Add simple workflow sample that mixes agents and executors (#1946)

* Add workflow mix agent executor samples * Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: Evan Mattson <35585003+moonbox3@users.noreply.github.com> Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
2026-06-16 21:04:09 +08:00 · 2025-11-05 14:40:25 -08:00
parent afd8b7ecb4
commit 9b471cc479
2 changed files with 107 additions and 5 deletions
@@ -0,0 +1,103 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import asyncio
+from typing import Never
+
+from agent_framework import (
+    AgentExecutorResponse,
+    Executor,
+    HostedCodeInterpreterTool,
+    WorkflowBuilder,
+    WorkflowContext,
+    handler,
+)
+from agent_framework.azure import AzureAIAgentClient
+from azure.identity.aio import AzureCliCredential
+
+"""
+This sample demonstrates how to create a workflow that combines an AI agent executor
+with a custom executor.
+
+The workflow consists of two stages:
+1. An AI agent with code interpreter capabilities that generates and executes Python code
+2. An evaluator executor that reviews the agent's output and provides a final assessment
+
+Key concepts demonstrated:
+- Creating an AI agent with tool capabilities (HostedCodeInterpreterTool)
+- Building workflows using WorkflowBuilder with an agent and a custom executor
+- Using the @handler decorator in the executor to process AgentExecutorResponse from the agent
+- Connecting workflow executors with edges to create a processing pipeline
+- Yielding final outputs from terminal executors
+- Non-streaming workflow execution and result collection
+
+Prerequisites:
+- Azure AI services configured with required environment variables
+- Azure CLI authentication (run 'az login' before executing)
+- Basic understanding of async Python and workflow concepts
+"""
+
+
+class Evaluator(Executor):
+    """Custom executor that evaluates the output from an AI agent.
+
+    This executor demonstrates how to:
+    - Create a custom workflow executor that processes agent responses
+    - Use the @handler decorator to define the processing logic
+    - Access agent execution details including response text and usage metrics
+    - Yield final results to complete the workflow execution
+
+    The evaluator checks if the agent successfully generated the Fibonacci sequence
+    and provides feedback on correctness along with resource consumption details.
+    """
+
+    @handler
+    async def handle(self, message: AgentExecutorResponse, ctx: WorkflowContext[Never, str]) -> None:
+        """Evaluate the agent's response and complete the workflow with a final assessment.
+
+        This handler:
+        1. Receives the AgentExecutorResponse containing the agent's complete interaction
+        2. Checks if the expected Fibonacci sequence appears in the response text
+        3. Extracts usage details (token consumption, execution time, etc.)
+        4. Yields a final evaluation string to complete the workflow
+
+        Args:
+            message: The response from the Azure AI agent containing text and metadata
+            ctx: Workflow context for yielding the final output string
+        """
+        target_text = "1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89"
+        correctness = target_text in message.agent_run_response.text
+        consumption = message.agent_run_response.usage_details
+        await ctx.yield_output(f"Correctness: {correctness}, Consumption: {consumption}")
+
+
+async def main():
+    async with (
+        AzureCliCredential() as credential,
+        AzureAIAgentClient(async_credential=credential) as chat_client,
+    ):
+        # Create an agent with code interpretation capabilities
+        agent = chat_client.create_agent(
+            name="CodingAgent",
+            instructions=("You are a helpful assistant that can write and execute Python code to solve problems."),
+            tools=HostedCodeInterpreterTool(),
+        )
+
+        # Build a workflow: Agent generates code -> Evaluator assesses results
+        # The agent will be wrapped in a special agent executor which produces AgentExecutorResponse
+        workflow = WorkflowBuilder().set_start_executor(agent).add_edge(agent, Evaluator(id="evaluator")).build()
+
+        # Execute the workflow with a specific coding task
+        results = await workflow.run(
+            "Generate the fibonacci numbers to 100 using python code, show the code and execute it."
+        )
+
+        # Extract and display the final evaluation
+        outputs = results.get_outputs()
+        if isinstance(outputs, list) and len(outputs) == 1:
+            print("Workflow results:", outputs[0])
+        else:
+            raise ValueError("Unexpected workflow outputs:", outputs)
+
+
+if __name__ == "__main__":
+    asyncio.run(main())