ai-web-builder

AI Web Builder

Open-source AI-powered website builder — Generate production-ready websites from natural language prompts using local-first AI. Supports Ollama, llama.cpp, OpenAI GPT-4o, Anthropic Claude, Google Gemini, Groq, DeepSeek, Mistral AI, Together AI, and OpenRouter. Desktop app (Windows EXE, macOS DMG, Linux AppImage) and browser-based web app included. Local-first 13-layer architecture runs entirely offline with no API keys required for basic usage. Features website cloning from URLs, multi-source image generation (ComfyUI, SVG, stock APIs), Docker deployment, and Unraid Community Apps support.

Overview
Features
Architecture
Quick Start
Build Instructions
- Windows
- macOS
- Linux
AI Provider Setup
13-Layer Architecture
Templates
Environment Variables
Troubleshooting
License

Overview

AI Web Builder is a universal, open-source website generator that transforms natural language descriptions into fully functional, production-grade websites. Built by synthesizing patterns from 46+ AI website builder projects, it combines the best features of each into a single unified platform.

Key differentiators:

Local-first — Run entirely offline with Ollama or llama.cpp. No API keys required for basic usage.
13 production layers — Every generated site includes configurable Frontend, API, Database, Auth, Deployment, CI/CD, Security, Rate Limiting, Caching, Load Balancing, Error Tracking, and Recovery layers.
Dual delivery — Use it in your browser or as a native desktop application.
Provider-agnostic — Switch between 10+ AI providers without changing your workflow.

Features

🤖 AI-Powered Generation

Transform plain English descriptions into complete, working websites. Describe what you want — “a SaaS landing page with pricing tiers and a contact form” — and AI generates every file.

🏗️ 13-Layer Architecture

Every generated website includes all 13 production layers, fully configurable before generation. Toggle layers on/off, adjust settings per layer.

🔌 10+ AI Providers

Local: Ollama, llama.cpp, ComfyUI (runs offline, zero cost)
Cloud: OpenAI (GPT-4o), Anthropic (Claude), Google (Gemini), Groq, DeepSeek, Mistral AI, Together AI, OpenRouter
BYOK: Bring your own API key for any provider

🌐 Website Cloning

Clone existing websites as structural reference for AI generation. Crawls pages (BFS), extracts layout/structure patterns, and feeds them as context to the AI prompt — no content copying, just architectural inspiration.

🖼️ Image Generation

Stock photos: Pexels and Unsplash APIs with automatic fallback to styled placeholders
Cloud AI: OpenAI DALL-E / GPT Image for AI-generated images
Local AI: ComfyUI (SDXL, SD1.5) for fully offline image generation
SVG Generation: 8 zero-dependency SVG generators (gradient, dots, geometric, waves, grid, crosshatch, abstract, hero) — no API key required

📦 10 Starter Templates

Pre-built templates for SaaS landing pages, e-commerce stores, creative portfolios, blogs, documentation sites, agency websites, local businesses, and more.

⚙️ Fully Customizable Settings

Eight categories of settings accessible from the settings panel:

General, Providers, Generation, Layers, Assets, Editor, Export, Advanced

🖥️ Desktop + Web + Docker

Web app: Next.js 15, accessible from any browser
Desktop app: Electron (Next.js standalone output), with native file dialogs, system menus, and offline capability
Cross-platform: Windows EXE (NSIS installer, 143 MB), macOS DMG, Linux AppImage
Docker: Multi-stage Dockerfile, production docker-compose.yml (web + Ollama sidecar with GPU passthrough), dev docker-compose.dev.yml (hot reload)
Unraid: Community Applications template with 20+ configurable environment fields

Architecture

ai-web-builder/
├── apps/                          # Application entrypoints
│   ├── web/                       # Next.js 15 browser application
│   │   ├── src/
│   │   │   ├── app/               # App router pages
│   │   │   └── components/        # React components
│   │   └── package.json
│   └── desktop/                   # Electron desktop application
│       ├── electron/              # Main process & preload scripts
│       │   ├── main.js            # Electron main process
│       │   └── preload.js         # Context bridge for IPC
│       └── package.json
├── packages/                      # Shared libraries
│   ├── core/                      # Engine: providers, generator, layers, assets, settings
│   │   └── src/
│   │       ├── providers/         # 10 AI provider integrations
│   │       ├── generator/         # Prompt builder, code engine, templates
│   │       ├── layers/            # 13-layer definition & applier
│   │       ├── assets/            # Image generation (ComfyUI, SVG, stock APIs)
│   │       ├── clone/             # Website cloning (fetcher, parser, crawler)
│   │       ├── settings/          # Settings schema & defaults
│   │       └── templates/         # 10 starter template definitions
│   ├── ui/                        # Shared React UI components
│   └── config/                    # Shared configuration & constants
├── clones/                        # 46 reference implementation repos
├── scripts/
│   └── setup.ps1                  # PowerShell setup helper
├── package.json                   # Root workspace config
├── pnpm-workspace.yaml            # pnpm workspace definition
└── turbo.json                     # Turborepo pipeline

Quick Start

Prerequisites

Node.js >= 24.0.0
pnpm >= 10.0.0 (install via npm install -g pnpm)
Git (for cloning)
Optional: Ollama for local AI inference

Install & Run

# 1. Clone the repository
git clone https://github.com/wildfirebill-gen-ai-web/ai-web-builder.git
cd ai-web-builder

# 2. Install dependencies
pnpm install

# 3. Start the web app (opens at http://localhost:3000)
pnpm web

# 4. Or start the desktop app
pnpm desktop

# 5. Or run via Docker (web + Ollama)
docker compose up -d

Web app: http://localhost:3000
Docker: docker compose up -d (includes Ollama sidecar)
For local AI providers, ensure Ollama is running (ollama serve or use Docker)

Build Instructions

Build standalone executables for distribution on all platforms.

Windows Build

Build a native Windows .exe installer (NSIS):

# Requirements: Windows 10+, Node.js 24+, pnpm

# Build the desktop EXE
pnpm desktop:build

# Or run the build script
.\scripts\setup.ps1 -BuildDesktop

Output: apps/desktop/release/AI Web Builder Setup X.X.X.exe

Windows build flags:

# 64-bit only (default)
pnpm --filter @awb/desktop build:win

# 32-bit
pnpm --filter @awb/desktop exec electron-builder --win --ia32

# Both architectures
pnpm --filter @awb/desktop exec electron-builder --win --x64 --ia32

Portable version (no installer):

pnpm --filter @awb/desktop exec electron-builder --win portable

macOS Build

Build a native macOS .dmg or .app bundle:

# Requirements: macOS 12+, Node.js 24+, pnpm, Xcode CLI tools

# Build DMG installer
pnpm desktop:build

# Or specifically:
pnpm --filter @awb/desktop build:mac

Output: apps/desktop/release/AI Web Builder-X.X.X.dmg

macOS build options:

# DMG (default)
pnpm --filter @awb/desktop exec electron-builder --mac

# App bundle only (no DMG)
pnpm --filter @awb/desktop exec electron-builder --mac --config.mac.target=zip

# Apple Silicon + Intel universal binary
pnpm --filter @awb/desktop exec electron-builder --mac --universal

# Notarize for distribution (requires Apple Developer account)
# Set CSC_LINK and CSC_KEY_PASSWORD environment variables
pnpm --filter @awb/desktop exec electron-builder --mac --publish=always

Linux Build

Build a native Linux .AppImage or .deb package:

# Requirements: Linux (Ubuntu 20.04+, Fedora 36+, etc.), Node.js 24+, pnpm

# Build AppImage (portable)
pnpm desktop:build

# Or specifically:
pnpm --filter @awb/desktop build:linux

Output: apps/desktop/release/AI Web Builder-X.X.X.AppImage

Linux build options:

# AppImage (portable, default)
pnpm --filter @awb/desktop exec electron-builder --linux

# Debian/Ubuntu package
pnpm --filter @awb/desktop exec electron-builder --linux deb

# Red Hat/Fedora package
pnpm --filter @awb/desktop exec electron-builder --linux rpm

# Snap package
pnpm --filter @awb/desktop exec electron-builder --linux snap

# All formats
pnpm --filter @awb/desktop exec electron-builder --linux AppImage deb rpm

Linux dependencies for building:

# Ubuntu/Debian
sudo apt-get install -y rpm fakeroot dpkg-dev

# Fedora
sudo dnf install -y rpm-build dpkg-dev

# Arch
sudo pacman -S rpm-tools dpkg

Docker

Run the AI Web Builder web app and Ollama in containers. The desktop app (Electron) cannot run in Docker — use the web app via Docker instead.

Prerequisites

Docker 24+
Docker Compose v2+
GPU support (optional): NVIDIA Container Toolkit for GPU-accelerated Ollama

Production Stack

Starts the web app on http://localhost:3000 with an Ollama sidecar.

docker compose up -d

After starting, pull a model into Ollama:

docker exec ollama ollama pull qwen2.5-coder

Development Stack

Hot-reloading web app with Ollama sidecar. Source code is mounted from your host.

docker compose -f docker-compose.dev.yml up -d

Web App Only (Standalone)

Run the web app without Ollama (use cloud providers instead):

docker build -t ai-web-builder .
docker run -p 3000:3000 \
  -e OPENAI_API_KEY=sk-... \
  ai-web-builder

With Cloud Providers

Pass API keys via .env file or environment variables:

# Using .env file (auto-loaded by docker compose)
echo "OPENAI_API_KEY=sk-..." >> .env
docker compose up -d

# Or pass inline
OPENAI_API_KEY=sk-... docker compose up -d

GPU Acceleration for Ollama

Enable GPU inference inside the container:

# docker-compose.yml (already configured)
ollama:
  deploy:
    resources:
      reservations:
        devices:
          - driver: nvidia
            count: all
            capabilities: [gpu]

Requires NVIDIA Container Toolkit.

Docker Compose Reference

Command	Description
`docker compose up -d`	Start production stack
`docker compose -f docker-compose.dev.yml up -d`	Start dev stack with hot reload
`docker compose down`	Stop all containers
`docker compose logs -f`	Follow logs
`docker exec ollama ollama pull <model>`	Pull a model into Ollama
`docker compose build --no-cache web`	Rebuild web image from scratch

Unraid Deployment

A Community Applications template is at unraid/templates/ai-web-builder.xml. The image is automatically built and pushed to ghcr.io/wildfirebill-gen-ai-web/ai-web-builder by CI — no manual build needed.

Install: Open the Apps tab in Unraid → search “AI Web Builder” → Install.

Manual install: Copy unraid/templates/ai-web-builder.xml to /boot/config/plugins/dockerMan/templates-user/ on your Unraid server, then add the container from the Docker UI.

Cross-Platform Build (CI)

Pre-configured GitHub Actions workflows in .github/workflows/:

Workflow	Triggers	Output
`ci.yml`	PRs and pushes to `main`	TypeScript typecheck + lint across all packages
`docker.yml`	PRs/main pushes and semver tags `v*`	Multi-arch Docker image on GHCR (latest, semver, date tags)
`build.yml`	Published releases or manual dispatch	Windows EXE (NSIS), macOS DMG, Linux AppImage uploaded as artifacts

AI Provider Setup

Local Providers (No API Key Required)

Provider	Setup	Default URL
Ollama	Install Ollama, pull a model (`ollama pull qwen2.5-coder`)	`http://localhost:11434`
llama.cpp	Build/install llama.cpp, start server (`./server`)	`http://localhost:8080`

Cloud Providers (API Key Required)

Set the corresponding environment variable:

Provider	Env Variable	Models	Get Key
OpenAI	`OPENAI_API_KEY`	GPT-4o, GPT-4o-mini, o3-mini	platform.openai.com
Anthropic	`ANTHROPIC_API_KEY`	Claude Sonnet 4, Claude 3.5 Sonnet, Haiku, Opus	console.anthropic.com
Google Gemini	`GEMINI_API_KEY`	Gemini 2.5 Pro, 2.5 Flash, 1.5 Pro	aistudio.google.com
Groq	`GROQ_API_KEY`	LLaMA 3.3, Mixtral, DeepSeek R1	console.groq.com
DeepSeek	`DEEPSEEK_API_KEY`	DeepSeek Chat, DeepSeek Reasoner	platform.deepseek.com
Mistral AI	`MISTRAL_API_KEY`	Mistral Large, Small, Codestral	console.mistral.ai
Together AI	`TOGETHER_API_KEY`	LLaMA 3.3, DeepSeek V3, Mixtral	together.ai
OpenRouter	`OPENROUTER_API_KEY`	200+ models (unified API)	openrouter.ai

Image Provider Setup

Provider	Env Variable	Get Key
Pexels	`PEXELS_API_KEY`	pexels.com/api
Unsplash	`UNSPLASH_ACCESS_KEY`	unsplash.com/developers
OpenAI Images	`OPENAI_API_KEY`	platform.openai.com
ComfyUI (local)	`COMFYUI_URL` (optional, default `http://localhost:8188`)	github.com/comfyanonymous/ComfyUI
SVG Generator (local, zero-deps)	None required	Built-in (8 generators: gradient, dots, geometric, waves, grid, crosshatch, abstract, hero)

All image providers fall back to styled SVG placeholders when no API key is configured.

13-Layer Architecture

Every website generated by AI Web Builder includes 13 configurable architecture layers. Each layer produces production-ready code files.

#	Layer	Category	Description	Generated Files
1	Frontend Foundations	Presentation	UI framework, styling, components, responsive design, accessibility	`layout.tsx`, `globals.css`, `tailwind.config.ts`
2	APIs & Backend Logic	Backend	API routes, server actions, business logic, webhooks	`api/route.ts`, `middleware.ts`
3	Database & Storage	Data	Schema, migrations, file storage, data seeding	`schema.prisma`, `seed.ts`
4	Auth & Permissions	Security	Authentication, RBAC, session management	`auth.ts`, `roles.ts`, `AuthGuard.tsx`
5	Hosting & Deployment	Operations	Hosting config, deployment scripts, domain setup	`vercel.json`, `Dockerfile`, `.env.example`
6	Cloud & Compute	Operations	Serverless functions, containers, GPU compute	`functions/`, `Dockerfile.cloud`
7	CI/CD & Version Control	Operations	Pipelines, testing, linting, automated releases	`.github/workflows/ci.yml`, `vitest.config.ts`
8	Security & RLS	Security	Security headers, CSP, row-level security, CSRF	`middleware.ts`, `security-headers.ts`, `rls-policies.sql`
9	Rate Limiting	Security	API throttling, abuse prevention	`rate-limit.ts`, `middleware.ts`
10	Caching & CDN	Performance	CDN config, cache strategy, image optimization	`cache.ts`, `next.config.ts`
11	Load Balancing & Scaling	Performance	Auto-scaling, health checks, sticky sessions	`health.ts`, `scaling-config.ts`
12	Error Tracking & Logs	Observability	Structured logging, error monitoring, metrics	`logger.ts`, `error-boundary.tsx`
13	Availability & Recovery	Operations	Backups, disaster recovery, uptime monitoring	`backup.ts`, `recovery.ts`

Templates

Template	Type	Pages	Best For
SaaS Landing Page	`landing-page`	home, features, pricing, faq	Marketing sites
SaaS Dashboard	`saas`	dashboard, analytics, users, billing	Web applications
E-Commerce Store	`ecommerce`	products, cart, checkout, orders	Online stores
Creative Portfolio	`portfolio`	projects, gallery, about, contact	Designers, artists
Blog / Magazine	`blog`	blog posts, categories, tags, search	Content sites
Web App (CRUD)	`webapp`	dashboard, items, settings	Data management
Documentation Site	`documentation`	guides, API reference, examples	Open-source projects
AI Product Landing	`landing-page`	demo, capabilities, pricing	AI/ML products
Agency Website	`multi-page`	services, case studies, team	Agencies
Local Business	`multi-page`	services, reviews, gallery, contact	Small businesses

Environment Variables

Create a .env file in any app directory:

# === AI PROVIDERS ===
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GEMINI_API_KEY=...
GROQ_API_KEY=gsk_...
DEEPSEEK_API_KEY=sk-...
MISTRAL_API_KEY=...
TOGETHER_API_KEY=...
OPENROUTER_API_KEY=sk-or-...

# === IMAGE PROVIDERS ===
PEXELS_API_KEY=...
UNSPLASH_ACCESS_KEY=...

# === LOCAL PROVIDERS (optional overrides) ===
OLLAMA_URL=http://localhost:11434
LLAMACPP_URL=http://localhost:8080
COMFYUI_URL=http://localhost:8188

# === APP CONFIG ===
NEXT_PUBLIC_APP_URL=http://localhost:3000

Troubleshooting

Problem	Solution
`pnpm install` fails	Ensure Node.js >= 24 is installed. Run `node --version` to check.
`OllamaProvider` connection error	Start Ollama: `ollama serve`. Verify at `http://localhost:11434`.
Electron build fails on Windows	Install Visual Studio Build Tools with “Desktop development with C++” workload.
macOS “unidentified developer” warning	Run `xattr -dr com.apple.quarantine /path/to/AI\ Web\ Builder.app`
Linux AppImage not running	Install FUSE: `sudo apt install fuse libfuse2`
Port 3000 already in use	Set `PORT=3001 pnpm web` or modify in `apps/web/package.json`
TypeScript errors after pull	Run `pnpm install` to ensure dependencies are up to date.
Blank screen in desktop app	Open DevTools (View → Toggle DevTools) and check Console for errors.

License

MIT — See LICENSE for details.

Built from patterns across 46 open-source AI website builder projects.

This site is open source. Improve this page.

ai-web-builder

AI Web Builder

Table of Contents

Overview

Features

🤖 AI-Powered Generation

🏗️ 13-Layer Architecture

🔌 10+ AI Providers

🌐 Website Cloning

🖼️ Image Generation

📦 10 Starter Templates

⚙️ Fully Customizable Settings

🖥️ Desktop + Web + Docker

Architecture

Quick Start

Prerequisites

Install & Run

Build Instructions

Windows Build

macOS Build

Linux Build

Docker

Prerequisites

Production Stack

Development Stack

Web App Only (Standalone)

With Cloud Providers

GPU Acceleration for Ollama

Docker Compose Reference

Unraid Deployment

Cross-Platform Build (CI)

AI Provider Setup

Local Providers (No API Key Required)

Cloud Providers (API Key Required)

Image Provider Setup

13-Layer Architecture

Templates

Environment Variables

Troubleshooting

License