Desenvolva e implemente agentes no Vertex AI Agent Engine

Esta página demonstra como criar e implementar um agente no Vertex AI Agent Engine Runtime através das seguintes frameworks de agentes:

Este guia de início rápido explica os seguintes passos:

  • Configure o seu projeto Google Cloud .

  • Instale o SDK Vertex AI para Python e a framework escolhida.

  • Desenvolver um agente de câmbio.

  • Implemente o agente no tempo de execução do Vertex AI Agent Engine.

  • Teste o agente implementado.

Para o guia de início rápido com o Agent Development Kit, consulte o artigo Desenvolva e implemente agentes no Vertex AI Agent Engine com o Agent Development Kit.

Antes de começar

  1. Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.
  2. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

    Roles required to select or create a project

    • Select a project: Selecting a project doesn't require a specific IAM role—you can select any project that you've been granted a role on.
    • Create a project: To create a project, you need the Project Creator role (roles/resourcemanager.projectCreator), which contains the resourcemanager.projects.create permission. Learn how to grant roles.

    Go to project selector

  3. Verify that billing is enabled for your Google Cloud project.

  4. Enable the Vertex AI and Cloud Storage APIs.

    Roles required to enable APIs

    To enable APIs, you need the Service Usage Admin IAM role (roles/serviceusage.serviceUsageAdmin), which contains the serviceusage.services.enable permission. Learn how to grant roles.

    Enable the APIs

  5. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

    Roles required to select or create a project

    • Select a project: Selecting a project doesn't require a specific IAM role—you can select any project that you've been granted a role on.
    • Create a project: To create a project, you need the Project Creator role (roles/resourcemanager.projectCreator), which contains the resourcemanager.projects.create permission. Learn how to grant roles.

    Go to project selector

  6. Verify that billing is enabled for your Google Cloud project.

  7. Enable the Vertex AI and Cloud Storage APIs.

    Roles required to enable APIs

    To enable APIs, you need the Service Usage Admin IAM role (roles/serviceusage.serviceUsageAdmin), which contains the serviceusage.services.enable permission. Learn how to grant roles.

    Enable the APIs

  8. Para receber as autorizações de que precisa para usar o Vertex AI Agent Engine, peça ao seu administrador para lhe conceder as seguintes funções de IAM no seu projeto:

    Para mais informações sobre a atribuição de funções, consulte o artigo Faça a gestão do acesso a projetos, pastas e organizações.

    Também pode conseguir as autorizações necessárias através de funções personalizadas ou outras funções predefinidas.

    Instale e inicialize o SDK Vertex AI para Python

    1. Execute o seguinte comando para instalar o SDK Vertex AI para Python e outros pacotes necessários:

      LangGraph

      pip install --upgrade --quiet google-cloud-aiplatform[agent_engines,langchain]>=1.112

      LangChain

      pip install --upgrade --quiet google-cloud-aiplatform[agent_engines,langchain]>=1.112

      AG2

      pip install --upgrade --quiet google-cloud-aiplatform[agent_engines,ag2]>=1.112

      LlamaIndex

      pip install --upgrade --quiet google-cloud-aiplatform[agent_engines,llama_index]>=1.112
    2. Autentique-se como utilizador

      Colab

      Execute o seguinte código:

      from google.colab import auth
      
      auth.authenticate_user(project_id="PROJECT_ID")
      

      Cloud Shell

      Não é necessária nenhuma ação.

      Shell local

      Execute o seguinte comando:

      gcloud auth application-default login

      Modo expresso

      Se estiver a usar o Vertex AI no modo expresso, não é necessária nenhuma ação.

    3. Execute o seguinte código para importar o Vertex AI Agent Engine e inicializar o SDK:

      1. (Opcional) Antes de testar um agente que desenvolve, tem de importar o Vertex AI Agent Engine e inicializar o SDK da seguinte forma:

        Projeto do Google Cloud

        import vertexai
        
        vertexai.init(
            project="PROJECT_ID",               # Your project ID.
            location="LOCATION",                # Your cloud region.
        )
        

        Onde:

        Modo expresso

        Se estiver a usar o Vertex AI no modo expresso, execute o seguinte código:

        import vertexai
        
        vertexai.init(
            api_key="API_KEY"
        )
        

        onde API_KEY é a chave da API que usa para autenticar o agente.

      2. Antes de implementar um agente, tem de importar o Vertex AI Agent Engine e inicializar o SDK da seguinte forma:

        Projeto do Google Cloud

        import vertexai
        
        client = vertexai.Client(
            project="PROJECT_ID",               # Your project ID.
            location="LOCATION",                # Your cloud region.
        )
        

        Onde:

        Modo expresso

        Se estiver a usar o Vertex AI no modo expresso, execute o seguinte código:

        import vertexai
        
        client = vertexai.Client(
            api_key="API_KEY"
        )
        

        onde API_KEY é a chave da API que usa para autenticar o agente.

    Desenvolva um agente

    1. Desenvolva uma ferramenta de câmbio para o seu agente:

      def get_exchange_rate(
          currency_from: str = "USD",
          currency_to: str = "EUR",
          currency_date: str = "latest",
      ):
          """Retrieves the exchange rate between two currencies on a specified date."""
          import requests
      
          response = requests.get(
              f"https://api.frankfurter.app/{currency_date}",
              params={"from": currency_from, "to": currency_to},
          )
          return response.json()
      
    2. Instancie um agente:

      LangGraph

      from vertexai import agent_engines
      
      agent = agent_engines.LanggraphAgent(
          model="gemini-2.0-flash",
          tools=[get_exchange_rate],
          model_kwargs={
              "temperature": 0.28,
              "max_output_tokens": 1000,
              "top_p": 0.95,
          },
      )
      

      LangChain

      from vertexai import agent_engines
      
      agent = agent_engines.LangchainAgent(
          model="gemini-2.0-flash",
          tools=[get_exchange_rate],
          model_kwargs={
              "temperature": 0.28,
              "max_output_tokens": 1000,
              "top_p": 0.95,
          },
      )
      

      AG2

      from vertexai import agent_engines
      
      agent = agent_engines.AG2Agent(
          model="gemini-2.0-flash",
          runnable_name="Get Exchange Rate Agent",
          tools=[get_exchange_rate],
      )
      

      LlamaIndex

      from vertexai.preview import reasoning_engines
      
      def runnable_with_tools_builder(model, runnable_kwargs=None, **kwargs):
          from llama_index.core.query_pipeline import QueryPipeline
          from llama_index.core.tools import FunctionTool
          from llama_index.core.agent import ReActAgent
      
          llama_index_tools = []
          for tool in runnable_kwargs.get("tools"):
              llama_index_tools.append(FunctionTool.from_defaults(tool))
          agent = ReActAgent.from_tools(llama_index_tools, llm=model, verbose=True)
          return QueryPipeline(modules = {"agent": agent})
      
      agent = reasoning_engines.LlamaIndexQueryPipelineAgent(
          model="gemini-2.0-flash",
          runnable_kwargs={"tools": [get_exchange_rate]},
          runnable_builder=runnable_with_tools_builder,
      )
      
    3. Teste o agente localmente:

      LangGraph

      agent.query(input={"messages": [
          ("user", "What is the exchange rate from US dollars to SEK today?"),
      ]})
      

      LangChain

      agent.query(
          input="What is the exchange rate from US dollars to SEK today?"
      )
      

      AG2

      agent.query(
          input="What is the exchange rate from US dollars to SEK today?"
      )
      

      LlamaIndex

      agent.query(
          input="What is the exchange rate from US dollars to SEK today?"
      )
      

    Implemente um agente

    Implemente o agente criando um recurso reasoningEngine no Vertex AI:

    LangGraph

    remote_agent = client.agent_engines.create(
        agent,
        config={
            "requirements": ["google-cloud-aiplatform[agent_engines,langchain]"],
        },
    )
    

    LangChain

    remote_agent = client.agent_engines.create(
        agent,
        config={
            "requirements": ["google-cloud-aiplatform[agent_engines,langchain]"],
        },
    )
    

    AG2

    remote_agent = client.agent_engines.create(
        agent,
        config={
            "requirements": ["google-cloud-aiplatform[agent_engines,ag2]"],
        },
    )
    

    LlamaIndex

    remote_agent = client.agent_engines.create(
        agent,
        config={
            "requirements": ["google-cloud-aiplatform[agent_engines,llama_index]"],
        },
    )
    

    Use um agente

    Teste o agente implementado enviando uma consulta:

    LangGraph

    remote_agent.query(input={"messages": [
        ("user", "What is the exchange rate from US dollars to SEK today?"),
    ]})
    

    LangChain

    remote_agent.query(
        input="What is the exchange rate from US dollars to SEK today?"
    )
    

    AG2

    remote_agent.query(
        input="What is the exchange rate from US dollars to SEK today?"
    )
    

    LlamaIndex

    remote_agent.query(
        input="What is the exchange rate from US dollars to SEK today?"
    )
    

    Limpar

    Para evitar incorrer em cobranças na sua Google Cloud conta pelos recursos usados nesta página, siga estes passos.

    remote_agent.delete(force=True)
    

    O que se segue?