Amazon Connect Integration

​

Overview

• Chat: Cekura starts a chat through one of your communications widgets, the same way a visitor on your website does, and plays the scenario as the customer. The chat reaches whatever your flow routes it to, such as an Amazon Lex bot or an AI agent.
• Voice: Cekura calls the phone number attached to your flow and plays the scenario as the caller.

​

Prerequisites

• An Amazon Connect instance with a published flow that handles chats.
• Access to the Amazon Connect admin website, to create a communications widget.
• For voice testing: a phone number claimed in Amazon Connect and attached to the flow.

​

Set up a communications widget in Amazon Connect

<script type="text/javascript">
  (function(w, d, x, id){
    ...
  })(window, document, 'amazon_connect', '<widget-id>');
  amazon_connect('snippetId', '<snippet-id>');
  ...
</script>

​

Connect Amazon Connect in Cekura

1

Select Amazon

In Cekura, go to Agents, create an agent, and select Amazon as the provider. If you don’t see it, click More options to expand the full provider list.

2

Enter the widget details

Under Connections → Chat, select Amazon Connect, then fill in Amazon Connect Settings:

Field	Value
Widget ID	The widget ID from the code snippet.
Snippet ID	The snippet ID from the code snippet.
Connect Host	Your instance’s domain, for example bright-dental.my.connect.aws.
AWS Region	Optional. Your instance’s AWS Region. Defaults to us-east-1.

3

Add the phone number (optional)

To also test over the phone, select Telephony under Connections → Voice and enter the number attached to your flow under Telephony Settings. Leave Inbound on so Cekura calls your number.

4

Create the agent and test the connection

Click Next, enter the agent’s name, language and description, and click Create. Cekura checks the configuration, saves the agent, and offers a test. Select Chat and start the test: Cekura starts a chat through your widget, sends a greeting, and waits up to 30 seconds for your flow to reply.

​

Chat testing

• Your flow’s messages are labeled Main Agent in the transcript, and Cekura’s are labeled Testing Agent.
• Interactive messages, such as list pickers and quick replies, appear as text: the title, then the choices after Options:. Cekura answers with a typed message, so your flow must accept a typed answer to those prompts.
• The chat ends when your flow ends it, when the scenario is complete, or after 2 minutes with no message from your flow.

​

Voice testing

​

Troubleshooting

• Amazon Connect refused to start a chat through the communications widget: check that the Widget ID and Snippet ID come from the same widget, that https://api.cekura.ai is in the widget’s allowed domains, and that the widget’s security is set to No. Also confirm that Connect Host and AWS Region match your instance, and that the widget’s flow is published.
• The chat started but its message stream could not be opened: run the test again, and confirm that AWS Region matches your instance.
• The chat started but your flow sent no message: make sure the flow replies to the customer’s first message and routes the chat to your bot or AI agent.
• Changes don’t take effect: Amazon Connect can take a few minutes to apply widget changes. Wait, then run the test again.

​

Configure through the API

{
  "name": "Support assistant",
  "description": "Answers customer chats and calls for appointments and billing questions.",
  "project": 123,
  "telephony": {
    "phone_number": "+15551234567",
    "inbound": true
  },
  "provider": {
    "type": "amazon_connect",
    "chat_agent_details": {
      "type": "amazon_connect",
      "config": {
        "widget_id": "<widget-id>",
        "snippet_id": "<snippet-id>",
        "connect_host": "your-instance.my.connect.aws",
        "region": "us-east-1"
      }
    }
  }
}