使用客户端评估 - 由 AWS AppConfig 提供支持

要将 Evidently 客户端评估 Lambda 层添加到您的 Lambda 函数并进行配置

1. 如果尚未创建 Lambda 函数，请创建一个。

2. 将客户端评估层添加到您的函数中。您可以指定它的 ARN，也可以从 AWS 层的列表中进行选择（如果还没有此 ARN）。有关更多信息，请参阅配置函数以使用层和 AWS AppConfig Lambda 扩展程序的可用版本。

3. 创建名为 EvidentlyAppConfigCachingAgentPolicy 且具有以下内容的 IAM policy，并将其附加到此函数的执行角色。有关更多信息，请参阅 Lambda 执行角色。

   {
       "Version": "2012-10-17",
       "Statement": [
           {
               "Sid": "VisualEditor0",
               "Effect": "Allow",
               "Action": [
                   "appconfig:GetLatestConfiguration",
                   "appconfig:StartConfigurationSession",
                   "evidently:PutProjectEvents"
               ],
               "Resource": "*"
           }
       ]
   }

4. 向您的 Lambda 函数添加所需的环境变量 AWS_APPCONFIG_EXTENSION_EVIDENTLY_CONFIGURATIONS。此环境变量用于指定 Evidently 项目与 AWS AppConfig 资源之间的映射。

   如果您将此函数用于一个 Evidently 项目，请将环境变量的值设置为：applications/APP_ID/environments/ENVIRONMENT_ID/configurations/PROJECT_NAME

   如果您将此函数用于多个 Evidently 项目，请使用逗号分隔各个值，如以下示例所示：applications/APP_ID_1/environments/ENVIRONMENT_ID_1/configurations/PROJECT_NAME_1, applications/APP_ID_2/environments/ENVIRONMENT_ID_2/configurations/PROJECT_NAME_2

5. （可选）设置其他环境变量。有关更多信息，请参阅配置 AWS AppConfig Lambda 扩展程序。

6. 在您的应用程序中，通过将 EvaluateFeature 发送到 localhost 进行本地 Evidently 评估。

   Python 示例：

   import boto3
   from botocore.config import Config
           
   def lambda_handler(event, context):
       local_client = boto3.client(
           'evidently',
           endpoint_url="http://localhost:2772",
           config=Config(inject_host_prefix=False)
       )
       response = local_client.evaluate_feature(
           project=event['project'],
           feature=event['feature'],
           entityId=event['entityId']
       )
       print(response)

   Node.js 示例：

   const AWS = require('aws-sdk');
   const evidently = new AWS.Evidently({
       region: "us-west-2",
       endpoint: "http://localhost:2772",
       hostPrefixEnabled: false
   });
   
   exports.handler = async (event) => {
       
       const evaluation = await evidently.evaluateFeature({
           project: 'John_ETCProject_Aug2022',
           feature: 'Feature_IceCreamFlavors',
           entityId: 'John'
       }).promise()
       
       console.log(evaluation)
       const response = {
           statusCode: 200,
           body: evaluation,
       };
       return response;
   };

   Kotlin 示例：

   String localhostEndpoint = "http://localhost:2772/"
   public AmazonCloudWatchEvidentlyClient getEvidentlyLocalClient() {
       return AmazonCloudWatchEvidentlyClientBuilder.standard()
           .withEndpointConfiguration(AwsClientBuilder.EndpointConfiguration(localhostEndpoint, region))
           .withClientConfiguration(ClientConfiguration().withDisableHostPrefixInjection(true))
           .withCredentials(credentialsProvider)
           .build();
   }
   
   AmazonCloudWatchEvidentlyClient evidently = getEvidentlyLocalClient();
   
   // EvaluateFeature via local client.
   EvaluateFeatureRequest evaluateFeatureRequest = new EvaluateFeatureRequest().builder()
      .withProject(${YOUR_PROJECT}) //Required.
      .withFeature(${YOUR_FEATURE}) //Required.
      .withEntityId(${YOUR_ENTITY_ID}) //Required.
      .withEvaluationContext(${YOUR_EVAL_CONTEXT}) //Optional: a JSON object of attributes that you can optionally pass in as part of the evaluation event sent to Evidently.
      .build();
      
   EvaluateFeatureResponse evaluateFeatureResponse = evidently.evaluateFeature(evaluateFeatureRequest);
   
   // PutProjectEvents via local client. 
   PutProjectEventsRequest putProjectEventsRequest = new PutProjectEventsRequest().builder()
       .withData(${YOUR_DATA})
       .withTimeStamp(${YOUR_TIMESTAMP})
       .withType(${YOUR_TYPE})
       .build();
       
   PutProjectEvents putProjectEventsResponse = evidently.putProjectEvents(putProjectEventsRequest);