Google Cloud Functions调用Recommendation AI predict API示例求助

前置配置 #

• 提前为服务账号授予roles/retail.predictCaller权限，这是调用预测接口的最小必要权限
• 云函数部署时绑定已授权的服务账号，无需在代码中硬编码服务账号密钥
• requirements.txt需添加以下依赖：

google-auth==2.23.0
requests==2.31.0

完整可运行代码 #

以下代码自动使用云函数绑定的服务账号完成身份验证，无需手动处理密钥逻辑：

import google.auth
from google.auth.transport.requests import AuthorizedSession
import json

# 全局初始化授权会话，冷启动仅执行一次，提升运行效率
credentials, project_id = google.auth.default(
    scopes=["https://www.googleapis.com/auth/cloud-platform"]
)
authed_session = AuthorizedSession(credentials)

# 需替换为实际项目参数
PROJECT_ID = "*你的GCP项目ID*"
REGION = "*模型部署区域，如global、us-central1*"
PLACEMENT_ID = "*你的预测模型Placement ID*"
ENDPOINT = f"https://retail.googleapis.com/v2/projects/{PROJECT_ID}/locations/{REGION}/placements/{PLACEMENT_ID}:predict"

def predict(request):
    # 若为HTTP触发，可从request中获取调用端传入的用户行为等参数
    request_json = request.get_json(silent=True) or {}
    
    # 构造Predict接口请求体
    predict_request = {
        "userEvent": {
            "eventType": "detail-page-view",
            "visitorId": request_json.get("visitor_id", "*测试用visitor ID*"),
            "userInfo": {
                "userId": request_json.get("user_id", ""),
                "ipAddress": request.get("origin", ""),
                "userAgent": request.headers.get("User-Agent", "")
            },
            "productDetails": request_json.get("product_details", [
                {"product": {"id": "*测试商品ID*"}}
            ])
        },
        "pageSize": 10,
        "filter": "",
        "params": {
            "strictFiltering": True,
            "returnProduct": True
        },
        "labels": {}
    }

    try:
        response = authed_session.post(
            ENDPOINT,
            json=predict_request,
            headers={"Content-Type": "application/json"}
        )
        response.raise_for_status()
        return (json.dumps(response.json()), 200, {"Content-Type": "application/json"})
    except Exception as e:
        return (json.dumps({"error": str(e)}), 500, {"Content-Type": "application/json"})

使用注意事项 #

• 代码中所有带*标记的占位内容，需替换为你自己的项目实际值后再部署
• 若要测试接口，可直接给HTTP触发的云函数发送POST请求，请求体传入自定义的用户行为、商品ID等参数
• 若返回权限错误，先确认云函数绑定的服务账号已授予Retail预测调用权限，且项目已启用Retail API

内容的提问来源于stack exchange，提问作者Rafael Silvestre