I wrote a tool, AWS Lambda Persistence, which is a Python module that enables AWS Lambda functions to maintain state and persist data between invocations by writing to a dict-like variable. It achieves this by using the DynamoDB free tier and so it can be used without adding any costs to your project.
https://github.com/gene1wood/aws-lambda-persistence
The [README gives details on how to use it in AWS Lambda](https://github.com/gene1wood/aws-lambda-persistence?tab=readme-ov-file#how-to-install-it).
Here's an example of what the AWS Lambda code would look like using the tool. This example uses the built-in AWS Lambda cache, and backs it with PersistentMap for when the cache is lost.
```python import datetime from aws_lambda_persistence import PersistentMap def lambda_handler(event, context): global data if 'data' not in globals(): # data isn't present in Lambda cache already data = PersistentMap() if 'expiration_datetime' in data: if data['expiration_datetime'] < datetime.datetime.now(): data.update({ 'special_value': expensive_function(event), 'expiration_datetime': datetime.datetime.now() + datetime.timedelta(hours=4) }) ```
And here's an example with a complete CloudFormation template
```yaml AWSTemplateFormatVersion: '2010-09-09' Resources: ExampleFunction: Type: AWS::Lambda::Function Properties: Code: ZipFile: | import datetime from aws_lambda_persistence import PersistentMap def lambda_handler(event, context): global data if 'data' not in globals(): data = PersistentMap() result = ("The persistent data wasn't in the Lambda cache so " "it was pulled from DynamoDB. ") else: result = "The persistent data was pulled from Lambda cache. " if 'last_run' in data: result += f"This function was last run at {data['last_run']}" else: result += f"This is the first time the function has run." data['last_run'] = datetime.datetime.now() return(result) Handler: index.lambda_handler Layers: - arn:aws:lambda:us-west-2:220481034027:layer:aws-lambda-persistence:2 PackageType: Zip Role: !GetAtt ExampleRole.Arn Runtime: python3.14 # On first run, if the DynamoDB doesn't exist, execution can take 11+ seconds # Clod starts with a pull from DynamoDB can take 4+ seconds Timeout: 30 ExampleRole: Type: AWS::IAM::Role Properties: AssumeRolePolicyDocument: Version: '2012-10-17' Statement: - Effect: Allow Principal: Service: - lambda.amazonaws.com Action: - sts:AssumeRole Policies: - PolicyName: AllowPersistentMap PolicyDocument: Version: '2012-10-17' Statement: - Effect: Allow Action: - dynamodb:CreateTable - dynamodb:TagResource - dynamodb:PutItem - dynamodb:DescribeTable - dynamodb:GetItem Resource: arn:aws:dynamodb:*:*:table/AWSLambdaPersistence ```

I wrote a tool, AWS Lambda Persistence, which is a Python module that enables AWS Lambda functions to maintain state and persist data between invocations by writing to a dict-like variable. It achieves this by using the DynamoDB free tier and so it can be used without adding any costs to your project.
https://github.com/gene1wood/aws-lambda-persistence
The [README gives details on how to use it in AWS Lambda](https://github.com/gene1wood/aws-lambda-persistence?tab=readme-ov-file#how-to-install-it).
Here's an example of what the AWS Lambda code would look like using the tool. This example uses the built-in AWS Lambda cache, and backs it with PersistentMap for when the cache is lost.
```python import datetime from aws_lambda_persistence import PersistentMap def lambda_handler(event, context): global data if 'data' not in globals(): # data isn't present in Lambda cache already data = PersistentMap() if 'expiration_datetime' in data: if data['expiration_datetime'] < datetime.datetime.now(): data.update({ 'special_value': expensive_function(event), 'expiration_datetime': datetime.datetime.now() + datetime.timedelta(hours=4) }) ```
And here's an example with a complete CloudFormation template
```yaml AWSTemplateFormatVersion: '2010-09-09' Resources: ExampleFunction: Type: AWS::Lambda::Function Properties: Code: ZipFile: | import datetime from aws_lambda_persistence import PersistentMap def lambda_handler(event, context): global data if 'data' not in globals(): data = PersistentMap() result = ("The persistent data wasn't in the Lambda cache so " "it was pulled from DynamoDB. ") else: result = "The persistent data was pulled from Lambda cache. " if 'last_run' in data: result += f"This function was last run at {data['last_run']}" else: result += f"This is the first time the function has run." data['last_run'] = datetime.datetime.now() return(result) Handler: index.lambda_handler Layers: - arn:aws:lambda:us-west-2:220481034027:layer:aws-lambda-persistence:2 PackageType: Zip Role: !GetAtt ExampleRole.Arn Runtime: python3.14 # On first run, if the DynamoDB doesn't exist, execution can take 11+ seconds # Clod starts with a pull from DynamoDB can take 4+ seconds Timeout: 30 ExampleRole: Type: AWS::IAM::Role Properties: AssumeRolePolicyDocument: Version: '2012-10-17' Statement: - Effect: Allow Principal: Service: - lambda.amazonaws.com Action: - sts:AssumeRole Policies: - PolicyName: AllowPersistentMap PolicyDocument: Version: '2012-10-17' Statement: - Effect: Allow Action: - dynamodb:CreateTable - dynamodb:TagResource - dynamodb:PutItem - dynamodb:DescribeTable - dynamodb:GetItem Resource: arn:aws:dynamodb:*:*:table/AWSLambdaPersistence ```