@eldrforge/kodrdriv v0.0.3

KodrDriv

KodrDriv is a powerful utility designed to automatically generate intelligent release notes and change logs from your Git repository. It analyzes commit history, pull requests, and related metadata to create comprehensive, well-structured documentation of your project's evolution. By leveraging advanced parsing and analysis techniques, it helps teams maintain clear visibility into their codebase's development history while reducing the manual effort typically required for changelog maintenance.

Installation

Install KodrDriv globally using npm:

npm install -g @eldrforge/kodrdriv

This will make the kodrdriv command available globally on your system.

Commands

KodrDriv provides two main commands:

Commit Command

Generate intelligent commit messages:

kodrdriv commit

!TIP

Working with Staged Changes

When you have staged changes using git add, the kodrdriv commit command will automatically analyze the diff of your staged changes. This allows you to selectively stage files and generate a commit message that specifically addresses those changes, rather than all uncommitted changes in your working directory.

!TIP

Quick Commit with --sendit

If you trust the quality of the generated commit messages, you can use the --sendit flag to automatically commit your changes with the generated message without review. This is useful for quick, routine changes where you want to streamline your workflow.

Release Command

Generate comprehensive release notes based on changes since the last release:

kodrdriv release

!TIP

Custom Release Range

The kodrdriv release command supports customizing the range of commits to analyze using the --from and --to options. By default, it compares changes between the main branch and HEAD, but you can specify any valid Git reference (branch, tag, or commit hash) for either endpoint. This flexibility allows you to generate release notes for specific version ranges or between different branches.

!TIP

Comparing Releases

You can use the --from and --to options to generate release notes comparing two different releases. For example, to see what changed between v1.0.0 and v1.1.0, you could use kodrdriv release --from v1.0.0 --to v1.1.0. This is particularly useful for creating detailed changelogs when preparing release documentation.

Command Line Options

KodrDriv provides several command line options to customize its behavior:

Basic Options

• --dry-run: Perform a dry run without saving files (default: false)
• --verbose: Enable verbose logging (default: false)
• --debug: Enable debug logging (default: false)
• --version: Display version information

Commit Command Options

• --cached: Use cached diff for generating commit messages
• --sendit: Commit with the generated message without review (default: false)
• --context <context>: Provide additional context (as a string or file path) to guide the commit message generation. This context is included in the prompt sent to the AI and can be used to specify the purpose, theme, or any special considerations for the commit.
• --message-limit <messageLimit>: Limit the number of recent commit messages (from git log) to include in the prompt for context (default: 10). This can help focus the AI on the most relevant recent changes.

OpenAI Configuration

• --openai-api-key <key>: OpenAI API key (can also be set via OPENAI_API_KEY environment variable)
• --model <model>: OpenAI model to use (default: 'gpt-4o-mini')

!NOTE

Security Considerations

The OpenAI API key should be handled securely. While the --openai-api-key option is available, it's recommended to use environment variables instead. Git Intelligent Change automatically loads environment variables from a .env file in your current working directory.

While environment variables are a common approach for configuration, they can still pose security risks if not properly managed. We strongly encourage users to utilize secure credential management solutions like 1Password, HashiCorp Vault, or other keystores to protect sensitive information. This helps prevent accidental exposure of API keys and other credentials in logs, process listings, or environment dumps.

Content Configuration

• -c, --content-types [types...]: Content types to include in the summary (default: 'diff')
  • Available types: 'log', 'diff'
  • Can specify multiple types: --content-types log diff

Instructions

• -i, --instructions <file>: Path to custom instructions file for the AI (default: './.kodrdriv/instructions.md')

Examples

Basic usage with default settings:

kodrdriv commit

Generate a commit message and automatically commit it:

kodrdriv commit --sendit

Generate release notes:

kodrdriv release

Generate a summary including both git log and diff information:

kodrdriv release --content-types log diff

Run in verbose mode with a custom OpenAI model:

kodrdriv commit --verbose --model gpt-4

Use custom instructions from a file:

kodrdriv release --instructions ./my-custom-instructions.md

Use custom context and message limit:

kodrdriv commit --context "Refactoring for performance" --message-limit 5
kodrdriv release --context "Quarterly release, focus on stability" --message-limit 20

Configuration Directory

KodrDriv uses a configuration directory to store custom settings, instructions, and other configuration files. You can specify a custom location using the --config-dir option:

kodrdriv --config-dir ~/custom-kodrdriv-config

By default, the configuration directory is set to .kodrdriv in your current working directory. This directory is created automatically if it doesn't exist.

The configuration directory structure is as follows:

.kodrdriv/
├── instructions/
│   ├── commit.md         # Override for commit instructions
│   ├── commit-pre.md     # Content prepended to default commit instructions
│   ├── commit-post.md    # Content appended to default commit instructions
│   ├── release.md        # Override for release instructions
│   ├── release-pre.md    # Content prepended to default release instructions
│   └── release-post.md   # Content appended to default release instructions
└── ...                   # Other configuration files

Default Instructions

KodrDriv comes with default instructions that guide the AI in generating release notes or change logs. These instructions are defined in the source code:

• Commit Instructions: The default instructions for commit message generation are defined in src/prompt/instructions/commit.ts.

• Release Instructions: The default instructions for release notes generation are defined in src/prompt/instructions/release.ts.

These instruction files contain detailed guidelines for the AI on how to format and structure the output, including examples and specific requirements for different types of changes.

Customizing Instructions

You can override these default instructions in several ways:

1. Command Line Option: Use the --instructions flag to specify a custom instructions file:

   kodrdriv --instructions ./my-custom-instructions.txt

2. Default Location: Even without specifying a command line option, Git Intelligent Change will automatically look for an instructions file at ./.kodrdriv/instructions.md in your current working directory.

3. File Format: While the default file is named instructions.md, you can use any text file with any extension. The content doesn't have to be in Markdown format - any plain text content will work. This gives you flexibility to use your preferred text editor or format for writing instructions.

Prompt Structure

When KodrDriv sends a request to the LLM, it structures the prompt using XML-like tags to organize different components of the input. The prompt is composed of three main sections:

<instructions>
[Your custom instructions or the default instructions]
</instructions>

<log>
[Git log output if --content-types includes 'log']
</log>

<diff>
[Git diff output if --content-types includes 'diff']
</diff>

Each section serves a specific purpose:

• <instructions>: Contains the guidance for the LLM on how to format and structure the output
• <log>: Contains the git log output, providing commit history and messages
• <diff>: Contains the git diff output, showing the actual code changes

Context

KodrDriv can use contextual information about your project to generate more meaningful commit messages and release notes. Context is provided through Markdown files stored in a dedicated directory.

Context Directory Structure

The structure of your context directory is entirely up to you. There are no strict requirements for how you organize your context files - you can structure them in whatever way makes the most sense for your project and team.

Here are two example approaches to organizing context files:

Hierarchical Structure Example

You can organize context in a hierarchical structure with subdirectories for different categories:

.kodrdriv/context/
├── context.md                # Main context file describing sections
├── people/                   # Directory for information about people
│   ├── context.md            # Description of the people section
│   ├── team-members.md       # Information about team members
│   └── contributors.md       # Information about contributors
├── projects/                 # Directory for project information
│   ├── context.md            # Description of the projects section
│   └── project-details.md    # Details about various projects
└── technologies/             # Directory for technical information
    ├── context.md            # Description of the technologies section
    ├── frameworks.md         # Information about frameworks used
    └── libraries.md          # Information about libraries used

Individual Records Example

Alternatively, you can use a flatter structure with individual files for each entity:

.kodrdriv/context/
├── context.md                # Main context file describing sections
├── people/                   # Directory for individual people information
│   ├── context.md            # Description of the people section
│   ├── john-doe.md           # Information specific to John Doe
│   ├── jane-smith.md         # Information specific to Jane Smith
│   └── alex-johnson.md       # Information specific to Alex Johnson

Choose the organization that works best for your needs. The system will process the context files regardless of the structure, as long as they follow the basic Markdown formatting guidelines.

Main Context File

The context.md file in each directory serves as an introduction to that section. The system loads this file first to understand the structure of the information. For example, a context.md file in the people directory might look like:

## People

This section contains subsections that have information about people.

Context Files

After loading the context.md file, the system reads all other Markdown files in the directory. It uses the first header in each file as the name of the section or subsection. For example:

## Team Members

- John Doe: Lead Developer, focuses on backend systems
- Jane Smith: UX Designer, specializes in responsive interfaces
- Alex Johnson: DevOps Engineer, manages deployment pipelines

Context Location

You can specify where to store your context files in two recommended ways:

1. Project Directory: Store context files in your project repository at .kodrdriv/context/. This is useful when the context is specific to the project and should be versioned with the code.

2. gitignore Directory: Alternatively, you can store context in your .gitignore directory if you want to keep it separate from your project files or if the context contains sensitive information that shouldn't be committed to the repository.

To specify a custom context directory, use the --context-dir option:

kodrdriv commit --context-dir ~/my-custom-context

By default, KodrDriv looks for context in the .kodrdriv/context directory within your project.

Configuration Directory

The configuration directory (configDir) allows you to further customize both commit and release instructions by adding pre and post content to the default instructions. This is done by creating additional files in your .kodrdriv/instructions directory:

Release Instructions

1. Pre-Content: Create a file named release-pre.md to add content that will be prepended to the default release instructions.
2. Post-Content: Create a file named release-post.md to add content that should be appended to the default release instructions.

Commit Instructions

1. Pre-Content: Create a file named commit-pre.md to add content that will be prepended to the default commit instructions.
2. Post-Content: Create a file named commit-post.md to add content that should be appended to the default commit instructions.

For example, if you want to add specific formatting requirements before the default release instructions, you could create .kodrdriv/instructions/release-pre.md, and if you want to add instructions to the end of the commit instrucitons, you would have a file in .kodrdriv/instructions/commit-post.md.

Overriding Default Instructions

While the pre and post content files provide a way to extend the default instructions, you can also completely replace them by creating either commit.md or release.md in your .kodrdriv/instructions directory. This gives you full control over the instruction content.

However, please note that completely replacing the default instructions should be done with caution. The default instructions are carefully crafted to:

• Ensure consistent formatting
• Maintain proper context awareness
• Follow best practices for commit messages and release notes
• Handle edge cases and special scenarios

By replacing these instructions entirely, you may lose these benefits and potentially create inconsistencies in your documentation. It's recommended to use the pre and post content files to extend the default instructions rather than replacing them entirely, unless you have a specific need to do so.

To enable instruction overrides, you'll need to use the --overrides flag when running the command.

About the Name

Ski carving and efficient software development have a lot in common. Carving uses edge control to follow a smooth, energy-efficient arc — just like automation uses clean, repeatable scripts to replace manual work. Both are about flow: linking turns or commits without hesitation. As carving unlocks speed and control, automation unlocks scalability and momentum. The result is clean tracks — razor-thin arcs on snow, or tidy diffs in code. And when you've mastered your craft, you don't stop to think about your last move. Your code leaves a clean trail — and your commit message can be automated straight from the diff. And — snowboarders carve too. Different board, same beauty. We won't hold it against you if you're dropping clean edges on a single plank.

Origin Story: kodrdriv

It always happened at the same moment.

You've just spent the entire day in a flow state — the kind that only comes when everything clicks. Whether it was writing code for a critical feature or hammering out chapters of a Markdown or AsciiDoc book, you were locked in. Maybe you were racing the clock to hit a deadline. Maybe you were just up late trying to carve something beautiful out of nothing. Either way, you went right up to the wire, focused, dialed in, exhausted but satisfied.

And then… Git hits you with the meta-question: "What did you do?"

That one prompt — to sum it all up in a commit message — feels totally out of place. It asks you to stop, zoom out, and articulate everything you've just done, right when your brain is at its least reflective. You're not in summary mode. You're still in it. Still shaping. Still carving.

And that's the thing: it sounds silly, like it shouldn't be a real problem. But every developer, every writer who lives in Git knows that exact moment. The friction is real. The context switch is jarring. It's like being asked to narrate your entire ski run after you've blasted through powder, dodged trees, hit the cliff drop — and now you're out of breath, standing at the bottom, being asked to give a PowerPoint.

That's why I built kodrdriv.

It's not just a tool — it's a mindset shift. The idea is simple: you've already carved your line in the snow. Your code is there. Your diffs are real. Instead of making you explain it, kodrdriv uses an LLM to read the trail you left behind and generate a clean, meaningful commit message. One that actually reflects your work — without breaking your flow or making you guess what mattered most.

Whether you're merging branches or writing books, kodrdriv is built for that end-of-day moment when you want to commit and move on — not pause for existential reflection. It reads the line you've drawn, and it helps you push it forward.

Release Command Options

• --from <from>: Branch or reference to generate release notes from
• --to <to>: Branch or reference to generate release notes to
• --context <context>: Provide additional context (as a string or file path) to guide the release notes generation. This context is included in the prompt sent to the AI and can be used to specify the purpose, theme, or any special considerations for the release.
• --message-limit <messageLimit>: Limit the number of recent commit messages (from git log) to include in the release notes prompt (default: 10).

Explanation

• The --context option allows you to inject custom context into the AI prompt, which can help tailor the generated commit message or release notes to your specific needs or project conventions. You can provide a string directly or a path to a file containing the context.
• The --message-limit option controls how many recent commit messages are included in the prompt. Reducing this number can make the summary more focused, while increasing it can provide broader historical context.