Descriptions:
Stephanie Nyarko demonstrates how to build a chat-driven website generator using n8n, Google Gemini, and GitHub Pages — turning a single natural language prompt into a fully deployed, shareable website URL with no drag-and-drop builders or templates.
The n8n workflow starts with a chat trigger node where users type descriptions like ‘create a one-page wellness site with soft pastels and a FAQ section.’ An AI agent node connected to Google Gemini 3.0 processes the prompt and, via a structured output parser, returns a clean JSON object with discrete fields: site description, tone, style, extra requirements, and a URL slug. A second HTTP request node hits the Gemini API directly with a front-end developer system prompt to generate a complete, self-contained index.html file (pure HTML and CSS, no external dependencies), which is then pushed to GitHub Pages to produce a live URL.
Nyarko covers practical configuration details including how to obtain and attach a Google Gemini API key, why the structured output parser is critical for passing clean field values between nodes, and how to set the correct POST request headers and body parameters. A free workflow template is available via her School community, and the video also touches on how to layer a custom frontend UI over the workflow for a more polished user-facing product.
📺 Source: Stephanie Nyarko · Published December 13, 2025
🏷️ Format: Hands On Build
