Roadmap for Agent Framework as Standalone Workflow Engine vs. Integration with Durable Functions?

[workmaf]

Love the unification of Semantic Kernel + AutoGen—it's a game-changer for AI orchestration. With built-in durability (checkpointing, retries) and workflow graphs now GA, it seems like Agent Framework is evolving toward a full-fledged workflow engine.

Quick clarification on positioning/roadmap:

• For production scenarios needing serverless durability + HTTP APIs (e.g., external triggers/monitoring), should devs use Durable Functions as a wrapper around Agent Framework agents (as in the OpenAI Agent SDK integration examples - https://techcommunity.microsoft.com/blog/appsonazureblog/openai-agent-sdk-integration-with-azure-durable-functions/4453402)?
• Or is the goal for Agent Framework to become a standalone, full workflow engine with native HTTP APIs, advanced fault tolerance (e.g., event sourcing), and deeper Azure Functions parity—reducing the need for wrappers?

[moonbox3]

Adding @alliscode