Setting up a Channel Processor

To start using channel flows, you first create a processor Lambda function to handle preprocessing for your use case. For example, you can update message content or metadata, deny messages and prevent them from being sent, or let the original message through.

Prerequisites

• The Lambda function must be in the same AWS account and the same AWS Regions as the AppInstance.

Granting invocation permissions

You must give the Amazon Chime SDK messaging service permission to invoke your Lambda resource. For more information about permissions, refer to Using resource-based policies for AWS Lambda. For example:

Principal: "messaging.chime.amazonaws.com"

Action: lambda:InvokeFunction

Effect: Allow

AWS:SourceAccount:Your AWSAccountId.

AWS:SourceArn: "arn:aws:chime:region:AWSAccountId: appInstance/"

Note

You can provide a specific app instance ID to invoke your processor, or use a wildcard to allow all Amazon Chime SDK app instances in an account to invoke your processor.

Granting callback permissions

You also need to allow your processor Lambda functions to call the ChannelFlowCallback API. For information about doing that, see AWS Lambda run role in the AWS Lambda developer guide.

You can add an Inline policy to your Lambda function's run role. This example allows the processor to invoke the ChannelFlowCallback API.

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "chime:ChannelFlowCallback"
            ],
            "Resource": [
                "arn:aws:chime:Region:AwsAccountId:appInstance/*"
            ]
        }
    ]
}

Note

Follow the best practices for Lambda functions. For more information, refer to these topics:

• Performance Efficiency Best Practices

• Best practices for working with AWS Lambda

• Configuring reserved concurrency

• Asynchronous invocation

Invoking processor Lambda functions

When a user sends a message, the following input request invokes the processor Lambda function.

{
    "EventType": "string"
    "CallbackId": "string"
    "ChannelMessage": {
        "MessageId": "string",
        "ChannelArn": "string",
        "Content": "string",
        "Metadata": "string",
        "Sender":{
            "Arn": "string", 
            "Name": "string"
        },
        "Persistence": "string",
        "LastEditedTimestamp": "string", 
        "Type": "string",
        "CreatedTimestamp": "string", 
    }
}

EventType

The event being sent to processor. The value is a CHANNEL_MESSAGE_EVENT constant.

CallbackId

The token used while calling the ChannelFlowCallback API from the processor.

ChannelMessage

ChannelArn   The ARN of the channel

Content   Message content to be processed

CreatedTimestamp   The time at which the message was created

LastEditedTimestamp   The time at which a message was edited

MessageId   The message identifier

Metadata   Message metadata to be processed

Persistence   Boolean that controls whether the message is persisted on the back end. Valid Values: PERSISTENT | NON_PERSISTENT

Sender   The message sender. Type: an identity object.

Type   The message type. ChannelFlow only supports the STANDARD message types. Valid Value: STANDARD

The processor function determines the following about each message.

• Whether to update the message content, metadata, or both

• Whether to deny a message

• Whether to leave a message unchanged

When processing finishes, the processor Lambda function sends the result back to the Amazon Chime SDK Messaging service so the message can be sent to all recipients. Message status is marked PENDING until the processor Lambda function sends back the results. The processor Lambda function has 48 hours to send back the results. We do not guarantee message delivery after that, and the ChannelFlowCallback API throws a Forbidden Exception error message. To send back the results, invoke the ChannelFlowCallback API.