AI Commit Message Generator for VS Code 🤖✨

Local AI Commit Message Generation - Transform code diffs into conventional commits using open-source models. Privacy-focused, offline-capable solution for developers.

Demo showing commit message generation process (left) and settings (right)

Features 🌟

• 🔒 Privacy First - No data leaves your machine
• ⚡ Multi-Backend Support - Compatible with popular AI runners
• 📜 Commit Standard Compliance - Conventional Commits 1.0.0
• 🖥️ Hardware Aware - Optimized for various setups
• 🌐 Model Agnostic - Use any compatible LLM

Quick Start 🚀

1. Install extension:

   code --install-extension Its-Satyajit.ai-commit-message

2. Set up AI backend:

   # For CPU-focused systems
   ollama pull phi-3
   
   # For GPU-equipped machines
   ollama pull deepseek-r1:8b

3. Generate your first AI commit via VS Code Source Control view

Hardware Requirements 🖥️

Tested Environment

OS: openSUSE Tumbleweed
CPU: Intel i7-8750H (6c/12t @4.1GHz)
GPU: NVIDIA GTX 1050 Ti Mobile 4GB
RAM: 16GB DDR4
Storage: NVMe SSD

Minimum Recommendations

• CPU: 4-core (2015+)
• RAM: 8GB
• Storage: SSD
• Node: ^22
• Vscode: ^1.92.0

Model Compatibility 🧠

Performance Characteristics

Model Family	Example Models	Speed*	Quality*	Use When...
Lightweight	phi-3, mistral	22 t/s	██▌	Quick iterations
Balanced	llama3, qwen	14 t/s	███▎	Daily development
Quality-Focused	deepseek-r1	7 t/s	████▋	Complex changes

* Metrics from personal testing on mobile GTX 1050 Ti (Q4_K_M quantization)

        Speed vs Quality Tradeoff

        ▲
        │ 
Quality │.....█████ (deepseek-r1)
        │...███     (llama3) 
        │.██▌       (phi-3)
        └───────────────────▶ Time

Configuration ⚙️

Backend Setup

Option 1: Ollama (Simplest)

curl -fsSL https://ollama.com/install.sh | sh
ollama serve

* For more info, visit Ollama

Option 2: LM Studio (Advanced)

lmstudio serve --model ./models/deepseek-r1.gguf --gpulayers 20

* For more info, visit LM Studio

Extension Settings

{
  "commitMessageGenerator.provider": "ollama",
  "commitMessageGenerator.apiUrl": "http://localhost:11434",
  "commitMessageGenerator.model": "deepseek-r1:8b",
  "commitMessageGenerator.temperature": 0.7,
  "commitMessageGenerator.maxTokens": 5000,
  "commitMessageGenerator.apiKey": "your_api_key (if required by your OpenAI-compatible/ollama endpoint)",
  "commitMessageGenerator.types": [
    "feat: A new feature",
    "fix: A bug fix",
    "chore: Maintenance tasks",
    "docs: Documentation updates",
  ],
  "commitMessageGenerator.scopes": ["ui", "api", "config"]
}

Optimization Guide

GPU Acceleration

# NVIDIA Settings
export OLLAMA_GPUS=1
export GGML_CUDA_OFFLOAD=20

# Memory Allocation (4GB VRAM example)
┌───────────────────────┐
│ GPU Layers: 18/20     │
│ Batch Size: 128       │
│ Threads: 6            │
└───────────────────────┘

Performance Tips

• Start with phi-3 for quick feedback
• Switch to deepseek-r1 for final commits
• Use --no-mmap if experiencing slowdowns
• Reduce GPU layers when memory constrained

Troubleshooting 🔧

Issue	First Steps	Advanced Fixes
Slow generation	1. Check CPU usage 2. Verify quantization	Use --no-mmap flag
Model loading fails	1. Confirm SHA256 checksum 2. Check disk space	Try different quantization
GPU not detected	1. Verify drivers 2. Check CUDA version	Set CUDA_VISIBLE_DEVICES=0

FAQ ❓

Why local AI instead of cloud services?

• Privacy: Code never leaves your machine
• Offline Use: Works without internet
• Cost: No API fees
• Customization: Use models tailored to your needs

How to choose between models?

Quick Sessions → phi-3/mistral:

• Prototyping
• Personal projects
• Low-resource machines

Important Commits → deepseek-r1:

• Production code
• Team projects
• Complex refactors

Legal & Ethics

Neutrality Statement

This project and I am not affiliated with, endorsed by, or sponsored by:
- Ollama
- LM Studio
- Any model creators

Mentioned tools/models are personal preferences based on technical merits.

Contributing 🤝

1. Fork repository
2. Install dependencies:

   npm install

3. Build extension:

   npm run package

4. Submit PR with changes

Full contribution guidelines

License 📄

MIT License - View License

---

Built by Developers, for Developers - From quick fixes to production-grade commits 💻⚡

Report Issue