Description

Leverage the new Semantic Kernel agent packages to implement a multi-agent chat system in the BookWorm.Chat service. The system should support:

• Multiple AI agents (e.g., book recommendation, Q&A, summarizer) collaborating in chat.
• Agent-to-agent (A2A) communication and orchestration.
• Integration with existing chat streaming, session, and state management infrastructure.

Problem or Need

The project needs to enable collaborative AI agent conversations, using the recently added Semantic Kernel agent packages, to power advanced multi-agent chat and orchestration scenarios.

Proposed Solution

• Design and implement agent registration/discovery so agents can be dynamically composed for a given chat.
• Enable agent-to-agent messaging using Semantic Kernel A2A.
• Allow agents to access chat session context/history and contribute responses.
• Route user prompts and agent responses through the SignalR chat streaming pipeline.
• Expose endpoints to create a multi-agent chat session and interact with it.
• Integrate agent orchestration with current Redis/MongoDB state and cancellation mechanisms.
• Provide basic sample agents (e.g., Book RecommenderAgent, QnAAgent) for demonstration.
• Add tests and documentation for multi-agent orchestration, chat flows, and extensibility.

Alternatives Considered

• Manual agent orchestration without Semantic Kernel (less extensible)
• Using a single agent per chat session

Documentation

• Semantic Kernel