PR Checklist

If you have any questions, you can refer to the Contributing Guide

What is the current behavior?

Currently, AI interaction logic is directly embedded in the Copilot.Vue component, making it difficult to reuse and mix multiple responsibilities. All model calls, response handling, state management, and MCP processing are concentrated in the same component, resulting in a bloated component that cannot reuse the same AI interaction functionality in other scenarios.

Issue Number

Example: #123

What is the new behavior?

1. Extracted AI interaction logic: Created an independent AIAgent class that encapsulates all core functionality for interacting with large language models:

   • Message history construction
   • Streaming response handling
   • MCP tool call processing
   • Error handling and interruption control

2. Support for Agent orchestration: Added an AgentOrchestrator class for managing and coordinating multiple specialized AI Agents:

   • Support for registering and managing multiple specialized Agents
   • Provides workflow patterns such as sequential execution and branching execution
   • Implementation of cross-agent data transfer and state control

3. Improved user experience: Fixed AI response status indication issues:

   • Correctly handling loading states
   • Implementing more granular UI state control through callback mechanisms

4. Example implementations: Provided examples of several Agent orchestration patterns to help developers understand and use them

5. Response interruption capability: Added user interface controls to stop AI responses in progress:
   A stop button was added that appears during streaming responses.

Does this PR introduce a breaking change?

• Yes
• No