Create the Lambda Function

Time Estimate: 10 - 15 minutes

In this section you will create the SQS Processor Lambda function.

  1. Make sure you are in N. Virginia region. Look for the Lambda service in the AWS Management console and click on the highlighted result to access the service.

  2. Follow the same steps as previously. Select the python runtime as shown below and configure the execution role to be the IAM role that that was created with CloudFormation when you were setting up the project. Click Create function.

    SQS Lambda
  3. Copy the code below. The code will pick up a message in the Sync SQS queue, call Textract to process a document and then update the DynamoDB Documents and Outputs table as soon as a document has been processed.

import boto3
from decimal import Decimal
import json
import os
from helper import AwsHelper, S3Helper, DynamoDBHelper
from og import OutputGenerator
import datastore

def callTextract(bucketName, objectName, detectText, detectForms, detectTables):
    textract = AwsHelper().getClient('textract')
    if(not detectForms and not detectTables):
        response = textract.detect_document_text(
                'S3Object': {
                    'Bucket': bucketName,
                    'Name': objectName
        features  = []
        response = textract.analyze_document(
                'S3Object': {
                    'Bucket': bucketName,
                    'Name': objectName

    return response

def processImage(documentId, features, bucketName, objectName, outputTableName, documentsTableName):

    detectText = "Text" in features
    detectForms = "Forms" in features
    detectTables = "Tables" in features

    response = callTextract(bucketName, objectName, detectText, detectForms, detectTables)

    dynamodb = AwsHelper().getResource("dynamodb")
    ddb = dynamodb.Table(outputTableName)

    print("Generating output for DocumentId: {}".format(documentId))

    opg = OutputGenerator(documentId, response, bucketName, objectName, detectForms, detectTables, ddb)

    print("DocumentId: {}".format(documentId))

    ds = datastore.DocumentStore(documentsTableName, outputTableName)

# --------------- Main handler ------------------

def processRequest(request):

    output = ""

    print("request: {}".format(request))

    bucketName = request['bucketName']
    objectName = request['objectName']
    features = request['features']
    documentId = request['documentId']
    outputTable = request['outputTable']
    documentsTable = request['documentsTable']
    documentsTable = request["documentsTable"]
    if(documentId and bucketName and objectName and features):
        print("DocumentId: {}, features: {}, Object: {}/{}".format(documentId, features, bucketName, objectName))

        processImage(documentId, features, bucketName, objectName, outputTable, documentsTable)

        output = "Document: {}, features: {}, Object: {}/{} processed.".format(documentId, features, bucketName, objectName)

    return {
        'statusCode': 200,
        'body': output

def lambda_handler(event, context):

    print("event: {}".format(event))
    message = json.loads(event['Records'][0]['body'])
    print("Message: {}".format(message))

    request = {}
    request["documentId"] = message['documentId']
    request["bucketName"] = message['bucketName']
    request["objectName"] = message['objectName']
    request["features"] = message['features']    
    request["outputTable"] = os.environ['OUTPUT_TABLE']
    request["documentsTable"] = os.environ['DOCUMENTS_TABLE']

    return processRequest(request)
#### Important code snippets in the Lambda function

* Below is where we call the textract ***detect_document_text*** API. This API call will return the raw text in an image in a JSON structure.

response = textract.detect_document_text(
                'S3Object': {
                    'Bucket': bucketName,
                    'Name': objectName

* Below if we detect more complicated structures like forms and tables we call the ***analyze_document*** API. This API call will return the raw text and the also the relationships between detected text.

response = textract.analyze_document(
            'S3Object': {
                'Bucket': bucketName,
                'Name': objectName
  1. Paste the code as shown below and click Save.

    SQS Lambda Code
  2. We need to configure some enviroment variables in order for our Lambda fucntion to be able to identify the DynamoDB tables. The enviroment variables we are going to configure are:

    DOCUMENTS_TABLE -> Documents table name
    OUTPUT_TABLE -> Ouputs table name
  3. Configure those as shown below. Then click Save.

    Env Variables
  4. Increase the time out of the Lambda function as shown below. This is to ensure the Lambda has enough time to finish processing before it times out. Then click Save.

    Time out 30 sec