Python: default to service managed threads for AzureAIAgentClient to prevent …

[lavaman131]

Motivation and Context

Short Summary

Fixes the warning when using AI Foundry Agents with AzureAIAgentClient:

agent-framework/python/packages/core/agent_framework/_clients.py:704 - WARNING]
When conversation_id is set, store must be True for service-managed threads. Automatically setting store=True.

Detailed Summary

This pull request introduces a minor update to the python/packages/azure-ai/agent_framework_azure_ai/_chat_client.py file to improve type safety and extend functionality. The most notable change is the addition of a new create_agent method, which delegates agent creation to the superclass, and type checking improvements for better developer experience.

Type checking improvements:

• Added an if TYPE_CHECKING block to import ChatAgent for type hints, improving static analysis and autocompletion without affecting runtime performance.
• Updated import statements to include TYPE_CHECKING from typing, supporting the new type hinting logic.

Functionality extension:

• Added a create_agent method to the chat client, allowing agent creation with optional storage and keyword arguments, and delegating the logic to the superclass.

[Copilot]

Pull Request Overview

This PR fixes a warning message that appears when using AI Foundry Agents with AzureAIAgentClient by defaulting to service-managed threads with store=True. The change ensures proper thread management for conversation-based interactions.

Key Changes:

• Added a create_agent method to AzureAIChatClient that defaults store=True for service-managed threads
• Improved type safety by adding TYPE_CHECKING imports for better static analysis

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/azure-ai/agent_framework_azure_ai
_chat_client.py	421	83	80%	244, 249, 280, 282, 356, 429, 529–530, 532, 535, 538–543, 545, 553, 565–567, 571, 574–575, 583–586, 596, 604–609, 616, 624–625, 633–634, 639–640, 644–651, 656–657, 665, 671, 679–681, 684, 742–744, 761, 773–777, 804–806, 809–810, 825, 830, 872–873, 894, 950, 969, 1035–1036, 1038, 1110
TOTAL	11653	1897	83%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
1371	98 💤	0 ❌	0 🔥	26.385s ⏱️

[eavanvalkenburg]

I don't know if this is the way, for one, this method lacks all metadata compared to the create_agent method on the BaseClient, so it won't give the right IDE info. It also introduces a difference in creating a agent through this method, vs directly, which I don't like. I'm also not convinced what problem this solves.