docs(a2a): Add VNC support for browser-agent with real-time GUI viewing

[edenreich]

Summary

This PR adds VNC server support to the A2A example, enabling real-time viewing of the browser-agent's GUI during automation tasks. This is particularly useful for debugging and observing browser automation workflows.

Changes

• New Dockerfile.vnc: Lightweight Alpine-based container running x11vnc to expose the browser's X11 display
• Updated docker-compose.yaml:
  • Added BROWSER_HEADLESS: false to browser-agent service to enable headed mode
  • Created shared x11-socket volume for X11 display communication
  • Added browser-vnc service exposing port 5900 for VNC connections
• Updated README.md: Added comprehensive documentation section on viewing the browser GUI, including:
  • Requirements (browser-agent v0.4.0+, Xvfb support)
  • Connection instructions for general VNC clients
  • macOS-specific instructions using built-in Screen Sharing

How It Works

1. The browser-agent container runs Xvfb to create a virtual X11 display (:99)
2. The X11 socket is shared via a Docker volume (x11-socket)
3. The browser-vnc container connects to the shared display and exposes it via VNC on port 5900
4. Users can connect using any VNC client (password: password)

Testing

• Verified VNC connection on macOS using built-in Screen Sharing
• Confirmed browser automation is visible in real-time
• Linting passes (markdownlint)

🤖 Generated with Claude Code

Co-Authored-By: Claude noreply@anthropic.com