mirror of
https://github.com/microsoft/agent-framework.git
synced 2026-06-16 21:04:09 +08:00
Jacob Alber
0086d38f58
* refactor: Normalize Run/RunStreaming with AIAgent * refactor: Clarify Session vs. Run -level concepts * Rename RunId to SessionId to better match Run/Session terminology in AIAgent * [BREAKING]: Will break existing checkpointed sessions in CosmosDb due to field rename * refactor: Rename and simplify interface around getting typed data out of ExternalRequest/Response * Also adds hints around using value types in PortableValue * refactor: Rename AddFanInEdge to AddFanInBarrierEdge This will prevent a breaking change later when we introduce a programmable FanIn edge, analogous to the FanOut edge's EdgeSelector. The goal, in the long run is to support a number of different FanIn scenarios, with naive FanIn (no barrier) by default, similar to FanOut. * refactor: AsAgent(this Workflow, ...) => AsAIAgent(...) * misc - part1: SwitchBuilder internal --------- Co-authored-by: Dmytro Struk <13853051+dmytrostruk@users.noreply.github.com>
0086d38f58
ยท
2026-02-20 02:05:18 +00:00
History
Group Chat with Tool Approval Sample
This sample demonstrates how to use GroupChatBuilder with tools that require human approval before execution. A group of specialized agents collaborate on a task, and sensitive tool calls trigger human-in-the-loop approval.
What This Sample Demonstrates
- Using a custom
GroupChatManagerwith agents that have approval-required tools - Handling
FunctionApprovalRequestContentin group chat scenarios - Multi-round group chat with tool approval interruption and resumption
- Integrating tool call approvals with multi-agent workflows where different agents have different levels of tool access
How It Works
- A
GroupChatBuilderworkflow is created with multiple specialized agents - A custom
DeploymentGroupChatManagerdetermines which agent speaks next based on conversation state - Agents collaborate on a software deployment task:
- QA Engineer: Runs automated tests
- DevOps Engineer: Checks staging status, creates rollback plan, and deploys to production
- When the deployment agent tries to deploy to production, it triggers an approval request
- The sample simulates human approval and the workflow completes
Key Components
Approval-Required Tools
The DeployToProduction function is wrapped with ApprovalRequiredAIFunction to require human approval:
new ApprovalRequiredAIFunction(AIFunctionFactory.Create(DeployToProduction))
Custom Group Chat Manager
The DeploymentGroupChatManager implements custom speaker selection logic:
- First iteration: QA Engineer runs tests
- Subsequent iterations: DevOps Engineer handles deployment tasks
Approval Handling
The sample demonstrates continuous event-driven execution with inline approval handling:
- The workflow runs in a single event loop.
- When an approval-required tool is invoked, the loop surfaces an approval request, processes the (simulated) human response, and then continues execution without starting a separate phase.
Prerequisites
- Azure OpenAI or OpenAI configured with the required environment variables
AZURE_OPENAI_ENDPOINTenvironment variable setAZURE_OPENAI_DEPLOYMENT_NAMEenvironment variable (defaults to "gpt-4o-mini")
Running the Sample
dotnet run
Expected Output
The sample will show:
- QA Engineer running tests
- DevOps Engineer checking staging and creating rollback plan
- An approval request for production deployment
- Simulated approval response
- DevOps Engineer completing the deployment
- Workflow completion message
Related Samples
- Agent Function Tools with Approvals - Basic function approval pattern
- Agent Workflow Patterns - Group chat without approvals
- Human-in-the-Loop Basic - Workflow-level human interaction