Python: Enhance Copilot Studio Sample with AgentThread, Group Chat Support, and Rich Content

[meghapatilcode]

Motivation and Context

This PR enhances the existing Copilot Studio sample by aligning it with recent Semantic Kernel architecture changes and extending its capabilities for broader usage scenarios. Specifically, it introduces three key improvements:

• Agent Thread Implementation: Adds a CopilotAgentThread implementation to align with the introduction of agent threads in Semantic Kernel v1.26.0.

• Channel Support for Group Chat: To enable Copilot Studio agents to participate in group conversations, this update includes a new CopilotStudioAgentChannel class. This allows channel-based invocation of agents.

• Rich Content Handling: A new CopilotMessageContent class is introduced to handle the variety of content types returned by Copilot Studio agents. This includes not only plain text, but also adaptive cards and suggested actions, providing richer and more interactive responses in conversation flows.

Description

The following files were introduced or updated as part of this change to support Copilot Studio agent integration:

• directline_client.py
  A new utility module that encapsulates all Direct Line API operations. It handles authentication, starts conversations, posts user activities, and retrieves bot responses using watermark-based polling. By extracting this logic from the agent, it provides clear separation between transport-level concerns and agent orchestration.

• copilot_agent.py
  Defines the CopilotAgent, responsible for orchestrating interactions with a Copilot Studio bot via the Direct Line API. It serializes user messages, handles asynchronous polling for responses, and converts bot activities into structured message content.

• copilot_agent_thread.py
  Implements the AgentThread interface for Copilot Studio conversations. It manages Direct Line-specific context such as conversation ID and watermark.

• copilot_agent_channel.py
  Adds a CopilotStudioAgentChannel class, allowing Copilot Studio agents to participate in multi-agent group chats via the new channel-based invocation system.

• copilot_message_content.py
  Introduces CopilotMessageContent, an extension of ChatMessageContent that can represent rich message types from Copilot Studio—including plain text, adaptive cards, and suggested actions.

---

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines
• I didn't break anyone 😄

[moonbox3]

@meghapatilcode, sorry it's taken so long to get back on this: for agent orchestration, we're moving away from the pattern of using agent channels as part of group chat. Our new orchestration patterns leverage an InProcessRuntime, and any agent that inherits from our agent base class, is able to take part in the orchestration. Additionally, as the Copilot SDK/Agents SDK releases further updates related to using Copilot Agents through code via their clients, we're going to move that direction as it simplifies the interaction with the Copilot Studio Agent (there's no need to directly handle the DirectLineClient).

[moonbox3]

@meghapatilcode we appreciate the time you took to work on these updates. Since we have an SK abstraction for the Copilot Studio Agent, we would prefer to put any new dev resources on improving that versus communicating to customers to use this older pattern of interacting with CPS agents. Thank you for your support.