GitWit Agent is a container-based AI agent designed to automate the creation of GitHub repositories and commits based on natural language descriptions. It's targeted at developers and users who want to quickly scaffold projects or implement features without manual coding, leveraging AI to write code that writes code.

How It Works

GitWit operates by executing shell commands within a temporary Docker sandbox. It uses an LLM (defaulting to OpenAI's gpt-3.5-turbo) to generate a plan, which is then translated into a shell script. This script is executed inside the Docker container to perform Git operations and code generation. The agent can create new repositories or add commits to existing ones, with a retrieval-augmented generation approach to select relevant files for context.

Quick Start & Requirements

• Install: Clone the repo, copy .env.example to .env and fill in GITHUB_USERNAME, GITHUB_TOKEN, and OPENAI_API_KEY. Run npm install.
• Prerequisites: NodeJS (v18), Docker, OpenAI API key, GitHub Personal Access Token.
• Run: npm run start for new repos, npm run start -- --branch for new branches.
• Docs: app.gitwit.dev

Highlighted Details

• Automates repository and commit creation via AI.
• Operates in a sandboxed Docker environment for flexibility.
• Supports creating new GitHub repositories or new branches on existing ones.
• Can be configured to use different LLMs via LangChain.
• Examples include generating HTML/JS apps, Python scripts, and ReactJS apps.

Maintenance & Community

The project is maintained by jamesmurdza. Further community or roadmap information is not detailed in the README.

Licensing & Compatibility

The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The agent's effectiveness is dependent on the LLM's ability to generate correct and functional code. Configuration for LLMs other than OpenAI is available on a separate branch. The README does not detail specific limitations regarding supported programming languages or complexity of tasks.