Git Commit Message Generator

Generate clear, conventional Git commit messages directly inside VS Code using your preferred AI provider.

Table of Contents

• What this extension does
• Who this is for
• Key features
• Quick start (5 minutes)
• How to use
• Configuration
  • 1) Core settings
  • 2) Provider settings
  • 3) Protocol settings
  • 4) Example settings.json (Gemini)
• Supported providers
• Tips for better commit messages
• Troubleshooting
• FAQ
• Contributing
• License

What this extension does

This extension reads your staged Git changes, sends a summary/diff to a configured LLM, and writes a commit message suggestion that follows Conventional Commits style.

In short, it helps you:

• write commit messages faster,
• keep message style consistent,
• reduce mental overhead when committing often.

Who this is for

• Developers who want cleaner commit history.
• Teams that use Conventional Commits (feat:, fix:, chore:...).
• Beginners who are unsure how to phrase good commit messages.

Key features

• ✅ One-click commit message generation from staged changes.
• ✅ Conventional-Commit-friendly output.
• ✅ Multiple providers supported (cloud and local).
• ✅ Configurable prompt, system instruction, and generation parameters.
• ✅ Works from Source Control UI in VS Code.

Quick start (5 minutes)

1. Install Git Commit Message Generator from your extension source.
2. Open a Git repository in VS Code.
3. Open Settings (JSON).
4. Add your provider config (example below uses Gemini).
5. Stage your files in Source Control.
6. Click Generate Commit Message in the Source Control toolbar.

Command ID used by this extension:

• git-commit-generator.generateCommitMessage

How to use

1. Make code changes.
2. Stage files in Source Control (+) or via terminal (git add ...).
3. Run Generate Commit Message from the Source Control toolbar.
4. Review the generated message.
5. Edit if needed, then commit.

Important: the generator works on staged changes. If nothing is staged, no message will be generated.

Configuration

All settings are under the namespace:

• git-commit-generator.*

1) Core settings

• git-commit-generator.llm.provider
  Select active provider. Options: aliyun, openai, ollama, deepseek, anthropic, tencent, siliconflow, volcengine, gemini, custom

• git-commit-generator.llm.prompt
  Prompt template used for message generation.

• git-commit-generator.llm.system
  System instruction for commit style.

• git-commit-generator.llm.temperature
  Output randomness (0 to 1). Lower = more deterministic.

• git-commit-generator.llm.top_p
  Nucleus sampling value (0 to 1).

• git-commit-generator.llm.max_tokens
  Maximum generated token count.

2) Provider settings

Each provider uses this pattern:

• git-commit-generator.providers.<provider>.url
• git-commit-generator.providers.<provider>.model
• git-commit-generator.providers.<provider>.apiKey

3) Protocol settings

Some providers also include a protocol selector:

• git-commit-generator.providers.gemini.protocol
  • Available value: gemini
• git-commit-generator.providers.custom.protocol
  • Available values: openai, ollama

Examples of <provider> values:

• aliyun
• openai
• ollama
• deepseek
• anthropic
• tencent
• siliconflow
• volcengine
• gemini
• custom

4) Example settings.json (Gemini)

{
  "git-commit-generator.llm.provider": "gemini",
  "git-commit-generator.providers.gemini.apiKey": "YOUR_API_KEY",
  "git-commit-generator.providers.gemini.model": "gemini-2.5-flash",
  "git-commit-generator.providers.gemini.protocol": "gemini"
}

You can also set custom OpenAI-compatible or Ollama-compatible endpoints:

{
  "git-commit-generator.llm.provider": "custom",
  "git-commit-generator.providers.custom.url": "https://your-endpoint.example.com/v1/chat/completions",
  "git-commit-generator.providers.custom.model": "your-model-name",
  "git-commit-generator.providers.custom.apiKey": "YOUR_API_KEY",
  "git-commit-generator.providers.custom.protocol": "openai"
}

Supported providers

• Alibaba Bailian (aliyun)
• OpenAI (openai)
• Ollama local (ollama)
• DeepSeek (deepseek)
• Anthropic (anthropic)
• Tencent Hunyuan (tencent)
• SiliconFlow (siliconflow)
• Volcengine (volcengine)
• Gemini (gemini)
• Custom endpoint (custom)

Tips for better commit messages

• Stage related files together (one logical change per commit).
• Keep commits small and focused.
• Review generated text before committing.
• Use scope when useful, e.g. fix(auth): ....
• Prefer imperative style, e.g. “add”, “fix”, “refactor”.

Troubleshooting

1) "No staged files" warning

Cause: no files are staged.

Fix:

• Stage files in Source Control, then run generator again.
• Or use terminal: git add .

2) Output language/style is not what you want

Fix:

• Customize git-commit-generator.llm.prompt
• Customize git-commit-generator.llm.system
• Lower temperature for more consistent output

3) API request failed

Check:

• Correct provider selected in git-commit-generator.llm.provider
• Valid apiKey
• Correct url and model
• Internet/network access for cloud providers

4) Local Ollama not responding

Check:

• Ollama service is running
• URL is reachable (commonly http://localhost:11434/api/generate)
• Model is installed locally

FAQ

Does this extension commit automatically?

No. It only generates the commit message. You still review and commit manually.

Can I use my own OpenAI-compatible endpoint?

Yes. Use provider custom and set your endpoint/model/api key.

Does it support team conventions?

Yes. You can enforce team style via llm.system and llm.prompt.

Contributing

Issues and pull requests are welcome.

If you plan to contribute code, please:

1. Open an issue describing the change.
2. Keep PRs focused and small.
3. Include clear testing notes.

Credits

Commit message format inspired by Conventional Commits.

Based on the original extension by chenkai2.
Modified and maintained by klaveriuzent.

License

MIT License