and skips message generation—because the agent itself has already generated a good message. Detection has three layers:

  • Environment variables: Scan all environment variables for AI agent keywords (CLAUDE, COPILOT, CURSOR, AIDER, CODEX, OPENAI, AGENT, BOT, etc.).
  • Git configuration: Check user.name and user.email for AI patterns (e.g., "Claude", "Copilot", "ai@anthropic.com").
  • Message quality: If the commit message already conforms to Conventional Commits format (feat:, fix:, etc.), Diffron skips. You can add custom detection patterns, configure via git config, or test manually.

Use cases Suitable for developers who want to automatically generate standardized Git commit messages and PR descriptions locally without the cloud, especially for users using AMD Ryzen AI PCs with a local Lemonade LLM server, and workflows using GitHub Desktop.

Limitations and notes

  • Requires AMD Lemonade Server to be installed and running first, and a model to be downloaded.
  • Depends on lemonade-python-sdk.
  • Skips generation when AI coding agent environment variables are detected or the commit message already conforms to Conventional Commits format.
  • Known issues include: Lemonade not detected, Hooks not working, model not found (404), wrong model used (the DIFFRON_MODEL environment variable overrides the default), Diffron not generating a message (AI agent detected). See docs/SETUP.md for full troubleshooting.
  • Licensed under the MIT License.
  • Version: 0.1.9, last updated: 2026-07-11.