AI Dashboard Canvas Agent

https://github.com/user-attachments/assets/9201d528-573f-43cc-9d31-571c362318a7

An agent that populates live charts, metrics, and real-time data into a Canvas dashboard instead of just streaming text. Built with CopilotKit, AG-UI, and Google's ADK.

Gen UI concept — agentic canvas. The chat is a thin sidebar; the real surface is a persistent Canvas the agent writes into. Charts, KPIs, and panels are addressable artifacts the agent can place, update, and rearrange across turns — closer to a collaborator using a whiteboard than a chatbot returning replies.

🔧 Quickstart

# from the awesome-llm-apps repo root
cd generative_ui_agents/ai-dashboard-canvas-agent

# install JS deps + agent
pnpm install        # or npm/yarn/bun

# install Python deps separately for the ADK agent
pnpm install:agent

# set your Google API key
cp .env.example .env
# edit .env and set GOOGLE_API_KEY=...

# start UI + agent together
pnpm run dev

📦 Prerequisites

• Node.js 18+
• Python 3.8+
• Google Makersuite API Key → get one here
• Any package manager (pnpm recommended)

💡 Lockfiles (package-lock.json, yarn.lock, etc.) are gitignored — each dev manages their own.

🛠 Available Scripts

• dev → Start UI + agent (default)
• dev:debug → Start with debug logging
• dev:ui → Run just the Next.js app
• dev:agent → Run just the ADK agent
• build / start → Production build + server
• lint → Run ESLint
• install:agent → Install Python deps inside agent/.venv

🎨 Customization

• Main UI → src/app/page.tsx
• Change theme/colors and sidebar appearance
• Add new visualization components
• Extend agent logic in /agent

📚 Docs

• ADK
• CopilotKit
• AG-UI