Autochitect

Autonomous AI Architect for codebase analysis and C4 reporting.

Installation

npm install -g autochitect

Or run directly via npx:

npx autochitect <repo-url>

Setup

Autochitect requires a Google Gemini API Key. Set it in your environment:

export GOOGLE_API_KEY=your_key_here

Usage

Analyze a repository:

autochitect https://github.com/user/repo

Options

• -t, --token <token>: GitHub Personal Access Token (for private repos).
• -e, --eval: Run in local evaluation mode (uses .test-workspace directory).
• -p, --provider <provider>: LLM Provider (google-genai, google-vertexai, anthropic, openai, ollama).
• -m, --model <model>: Specific LLM model name (e.g., claude-3-5-sonnet, llama3.1).
• -v, --version: Show version number.
• -h, --help: Show help.

Supported Providers

Autochitect will automatically detect your provider based on available environment variables, but you can override it using the --provider flag.

| Provider | Environment Variable | Default Model (High) | |----------|----------------------|----------------------| | Gemini (API) | GOOGLE_API_KEY | gemini-1.5-pro | | Claude | ANTHROPIC_API_KEY | claude-3-5-sonnet | | OpenAI | OPENAI_API_KEY | gpt-4o | | Vertex AI | GOOGLE_APPLICATION_CREDENTIALS | gemini-1.5-pro | | Ollama | OLLAMA_HOST (optional) | llama3.1 |

Output

The agent will generate a report.json in your current working directory. This file can be loaded into the Autochitect Web App for review and validation.