Serverless translation example

This example demonstrated the capabilities and benefits of managed services on AWS. The solution described below is a very simple example and will lead to the following architecture:

Create S3 Bucket

1. Create an empty bucket with default values

Configure SNS

1. Create a new SNS topic with default settings
2. Create a new subscription and select Email as protocol and enter your E-Mail address as Endpoint
3. Confirm the subscription in the E-Mail you've received
4. From the Topics section in SNS, copy your Topic ARN (should look something like this arn:aws:sns:aws-region:aws-account-id:topicname)

Lambda function

1. Create new Lambda function (Author from scratch)
2. Choose a name, for Runtime select Python 3.12 and create the function
3. Replace existing function code and deploy the Lambda function
4. Review the Python code to understand what's supposed to happen during the invocation

import boto3
import logging
import os

# Clients
s3 = boto3.client('s3')
translate = boto3.client('translate')
sns = boto3.client('sns')

# Configure logger
logger = logging.getLogger()
logger.setLevel(logging.INFO)

def lambda_handler(event, context):
    logger.info('Raw Event: {}'.format(event))
    print(event)
    
    # Extract Bucket- and Object names from Event Information
    bucket_name = event['Records'][0]['s3']['bucket']['name']
    object_name = event['Records'][0]['s3']['object']['key']

    # Get the object from S3
    fileobj = s3.get_object(
        Bucket=bucket_name,
        Key=object_name
    ) 
    
    # Open the file object and read it into the variable filedata. 
    filedata = fileobj['Body'].read()
    
    # file data will be a binary stream.  We have to decode it and store it as string
    content = str(filedata.decode('utf-8'))

    # Send content from the input file to AWS translate
    response_translate = translate.translate_text(
        Text=content,
        SourceLanguageCode='auto',
        TargetLanguageCode='de'
    )
    
    # Send the translated text to the topic recipient
    response_sns = sns.publish(
        TopicArn=os.environ['sns_topic'],
        Message=response_translate['TranslatedText']
    )
    
    return {
        'statusCode': 200,
        'body': response_sns
    }

1. Add a S3 trigger to your Lambda function and select the bucket created previously
2. Add an environment variable key sns_topic and paste your SNS Topic ARN
3. Under Configuration click on the Role name which will open a new windows in IAM

IAM Configuration

The Lambda function comes with a pre-created IAM policy that allows to write log files to CloudWatch Logs. But the function needs also to a/ read from the S3 bucket, b/ detect which language has been selected, c/ translate the payload via Amazon Translate and d/ send the translated text via Amazon SNS. Because of the principle of least priviledge, we need to explicitely allow the Lambda function access to these services and resources. Due to simplicity, we consolidate these actions in one additional IAM policy.

Create IAM policy

• Click on Add permissions and Create inline policy to allow your Lambda function to read from your S3 bucket

S3 Bucket IAM policy

1. Under Service select S3
2. Under Actions select GetObject API call
3. Under Resources select Specific and click on Add ARN
4. Provide your S3 Bucket name, select Any for Object name
5. Click on Add additional permissions

Amazon Comprehend IAM policy

1. Under Service select Comprehend
2. Under Actions select DetectDominantLanguage API call
3. Click on Add additional permissions

Amazon Translate IAM policy

1. Under Service select Translate
2. Under Actions select TranslateText API call
3. Click on Add additional permissions

Amazon SNS IAM policy

1. Under Service select SNS
2. Under Actions select Publish API call
3. Under Resources enter your topic ARN like arn:aws:sns:aws-region:aws-account-id:topicname
4. Click on Review policy
5. Provide a name e.g. "lambda_translate_policy"

You should now have 3 inline IAM policies in the IAM role and if you look at the JSON file, it should look similar to this one (as example)

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "VisualEditor0",
            "Effect": "Allow",
            "Action": [
                "comprehend:DetectDominantLanguage",
                "translate:TranslateText"
            ],
            "Resource": "*"
        },
        {
            "Sid": "VisualEditor1",
            "Effect": "Allow",
            "Action": [
                "s3:GetObject",
                "sns:Publish"
            ],
            "Resource": [
                "arn:aws:sns:<your-region>:<your-account>:<your-topic>",
                "arn:aws:s3:::<your-bucket-name>/*"
            ]
        }
    ]
}

Translate an example text

Create an example text (small sentence) in English in text editor and store it locally. Go back to S3, browse to your bucket and upload or drag/drop your file into your bucket. You should receive a translation of that text shortly after via E-Mail.