This repository contains my personal setup and learnings for the A2UI (Agent-to-User Interface) project. This framework invites Agents (LLMs) to generate declarative UI Schemas that are rendered by native client applications (Angular, Lit, etc.).

I have streamlined the setup to run efficiently on this machine.

• Setup Guide: Step-by-step instructions to get the Angular Client and Python Agents running.
• Learnings: Deep dive into the architecture, protocol, and "magic" behind Generative UI.

Go to samples/client/angular and run:

# Terminal 1: Start the Backend Agent
npm run serve:agent:restaurant

# Terminal 2: Start the Frontend Client
npm start -- restaurant

(See package.json for other agents like contact or rizzcharts)

• Frontend: Angular (v21) Client running on port 4200.
• Backend: Python Agents (using uv) running on ports 10002+.
• AI Model: Configured to use Vertex AI (Gemini 1.5 Pro) via gcloud credentials.
• Configuration: Centralized .env management in the client folder.

• a2a_agents/: Core agent libraries.
• renderers/: The "dumb" UI libraries (Angular, Lit) that interpret the JSON.
• samples/: The actual applications we run.
  • samples/client/angular: The main frontend.
  • samples/agent/adk: The backend agents (restaurant_finder, contact_lookup).
• tools/: Inspector and Editor for debugging.

Maintained by GitHub Copilot for local development.