Agent Orchestration in web application with multiple user

[ultrasimulation]

Hello, I need to develop an agentic web application that will be used by multiple users. For the backend I am thinking to build an ASP.NET Core WebAPI with a "/message" endpoint.
-What are the best practice in terms of Dependency Injection for the orchestration?
-How can I reconstruct the state of the orchestration while interacting with a specific user (e.g. the chat history)?
-Thinking to orchestrations like the handoff that could have an human in the loop. Does it means that i need to keep the HandoffOrchestration object "active" on the server (e.g. in a static memory structure) and the InProcessRuntime running in a Task while interacting with the user via HTTP?

Documentation that provides guidance on this, sample and best practices would be very useful.

[TaoChenOSU]

Hi @ultrasimulation,

Thank you for your question!

We are actively working on the context management story for the multi-agent orchestrations. For example, you can pause it and resume it.

For now, you'd have to keep the orchestration instance active on the server while waiting for user input.

[pangjianxin]

any news here?