Repository navigation
Architecture for long running (background) agents #3966
Unanswered
Toni Wenzel (twenzel)
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi there,
I just wondering what is a working architecture for long running agents with Microsoft Agent Framework?
Scenario:
I've a frontend where the user can start a agent task which might take very long.
For this task I create a record in the DB with several information.
Requested functionality:
The agent request/task should be processed without a running client/frontend -> in the background
A frontend should be able to connect to this background job an should retrieve current messages (function calls, messages from LLM etc) to show this in the UI and interact with the agent for any user feedback or approvals
I would love to use AG-UI for the frontend to Agent connection because this seems the way to go.
What are my options?
Start the agent (with streaming) via a background job (job with hangfire or via message broker)?
Persists all messages in the DB with ChatHistoryProvider or custom AgentSessionStore.?
But how can I "connect" to the current execution of the agent? Maybe using a conversationId?
All reactions