Sekventiell orkestrering

Definiera dina agenter

Agenter är specialiserade entiteter som bearbetar uppgifter i följd. Här definierar vi tre agenter: en analytiker, en copywriter och en redigerare.

using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Agents;
using Microsoft.SemanticKernel.Agents.Orchestration;
using Microsoft.SemanticKernel.Agents.Orchestration.Sequential;
using Microsoft.SemanticKernel.Agents.Runtime.InProcess;

// Create a kernel with an AI service
Kernel kernel = ...;

ChatCompletionAgent analystAgent = new ChatCompletionAgent {
    Name = "Analyst",
    Instructions = "You are a marketing analyst. Given a product description, identify:\n- Key features\n- Target audience\n- Unique selling points",
    Kernel = kernel,
};

ChatCompletionAgent writerAgent = new ChatCompletionAgent {
    Name = "Copywriter",
    Instructions = "You are a marketing copywriter. Given a block of text describing features, audience, and USPs, compose a compelling marketing copy (like a newsletter section) that highlights these points. Output should be short (around 150 words), output just the copy as a single text block.",
    Kernel = kernel,
};

ChatCompletionAgent editorAgent = new ChatCompletionAgent {
    Name = "Editor",
    Instructions = "You are an editor. Given the draft copy, correct grammar, improve clarity, ensure consistent tone, give format and make it polished. Output the final improved copy as a single text block.",
    Kernel = kernel,
};

Valfritt: Observera agenternas svar

Du kan skapa ett återanrop för att samla in agentsvar när sekvensen fortskrider via egenskapen ResponseCallback .

ChatHistory history = [];

ValueTask responseCallback(ChatMessageContent response)
{
    history.Add(response);
    return ValueTask.CompletedTask;
}

Ställ in sekventiell orkestrering

Skapa ett SequentialOrchestration objekt, skicka in agenterna och det valfria återanropet för svar.

SequentialOrchestration orchestration = new(analystAgent, writerAgent, editorAgent)
{
    ResponseCallback = responseCallback,
};

Starta körningen

En körtidsmiljö krävs för att hantera agenternas exekvering. Här använder vi InProcessRuntime och startar den innan vi anropar orkestreringen.

InProcessRuntime runtime = new InProcessRuntime();
await runtime.StartAsync();

Anropa den orkestreringen

Anropa orkestreringen med din första uppgift (t.ex. en produktbeskrivning). Utdata flödar genom varje agent i följd.

var result = await orchestration.InvokeAsync(
    "An eco-friendly stainless steel water bottle that keeps drinks cold for 24 hours",
    runtime);

Samla in resultat

Vänta tills orkestreringen är klar och hämta det slutgiltiga resultatet.

string output = await result.GetValueAsync(TimeSpan.FromSeconds(20));
Console.WriteLine($"\n# RESULT: {text}");
Console.WriteLine("\n\nORCHESTRATION HISTORY");
foreach (ChatMessageContent message in history)
{
    this.WriteAgentChatMessage(message);
}

Valfritt: Stoppa körningen

När bearbetningen är klar stoppar du körningen för att rensa resurser.

await runtime.RunUntilIdleAsync();

Exempelutdata

# RESULT: Introducing our Eco-Friendly Stainless Steel Water Bottles – the perfect companion for those who care about the planet while staying hydrated! Our bottles ...


ORCHESTRATION HISTORY

# Assistant - Analyst: **Key Features:**
- Made from eco-friendly stainless steel
- Insulation technology that maintains cold temperatures for up to 24 hours
- Reusable and sustainable design
- Various sizes and colors available (assumed based on typical offerings)
- Leak-proof cap
- BPA-free ...

# Assistant - copywriter: Introducing our Eco-Friendly Stainless ...

# Assistant - editor: Introducing our Eco-Friendly Stainless Steel Water Bottles – the perfect companion for those who care about the planet while staying hydrated! Our bottles ...

Definiera dina agenter

Varje agent i sekvensen har ett specifikt ansvar. I det här exemplet har vi:

• ConceptExtractorAgent: Extraherar viktiga funktioner, målgrupp och unika försäljningsställen från en produktbeskrivning.
• WriterAgent: Skriver en marknadsföringskopia baserat på den extraherade informationen.
• FormatProofAgent: Redigerar och polerar utkastkopian för tydlighet och konsekvens.

from semantic_kernel.agents import Agent, ChatCompletionAgent
from semantic_kernel.connectors.ai.open_ai import AzureChatCompletion

def get_agents() -> list[Agent]:
    concept_extractor_agent = ChatCompletionAgent(
        name="ConceptExtractorAgent",
        instructions=(
            "You are a marketing analyst. Given a product description, identify:\n"
            "- Key features\n"
            "- Target audience\n"
            "- Unique selling points\n\n"
        ),
        service=AzureChatCompletion(),
    )
    writer_agent = ChatCompletionAgent(
        name="WriterAgent",
        instructions=(
            "You are a marketing copywriter. Given a block of text describing features, audience, and USPs, "
            "compose a compelling marketing copy (like a newsletter section) that highlights these points. "
            "Output should be short (around 150 words), output just the copy as a single text block."
        ),
        service=AzureChatCompletion(),
    )
    format_proof_agent = ChatCompletionAgent(
        name="FormatProofAgent",
        instructions=(
            "You are an editor. Given the draft copy, correct grammar, improve clarity, ensure consistent tone, "
            "give format and make it polished. Output the final improved copy as a single text block."
        ),
        service=AzureChatCompletion(),
    )
    return [concept_extractor_agent, writer_agent, format_proof_agent]

Valfritt: Observera agenternas svar

Du kan definiera ett återanrop för att observera och skriva ut utdata från varje agent när sekvensen fortskrider.

from semantic_kernel.contents import ChatMessageContent

def agent_response_callback(message: ChatMessageContent) -> None:
    print(f"# {message.name}\n{message.content}")

Ställ in sekventiell orkestrering

SekventielltOrchestration-objekt som skickar in agenterna och det valfria återanropet för svar.

from semantic_kernel.agents import SequentialOrchestration

agents = get_agents()
sequential_orchestration = SequentialOrchestration(
    members=agents,
    agent_response_callback=agent_response_callback,
)

Starta körningen

Starta körmiljön för att hantera agentens körning.

from semantic_kernel.agents.runtime import InProcessRuntime

runtime = InProcessRuntime()
runtime.start()

Anropa den orkestreringen

Anropa orkestreringen med din första uppgift (t.ex. en produktbeskrivning). Utdata flödar genom varje agent i följd.

orchestration_result = await sequential_orchestration.invoke(
    task="An eco-friendly stainless steel water bottle that keeps drinks cold for 24 hours",
    runtime=runtime,
)

Samla in resultat

Vänta tills orkestreringen har slutförts.

value = await orchestration_result.get(timeout=20)
print(f"***** Final Result *****\n{value}")

Valfritt: Stoppa körningen

När bearbetningen är klar stoppar du körningen för att rensa resurser.

await runtime.stop_when_idle()

Exempelutdata

# ConceptExtractorAgent
- Key Features:
- Made of eco-friendly stainless steel
- Keeps drinks cold for 24 hours
...
# WriterAgent
Keep your beverages refreshingly chilled all day long with our eco-friendly stainless steel bottles...
# FormatProofAgent
Keep your beverages refreshingly chilled all day long with our eco-friendly stainless steel bottles...
***** Final Result *****
Keep your beverages refreshingly chilled all day long with our eco-friendly stainless steel bottles...