Partager via


Agents déclaratifs

Les agents déclaratifs vous permettent de définir la configuration de l’agent à l’aide de fichiers YAML ou JSON au lieu d’écrire du code programmatique. Cette approche facilite la définition, la modification et le partage des agents entre les équipes.

L’exemple suivant montre comment créer un agent déclaratif à partir d’une configuration YAML :

using Azure.AI.Projects;
using Azure.Identity;
using Microsoft.Agents.AI;

// Create the chat client
IChatClient chatClient = new AIProjectClient(
    new Uri("<your-foundry-project-endpoint>"),
    new DefaultAzureCredential())
        .GetProjectOpenAIClient()
        .GetProjectResponsesClient()
        .AsIChatClient("gpt-4o-mini");

// Define the agent using a YAML definition.
var yamlDefinition =
    """
    kind: Prompt
    name: Assistant
    description: Helpful assistant
    instructions: You are a helpful assistant. You answer questions in the language specified by the user. You return your answers in a JSON format.
    model:
        options:
            temperature: 0.9
            topP: 0.95
    outputSchema:
        properties:
            language:
                type: string
                required: true
                description: The language of the answer.
            answer:
                type: string
                required: true
                description: The answer text.
    """;

// Create the agent from the YAML definition.
var agentFactory = new ChatClientPromptAgentFactory(chatClient);
var agent = await agentFactory.CreateFromYamlAsync(yamlDefinition);

// Invoke the agent and output the text result.
Console.WriteLine(await agent!.RunAsync("Tell me a joke about a pirate in English."));

// Invoke the agent with streaming support.
await foreach (var update in agent!.RunStreamingAsync("Tell me a joke about a pirate in French."))
{
    Console.WriteLine(update);
}

Avertissement

DefaultAzureCredential est pratique pour le développement, mais nécessite une considération minutieuse en production. En production, envisagez d’utiliser des informations d’identification spécifiques (par exemple ManagedIdentityCredential) pour éviter les problèmes de latence, la détection involontaire des informations d’identification et les risques de sécurité potentiels liés aux mécanismes de secours.

Définir un agent inline avec YAML

Vous pouvez définir la spécification YAML complète en tant que chaîne directement dans votre code :

import asyncio

from agent_framework.declarative import AgentFactory
from azure.identity.aio import AzureCliCredential


async def main():
    """Create an agent from an inline YAML definition and run it."""
    yaml_definition = """kind: Prompt
name: DiagnosticAgent
displayName: Diagnostic Assistant
instructions: Specialized diagnostic and issue detection agent for systems with critical error protocol and automatic handoff capabilities
description: An agent that performs diagnostics on systems and can escalate issues when critical errors are detected.

model:
  id: =Env.AZURE_OPENAI_MODEL
  connection:
    kind: remote
    endpoint: =Env.FOUNDRY_PROJECT_ENDPOINT
"""
    async with (
        AzureCliCredential() as credential,
        AgentFactory(client_kwargs={"credential": credential}).create_agent_from_yaml(yaml_definition) as agent,
    ):
        response = await agent.run("What can you do for me?")
        print("Agent response:", response.text)


if __name__ == "__main__":
    asyncio.run(main())

Charger un agent à partir d’un fichier YAML

Vous pouvez également charger la définition YAML à partir d’un fichier :

import asyncio
from pathlib import Path

from agent_framework.declarative import AgentFactory
from azure.identity import AzureCliCredential


async def main():
    """Create an agent from a declarative YAML file and run it."""
    yaml_path = Path(__file__).parent / "agent-config.yaml"

    with yaml_path.open("r") as f:
        yaml_str = f.read()

    agent = AgentFactory(client_kwargs={"credential": AzureCliCredential()}).create_agent_from_yaml(yaml_str)
    response = await agent.run("Why is the sky blue?")
    print("Agent response:", response.text)


if __name__ == "__main__":
    asyncio.run(main())

Prochaines étapes