如何在Google ADK Agent中调用Vertex AI微调的Gemini模型

解决方案概述 #

Google ADK的Agent类支持传入自定义的genai.Client实例，通过该客户端可连接到Vertex AI上的微调模型端点。需要修改agent.py中的客户端配置和Agent初始化参数，替换原API Key认证方式为Vertex AI身份验证。

修改后的完整agent.py代码 #

import os
from google.adk.agents import Agent
from google.adk.tools import google_search
from google.adk.tools.agent_tool import AgentTool
from google import genai
from google.genai import types
from google.adk.sessions import InMemorySessionService
from google.adk.runners import Runner

# --- 配置Vertex AI客户端 ---
PROJECT_ID = "9456734556734"
LOCATION = "us-central1"

# 初始化Vertex AI客户端（依赖GCP身份验证，无需API Key）
# 本地运行需设置GOOGLE_APPLICATION_CREDENTIALS指向服务账号JSON文件
client = genai.Client(
    vertexai=True,
    project=PROJECT_ID,
    location=LOCATION,
)

# --- 配置生成参数（与微调代码保持一致） ---
generation_config = types.GenerateContentConfig(
    temperature=0.7,
    top_p=0.95,
    max_output_tokens=8192,
    safety_settings=[
        types.SafetySetting(category="HARM_CATEGORY_HATE_SPEECH", threshold="OFF"),
        types.SafetySetting(category="HARM_CATEGORY_DANGEROUS_CONTENT", threshold="OFF"),
        types.SafetySetting(category="HARM_CATEGORY_SEXUALLY_EXPLICIT", threshold="OFF"),
        types.SafetySetting(category="HARM_CATEGORY_HARASSMENT", threshold="OFF"),
    ]
)

# --- Root Agent ---
root_agent = Agent(
    name="RootAgent",
    # 替换为你的微调模型端点路径
    model="projects/9456734556734/locations/us-central1/endpoints/5797762931831894471",
    client=client,
    description="AI Agent",
    instruction="""
    You are a helpful AI assistant.
    """,
    tools=[google_search],
    generation_config=generation_config
)

# --- Root Agent Runner配置 ---
APP_NAME = "snack_creations_app"
USER_ID = "12345"
SESSION_ID = "123344"

session_service = InMemorySessionService()
session = session_service.create_session(
    app_name=APP_NAME, user_id=USER_ID, session_id=SESSION_ID
)
runner = Runner(
    agent=root_agent, app_name=APP_NAME, session_service=session_service
)

# Agent交互逻辑
def call_agent(query):
    content = types.Content(role="user", parts=[types.Part(text=query)])
    events = runner.run(user_id=USER_ID, session_id=SESSION_ID, new_message=content)

    final_response = None
    for event in events:
        if event.is_final_response():
            final_response = event.content.parts[0].text
            print("Agent Response: ", final_response)
            return final_response
    return final_response

# 启动交互式对话
print("Starting the interactive conversation (type 'quit' to exit).")
while True:
    user_query = input("You: ")
    if user_query.lower() == 'quit':
        break
    call_agent(user_query)
    print("\n")

print("Conversation ended.")

关键修改点说明 #

1. 替换认证方式：移除原GOOGLE_API_KEY配置，改用Vertex AI客户端，依赖GCP身份验证（服务账号或环境默认凭据）。
2. 传入自定义客户端：创建genai.Client实例并传入Agent的client参数，让Agent使用Vertex AI的模型端点。
3. 同步生成参数：将微调代码中的GenerateContentConfig设置到Agent的generation_config参数，保持生成规则一致。
4. 指定模型端点：将Agent的model参数替换为你的Vertex AI微调模型端点路径。

注意事项 #

• GCP权限配置：确保运行代码的身份（服务账号或用户）拥有aiplatform.endpoints.predict及Vertex AI相关访问权限。
• 依赖版本：安装最新版依赖避免兼容性问题：

  pip install --upgrade google-adk google-genai
  

• 本地身份验证：本地运行时设置环境变量指向服务账号JSON文件：

  export GOOGLE_APPLICATION_CREDENTIALS="/path/to/your/service-account-key.json"
  

内容的提问来源于stack exchange，提问作者Seth