Summary

Fabricate is an experimental research tool for generating synthetic GitHub personas via AI-driven repositories. It targets researchers and engineers studying AI-generated code, GitHub activity patterns, and detection methods, offering a novel way to produce varied, realistic codebases and commit histories for research.

How It Works

The tool uses Anthropic's Claude API to generate unique project concepts, code, and commit histories. It fabricates incremental commits with realistic messages over configurable time periods, creating synthetic GitHub activity that mimics genuine contributions across various languages and complexities. This offers a novel method for generating research data.

Quick Start & Requirements

Installation requires cloning the repo and using uv for setup (uv venv, source .venv/bin/activate, uv pip install -e .). Prerequisites include an Anthropic API Key and a GitHub Personal Access Token (with repo, delete_repo scopes), configured via environment variables or a .env file. Usage is via the fabricate generate CLI command.

Highlighted Details

• Customizable generation: Specify languages, technologies, project categories, repo count, history depth, and commit ranges via CLI flags.
• Diverse project types: Generates CLI tools, web APIs, libraries, data processing utilities, ML projects, and more.
• Varied complexity: Projects range from simple utilities (2-5 files) to complex applications (10-30 files).
• Realistic commit histories: Fabricates timestamped commits with descriptive messages for each repository.

Maintenance & Community

This is an experimental research project. Issues and PRs are welcomed for improvements, but no specific community channels or dedicated maintainer information are provided in the README.

Licensing & Compatibility

Released under the MIT License, this permissive license generally allows for commercial use and integration with closed-source projects without significant restrictions.

Limitations & Caveats

Usage incurs Anthropic API costs and is subject to GitHub API rate limits. Generated code may occasionally fail to compile. The README strongly cautions against fraudulent applications or malicious use, emphasizing its intended purpose for research and education.