Benito Code Generator is a workspace-based prompt-to-website studio built for fast frontend concepting. A user describes what they want, the app generates editable index.html, style.css, and script.js files, and the project can be reviewed, repaired, polished, or previewed live without leaving the workspace.
Problem
Early website ideas usually break in one of two ways: either they stay stuck as plain chat output, or they become generic templates that are hard to refine and even harder to present professionally.
Solution
Benito Code Generator turns prompting into an editable working session. Instead of stopping at generated text, it keeps the result inside a small coding workspace with diagnostics, direct file editing, and a live preview that opens only when the user wants it.
Who it is for
- Students building portfolio-grade web concepts quickly
- Freelancers shaping ideas into frontend drafts for clients
- Founders who need fast visual prototypes before full engineering
- Developers who want a lightweight prompt-to-code playground
- Showcase page: georgebenedict77.github.io/benito-code-generator
- Static generated sample: georgebenedict77.github.io/benito-code-generator/demo/
- Release notes: v1.0.1
The full AI workspace runs locally because it depends on a Node server and optional OpenAI credentials. The GitHub Pages site is the public showcase and sample-output companion.
- Accepts natural-language prompts for any website style or business category
- Generates browser-ready
index.html,style.css, andscript.js - Lets users edit generated files directly in a code workspace
- Shows diagnostics and review notes for the current project
- Runs repair, responsive, polish, and review actions
- Opens a live preview on demand instead of permanently taking screen space
- Supports a local studio mode and an optional OpenAI-backed mode
flowchart LR
User["User prompt or file edit"] --> UI["Browser workspace UI"]
UI --> Server["Node HTTP server"]
Server --> WM["Workspace manager"]
WM --> Files["Workspace files on disk"]
WM --> Local["Local generation engine"]
WM --> OpenAI["OpenAI Responses API (optional)"]
Files --> Preview["Live preview document"]
Preview --> UI
npm install
npm run startThen open:
http://localhost:3000
If 3000 is busy, the server will keep trying the next available local port.
Create a local .env or set environment variables before starting the server:
$env:OPENAI_API_KEY="your_api_key_here"
$env:OPENAI_MODEL="gpt-5.1"
npm run startWithout an API key, the app stays fully usable in local studio mode.
This repository now includes a lightweight smoke-check pipeline:
- JavaScript syntax validation for the server and browser files
- Server boot validation through the
/healthendpoint - Workspace API validation through
/api/workspace - Preview document validation through
/preview/project-001
Run locally:
npm run checkGitHub Actions runs the same validation on pushes and pull requests.
BENJOJI_BUILD AI/
|-- .github/workflows/ci.yml
|-- docs/
| |-- index.html
| |-- styles.css
| |-- demo/
| |-- positioning.md
| `-- release-notes/
|-- public/
| |-- index.html
| |-- app.css
| `-- app.js
|-- scripts/check.js
|-- server/
| |-- ai.js
| |-- fileWriter.js
| |-- openaiClient.js
| |-- previewServer.js
| `-- workspaceManager.js
|-- workspaces/project-001/
|-- CHANGELOG.md
|-- LICENSE
|-- package.json
`-- server.js
- The app separates prompting, editable files, diagnostics, and preview into a real workspace flow
- It avoids vendor lock-in by generating plain frontend files instead of a framework-specific sandbox
- It already supports an optional OpenAI runtime while preserving a local offline fallback
- The repo includes CI, release notes, licensing, and public showcase material for professional presentation
Released under the MIT License.