Ghost AI is an interactive system architecture builder designed for software teams. It addresses the challenge of visualizing and documenting complex system designs by leveraging AI to autonomously generate architecture diagrams from natural language prompts. The system allows for real-time collaborative refinement of these diagrams and automatically converts them into comprehensive technical specifications, streamlining the design and documentation process for development teams.

How It Works

The core of Ghost AI is an agentic planning application built with Next.js, React, and TypeScript. Users submit natural-language prompts, which are processed by a Google Gemini-powered AI agent. This agent autonomously places nodes and edges onto a shared React Flow canvas in real-time, facilitated by Trigger.dev for background tasks and Liveblocks for real-time collaboration features like synchronized state and presence. Once the visual architecture is finalized, a secondary AI task converts the graph into a detailed, multi-page Markdown technical specification.

Quick Start & Requirements

To set up locally, clone the repository (git clone https://github.com/adrianhajdin/ghost-ai.git), navigate into the directory (cd ghost-ai), and install dependencies (npm install). A .env file is required in the project root, containing credentials for Clerk, Liveblocks, Trigger.dev, Google AI Studio (for Gemini), and a PostgreSQL database URL. The project also requires Git, Node.js, and npm. Running the development server is done via npm run dev, and background AI tasks are initiated with npx trigger.dev@latest dev. A detailed step-by-step tutorial is available on the JavaScript Mastery YouTube channel.

Highlighted Details

• AI Architecture Agent: Utilizes Gemini to draw nodes and edges on a live canvas in real-time.
• Multiplayer Canvas: Offers full real-time collaboration powered by Liveblocks, including synchronized state, live cursors, and presence avatars.
• AI Spec Generation: A one-click feature converts the visual graph into a detailed Markdown technical specification using a Gemini-powered Trigger.dev task.
• Auto-Save Canvas: The canvas state is debounced and saved locally every 3 seconds of inactivity.
• Project Management: Includes a sidebar for creating projects and managing active rooms.

Maintenance & Community

The project highlights an active Discord community with over 50,000 members, available for assistance and bug reporting. The development is closely tied to tutorials provided by the JavaScript Mastery YouTube channel.

Licensing & Compatibility

The README does not explicitly state the project's license. This omission makes it impossible to determine compatibility for commercial use or closed-source linking without further clarification.

Limitations & Caveats

Setting up the project requires obtaining and configuring API keys for multiple external services (Clerk, Liveblocks, Trigger.dev, Google AI Studio) and setting up a PostgreSQL database, which can be a significant barrier to entry. The project is presented as a tutorial, suggesting it may be more illustrative than production-hardened. The absence of a clear open-source license is a critical limitation for adoption decisions.