💠 Real No Code

The UI is the Runtime. An architecture where the backend doesn't exist and the interface is a hallucination.

---

🚀 The Vision

Most "No-Code" tools are just visual programming with more friction — you're still building deterministic logic.

Real No Code is a subversive Proof of Concept that treats the LLM as a just-in-time rendering engine. Instead of a fixed UI, the application "hallucinates" the necessary interface, logic, and styling in real-time based on user intent. This moves software from Static Pixels to Fluid Probability.

---

🏗 How It Works

1. Intent Capture — The user provides a prompt (e.g., "A CRM for a pet shop").
2. JIT Generation — The orchestrator sends the intent to a provider (Gemini, Copilot, or Local LLM).
3. The Hallucination — The LLM returns a standalone HTML/JS/CSS blob.
4. Zero Middleware — The server serves this blob directly to the browser.
5. Recursive Exploration — Clicking elements on the generated page triggers a new request, evolving the interface dynamically.

---

🛠 Project State: Research Laboratory

This is a Research Preview. The goal is to prove that a zero-middleware architecture is possible by rendering LLM output directly to the browser.

• Short Term: A lightning-fast "Business Idea Generator."
• Long Term: Solving the "Security-Creativity Paradox" to allow generated interfaces to run safely in the wild.

See ROADMAP.md for the current (non-final) trajectory.

---

🔌 Supported Providers

Provider	Description
gemini	Google Gemini API — recommended for speed
copilot	GitHub Copilot CLI
local	LM Studio / OpenAI-compatible local endpoint — total privacy

Select a provider in the launcher UI or pass it as provider in POST requests.

---

🏃 Getting Started

1. Configuration

Copy .env.example to .env and fill in the values you need:

AI_DEFAULT_PROVIDER=gemini
AI_GEMINI_API_KEY=your-key-here

# Optional: local inference
AI_LM_STUDIO_BASE_URL=http://localhost:1234
AI_LM_STUDIO_MODEL=local-model

You can also use real OS environment variables instead of .env.

2. Run with Docker Compose (Recommended)

docker compose up --build

Open http://localhost:8080 to begin.

Local LM Studio: when using AI_DEFAULT_PROVIDER=local, the compose file automatically points AI_LM_STUDIO_BASE_URL to http://host.docker.internal:1234 — no extra configuration needed.

3. Run with Docker

docker build -t real-no-code .
docker run --rm -p 8080:8080 \
  -e AI_DEFAULT_PROVIDER=gemini \
  -e AI_GEMINI_API_KEY=your-key-here \
  real-no-code

4. Run with Gradle (Local)

gradlew.bat test
gradlew.bat bootRun

Open http://localhost:8080.

---

🏛 Architecture

• Zero Backend — No traditional database or server-side state logic.
• Pure HTML/JS — No intermediate frameworks (React/Vue) required for the end-user.
• Local-First — Designed to bridge to local LLMs (Ollama/LM Studio) for private execution.

Controllers

Route	Controller	Description
GET /	CopilotController	Main launcher UI
POST /generate	CopilotController	Generate a page from form input
GET /business	BusinessController	Business-mode UI
POST /business/generate	BusinessController	Generate a business idea page

---

🛡 The Security "Moat"

Executing LLM-generated HTML/JS is a security nightmare (XSS, prompt injection). Real No Code is a laboratory to try solving this. Currently designed for local-only prototyping and internal research.

---

⚖️ License

Licensed under GPL-3.0.

---

⚠️ Disclaimer

This project executes AI-generated code. Do not deploy to production. It is a tool for exploring the boundaries of software generation. Use at your own risk.