AI Code Tracker

Track and analyze the proportion of AI-assisted vs. human-written code in your Git repositories.

Heavily inspired by Aider's Release History

Features

• Analyze Git repositories to determine AI vs. human code contributions
• Track contributions over time with daily or weekly aggregation
• Smart file filtering for relevant source files
• Track AI commit metadata including time-prompting metrics (S/M/L/XL)
• Detailed statistics including lines added/deleted and percentage of AI changes
• Generate interactive visualizations:
  • AI Code Percentage by Day
  • New Lines of Code by Day (AI vs. Human)
  • Time Prompting Breakdown by Day

Installation

# Install from PyPI
pip install ai-code-tracker

# Or with uv
uv add ai-code-tracker

# Or clone the repository
git clone https://github.com/mrmattwright/ai-code-tracker
cd ai-code-tracker
uv sync

Usage

After installation, you can run the tool directly from the command line:

# Basic usage with date range
ai-code-tracker --start-date 2024-01-01 --end-date 2024-03-01

# Group by week instead of day
ai-code-tracker --start-date 2024-01-01 --group-by week

# Generate interactive charts
ai-code-tracker --start-date 2024-01-01 --output-dir ./charts

If you've cloned the repository instead:

# Using uv
uv run contribution_tracker.py --start-date 2024-01-01 --output-dir ./charts

Publishing to PyPI

If you're contributing to this project and need to publish a new version to PyPI:

1. Update the version number in pyproject.toml:

   [project]
   name = "ai-code-tracker"
   version = "0.1.1"  # Increment this version number

2. Build the distribution packages:

   uv run -m build

3. Upload to PyPI:

   uv run -m twine upload dist/*

   You'll need PyPI credentials or a token. For automated uploads, create a .pypirc file:

   [pypi]
   username = __token__
   password = your-token-here

Docker Usage

You can either build the image locally or use the pre-built image from Docker Hub:

# Using pre-built image
docker run -v $(pwd):/app mrmattwright/ai-code-tracker --start-date 2025-01-01

# Or build locally
docker build -t ai-code-tracker .

# Run with basic options (using local build)
docker run -v $(pwd):/app ai-code-tracker --start-date 2025-01-01

# Run with all options
docker run -v $(pwd):/app ai-code-tracker \
  --start-date 2025-01-01 \
  --end-date 2025-03-01 \
  --group-by week \
  --output-dir ./charts

# Run tests in container
docker run ai-code-tracker pytest

Git Configuration

To track AI vs human contributions, you need to set up your Git configuration to use different author information when committing AI-generated code. The tool identifies AI contributions by matching the author email in DEFAULT_AI_COMMITTER (default: "llm [email protected]").

Setting Up Multiple Git Configs

Create a .gitconfig for your AI commits:

[user]
    name = llm
    email = [email protected]

And one for your human commits:

[user]
    name = Your Name
    email = [email protected]

You can switch between configs using:

# For AI commits
git config --local include.path ../.gitconfig-ai

# For human commits
git config --local include.path ../.gitconfig-human

Or set per-command author:

# For one-off AI commits
git commit -m "AI generated change" --author="llm <[email protected]>"

Configuring the AI Committer

You can configure the AI committer via the AI_COMMITTER environment variable. By default, it is set to "llm <[email protected]>", but you can override this without modifying For example, if you're using direnv, add the following line to your .envrc:

export AI_COMMITTER="your_name <[email protected]>"

Commit Message Format

When committing AI-generated code, you need to include a time-prompting metric in your commit message:

git commit -m "feat: Add new feature

time-prompting: M"

Valid time-prompting values are:

• S: Small (< 5 minutes)
• M: Medium (5-15 minutes)
• L: Large (15-30 minutes)
• XL: Extra Large (> 30 minutes)

Enforcing Commit Message Format

You can enforce the time-prompting format for AI commits by adding a Git commit hook. Create .git/hooks/commit-msg:

#!/bin/bash

COMMIT_MSG_FILE="$1"
COMMITTER_EMAIL=$(git config --get user.email)
TARGET_EMAIL="[email protected]"

# Only enforce for AI commits
if [[ "$COMMITTER_EMAIL" == "$TARGET_EMAIL" ]]; then
    if ! grep -Eq "time-prompting:\s*(S|M|L|XL)\b" "$COMMIT_MSG_FILE"; then
        echo "ERROR: Commit message must include 'time-prompting: S|M|L|XL'"
        exit 1
    fi
fi

exit 0

Make it executable:

chmod +x .git/hooks/commit-msg

Options

• --start-date: Start date for analysis (YYYY-MM-DD) [required]
• --end-date: End date for analysis (YYYY-MM-DD) [default: current date]
• --group-by: Group results by 'day' or 'week' [default: day]
• --output-dir: Directory to save charts (enables chart generation) [optional]
• --chart-format: Chart output format (html, png, pdf) [default: html]

Sample Output

{
    "date": "2025-01-30",
    "ai_commits": 2,
    "total_commits": 7,
    "human_commits": 5,
    "ai_lines_added": 302,
    "ai_lines_deleted": 2,
    "ai_total_changes": 304,
    "human_lines_added": 337,
    "human_lines_deleted": 0,
    "human_total_changes": 337,
    "percentage_total_changes_ai": 47.43,
    "time_prompting_S": 1,
    "time_prompting_M": 1,
    "time_prompting_L": 0,
    "time_prompting_XL": 0
}

Generated Charts

When using the --output-dir option, the tool generates three interactive charts:

1. AI Code Percentage by Day: Line chart showing the percentage of AI-written code over time
2. New Lines of Code by Day: Stacked bar chart comparing AI and human-added lines
3. Time Prompting Breakdown: Stacked bar chart showing the distribution of time-prompting categories (S/M/L/XL)

Charts are generated in HTML format by default for maximum interactivity. PNG and PDF formats are also supported if the kaleido package is installed.

File Filtering

By default, the tool analyzes files with these extensions:

• .js
• .py
• .scm
• .sh
• Dockerfile
• .md
• .github/workflows/*.yml

Excluded patterns:

• tests/fixtures/watch/*
• **/prompts.py

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

License

MIT