✍️ CommitScribe - AI-Powered Commit Message Rewriting

CommitScribe is an Azure DevOps pipeline task that uses AI to automatically format and describe your commits. It supports multiple AI providers, including OpenAI, Anthropic, and Azure OpenAI.

The commit message and git diff are passed to the AI provider, which generates a new commit message based on the provided context. The new commit message is then used to update the commit message in the repository.

📦 Setup

1. Install the extension from the Visual Studio Marketplace
2. Configure your AI provider credentials:
   • Create a pipeline variable or variable group with your AI provider API key
   • Mark the variable as secret

📋 Prerequisites

1. Ensure your pipeline has a checkout step with persistCredentials enabled & fetchDepth set to 0:

steps:
    - checkout: self
      fetchDepth: "0"
      persistCredentials: true

2. Configure Git identity in your pipeline before the CommitScribe task:

steps:
    - script: |
          git config --global user.email "azure-pipelines@example.com"
          git config --global user.name "Azure Pipelines"
      displayName: "Configure Git"

3. Grant repository permissions:
   • For Azure Repos: Grant the build service account 'Contribute' & 'Force Push' permissions.
   • For GitHub repos: Configure GitHub service connection with write permissions
   • For other Git providers: Ensure the pipeline's service account has write access

Note

Why is force push required?

CommitScribe rewrites history. Use of the --force flag is necessary to persist updated commit messages in the target branch. This is a destructive operation, so use with caution! For more information, see Git's documentation on rewriting history.

🛠️ Basic Usage

Add the task to your pipeline:

steps:
    - checkout: self
      fetchDepth: "0"
      persistCredentials: true
    - script: |
          git config --global user.email "azure-pipelines@example.com"
          git config --global user.name "Azure Pipelines"
      displayName: "Configure Git"
    - task: commitscribe@0
      inputs:
          aiprovider: "openai" # Options: openai, anthropic, azure
          aiproviderkey: "$(YOUR_API_KEY)"

⚙️ Configuration Options

Input	Required	Default	Description
aiprovider	Yes	openai	AI provider (OpenAI, Anthropic, or Azure OpenAI)
aiproviderkey	Yes	-	API key for the selected provider
model	No	Provider-specific	Model to use (gpt-4 for OpenAI, claude-3-sonnet-20240229 for Anthropic)
azendpoint	No*	-	Azure OpenAI endpoint URL (*Required for Azure)
systempromptoverride	No	-	Override the default system prompt
projectspecificcontext	No	-	Additional project context for better results
maxtokens	No	2000	Maximum tokens for API requests
recurse	No	false	Process commits recursively until [commitscribe_signoff]

Warning

The recurse flag will parse your entire commit history and rewrite every commit message until it finds a commit message that contains [commitscribe_signoff]. This is useful for updating commit messages in bulk, but it can be dangerous if you're not careful. Use with caution!

🪄 Examples

OpenAI Configuration

- task: commitscribe@0
  inputs:
      aiprovider: "openai"
      aiproviderkey: "$(OPENAI_API_KEY)"
      model: "gpt-4"

Anthropic Configuration

- task: commitscribe@0
  inputs:
      aiprovider: "anthropic"
      aiproviderkey: "$(ANTHROPIC_API_KEY)"
      model: "claude-3-sonnet-20240229"

Azure OpenAI Configuration

- task: commitscribe@0
  inputs:
      aiprovider: "azure"
      aiproviderkey: "$(AZURE_OPENAI_KEY)"
      azendpoint: "https://your-resource.openai.azure.com/"
      model: "your-deployment-name"

✅ Best Practices

1. Always store API keys as secret variables
2. Use projectspecificcontext to provide domain-specific knowledge
3. Set appropriate maxtokens for your commit message size

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request.