copilotkit__copilotkit/examples/showcases/scene-creator

Scene Creator - CopilotKit + LangGraph + Gemini 3 Demo

A demo app showcasing CopilotKit integration with LangGraph and Google's Gemini 3 models. Generate AI-powered scenes by creating characters, backgrounds, and combining them together.

https://github.com/user-attachments/assets/3c60c2b8-5ccd-42f0-817f-0e5e22398a48

What This Demo Shows

• CopilotKit + LangGraph Integration - Connect a Python LangGraph agent to a Next.js frontend
• Shared State Pattern - Bidirectional state sync between frontend and agent
• Human-in-the-Loop (HITL) - Approve/reject AI actions before execution
• Generative UI - Real-time tool execution feedback in chat
• Dynamic API Keys - Pass API keys from frontend to agent at runtime
• Image Generation - Using Gemini 3 and Nano Banana (gemini-2.5-flash-image)

Demo Features

Feature	Description
Character Generation	Create characters with AI-generated images
Background Generation	Generate environments and scenes
Scene Composition	Combine characters and backgrounds
Image Editing	Modify generated images with natural language
HITL Approval	Review and approve image prompts before generation

Quick Start

Prerequisites

• Node.js 18+
• Python 3.10+
• Google AI API Key (get one here)

Setup

# 1. Install dependencies
npm install

# 2. Set up your API key
echo 'GOOGLE_API_KEY=your-key-here' > agent/.env

# 3. Start the app
npm run dev

Open http://localhost:3000 and start creating!

Project Structure

├── src/
│   ├── app/
│   │   ├── page.tsx          # Main UI with CopilotKit integration
│   │   └── api/copilotkit/   # CopilotKit API route
│   ├── components/
│   │   ├── ArtifactPanel.tsx # Display generated artifacts
│   │   ├── ApiKeyInput.tsx   # Dynamic API key management
│   │   └── CustomChatInput.tsx
│   └── lib/
│       └── types.ts          # Shared TypeScript types
├── agent/
│   ├── agent.py              # LangGraph agent with tools
│   ├── server.py             # Custom routes (static files)
│   └── langgraph.json        # LangGraph configuration

Key CopilotKit Patterns

1. Shared State (Frontend ↔ Agent)

// Frontend: src/app/page.tsx
const { state, setState } = useCoAgent<AgentState>({
  name: "sample_agent",
  initialState: { characters: [], backgrounds: [], scenes: [] },
});

// Update state from frontend
setState((prev) => ({ ...prev, apiKey: newKey }));

# Agent: agent/agent.py
class AgentState(MessagesState):
    characters: List[dict] = []
    backgrounds: List[dict] = []
    scenes: List[dict] = []
    apiKey: str = ""

# Read state in agent
api_key = state.get("apiKey", "")

2. Human-in-the-Loop (HITL)

// Frontend: Enable HITL for specific tool
useCopilotAction({
  name: "approve_image_prompt",
  disabled: true, // Agent calls this, not user
  handler: async ({ prompt }) => {
    // Show approval UI, return approved/rejected
  },
});

3. Generative UI

// Show real-time tool progress
useCopilotAction({
  name: "create_character",
  render: ({ status, result }) => (
    <ToolCard status={status} result={result} />
  ),
});

4. LangGraph Agent Tools

# agent/agent.py
@tool
async def create_character(
    name: str,
    description: str,
    prompt: str,
    state: Annotated[dict, InjectedState]  # Access shared state
) -> dict:
    api_key = state.get("apiKey", "")
    image_url = await generate_image(prompt, api_key=api_key)
    return {"name": name, "description": description, "imageUrl": image_url}

Deployment

Deploy the agent to Railway:

cd agent
railway link
railway up
railway variables --set "AGENT_URL=https://your-app.up.railway.app"
railway variables --set "GOOGLE_API_KEY=your-key"

See agent/DEPLOY.md for detailed deployment guide.

Tech Stack

Layer	Technology
Frontend	Next.js 16, React 19, Tailwind CSS 4
AI Integration	CopilotKit 1.10.6
Agent	Python, LangGraph 0.6.6
LLM	Gemini 3 Pro Preview
Image Gen	Nano Banana (gemini-2.5-flash-image)

Learn More

• CopilotKit Docs - Full CopilotKit documentation
• LangGraph + CopilotKit Guide - Integration guide
• Shared State Pattern - State synchronization

License

MIT

Built by Mark Morgan