GitHub - apideck-libraries/describer: Analyze codebases using AI - generate architectural overviews, documentation, explanations, bug reports and more

Describer

Analyze codebases using AI - generate architectural overviews, documentation, explanations, bug reports and more.

What It Does

Describer scans all files in a directory and uses Google's Gemini AI to generate an architectural overview or any other analysis of the codebase.

For example, you can:

• Generate markdown architectural overviews
• Get a summary of the codebase
• Analyze code patterns
• Document the codebase structure
• Find potential bugs or issues
• Generate test ideas

Prerequisites

• Python 3.6+
• A Google AI Studio API key for Gemini

Installation

Note: This package is currently in early development and not yet published to PyPI.

Install directly from GitHub:

pip install git+https://github.com/apideck-libraries/describer.git

Then set up your Gemini API key:

When prompted, enter your API key from Google AI Studio (https://makersuite.google.com/app/apikey).

Usage

# Generate an architectural overview of a directory
describer path/to/directory

# Generate a specific type of analysis
describer path/to/directory "summarize this codebase in 3 bullet points"
describer path/to/directory "identify potential bugs in this code"
describer path/to/directory "document this codebase as if for a new developer"

# Use a specific model
describer -m gemini-1.5-pro-latest path/to/directory

# Save output to a markdown file
describer path/to/directory -o output.md
describer path/to/directory "code review" --output review-results.md

# Include files that would normally be excluded by .gitignore
describer path/to/directory --ignore-gitignore

# Exclude specific file patterns
describer path/to/directory --exclude "*.test.ts"
describer path/to/directory --exclude "*.min.js"

# Don't show file count information
describer path/to/directory --quiet

By default, Describer respects .gitignore rules when scanning files, which means files and directories listed in your .gitignore file(s) will be excluded from the analysis. Use the --ignore-gitignore flag if you want to include these files.

You can also use the --exclude flag to exclude specific files or patterns from the analysis. This is useful for filtering out test files, generated code, or other files that might not be relevant to your analysis.

The tool will display the number of files being analyzed, which can help you confirm that the correct files are being included and that .gitignore rules are being applied as expected.

How It Works

This project combines these tools:

1. files-to-prompt: Collects all files in a directory and formats them for processing by an LLM.
2. llm: A command-line tool for interacting with various LLMs.
3. llm-gemini: A plugin for the llm tool to use Google's Gemini models.

Development

If you want to contribute to this project, follow these steps to set up your development environment:

1. Clone the repository:

   git clone https://github.com/apideck-libraries/describer.git
   cd describer

2. Create a virtual environment:

3. Activate the virtual environment:

   • On macOS/Linux:
   • On Windows:

4. Install the package in development mode:

5. Install development dependencies:

   pip install build twine pytest

6. Set up your Gemini API key:

7. Run tests:

   python -m unittest discover tests

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Credits

This library builds on the incredible work of Simon Willison and his open-source tools. While this CLI primarily strings together three of his existing tools, the goal is to make them more accessible and streamlined for specific use cases to analyze large codebases.