Large model application management

Step 1: Create a large model application

In the Intelligent Contact Center > Large Model Communication > Communication Intelligence Engine > Large Model Application Management. Click Create Large Model Application. In the dialog box, configure the basic information and click Submit to create the application.

Step 2: Configure inbound calls (optional)

After creating the large model application, you can configure its inbound call settings.

1. On the Large Model Application Management > Inbound Call Configuration to open the configuration page.

2. Choose to use an Alibaba Cloud number or a customer-provided line number. Click Add Number.

   • Alibaba Cloud number: A number you have applied for in Voice Messaging Service.

     Important

     Number types include Inbound, Outbound, and Inbound + Outbound. You must select a number that supports inbound calls, such as an Inbound or Inbound + Outbound number. If you do not have a number that supports inbound calls, go to the Voice Messaging Service console > Real Number Application page to apply for one. For details, see Real Number Application.

     In the Add Alibaba Cloud Number dialog box, select the checkboxes of the numbers you want to add, and then click OK.

   • Customer-provided line number: Use a number from your own line.

     Note

     The customer-provided line feature is currently available by invitation only. Before using it, contact your account manager to add your account to the allowlist. For more details, see Customer-Provided Lines.

     In the Add Customer-provided Line Number dialog box, select a Line Name, enter the inbound numbers in the Line Number input box. You can enter up to 10 numbers. Only digits are supported. Separate multiple numbers with line breaks. Then, click Add.

3. Callback configuration: When a user calls in, you can retrieve opening prompt variables based on the calling and called numbers.

   Turn on the Callback configuration, configure the parameters, and then click Enable.

   1. HTTP request URL: Required. Must start with http or https and be no more than 200 characters long.

   2. Response Variable Configuration: The application automatically reads variables from its text-based opening prompt. You can click Re-identify to fetch the latest variables.

      Note

      This feature can also be used without any variables, primarily to pass information such as the incoming line number.

      In the application creation form, the Opening prompt type supports both Text and Recording options. If you select Text, a tip below the input box indicates that the opening prompt supports variables in the format ${param}, for example: Hello, are you ${param}?. The Prompt input box also supports this variable format. At the bottom of the form, you can click the Select Calling Voice button to configure the voice.

   3. Example: Fill in the fields as shown in the example, click the Test button, and enter the calling number and called number in the dialog box.

   4. Timeout Setting: Required. Configure the timeout period for reading variables. If a timeout occurs, the default variable values are used.

Step 3: Configure transfer to human agent (optional)

After creating the large model application, you can configure settings to transfer calls to a human agent.

1. On the Large Model Application Management > Transfer to Human Agent Configuration to open the configuration dialog box.

2. Enable the Transfer to human agent feature.

Parameters

3. Click Confirm.

Step 4: Configure conversation attributes

On the Large Model Application Management > Conversation Attribute Configuration. In the dialog box that appears, configure the conversation attributes and click OK.

Step 5: Associate a summary agent (optional)

Prerequisites: You have created a Xiaoji Agent. See Xiaoji Agent.

1. Go to the Artificial Intelligence Cloud Call Service console. In the navigation bar, choose LLM Communication > Communication Intelligence Engine > LLM Application Management.

2. Find your application and click Xiaoji Agent Configuration in the Actions column.

3. Set Associate Xiaoji Agent to Yes and select the corresponding Xiaoji Agent.

   Associating a Xiaoji Agent incurs additional fees. See Billing Details for pricing. If you don’t have a suitable Xiaoji Agent, click Create Xiaoji Agent to create one.

Step 6: View large model application details

After configuring the conversation attributes, you can view the application's details. On the Large Model Application Management tab, click Details for the created application to view its details in a dialog box.

The details page consists of two parts: Basic Information (Application name, Application code, Model name, Model code, Qualification, Script, Use case, Scenario, Opening prompt, Prompt, Calling voice, Volume, Speech rate, Base model version, Maximum concurrency, Hotword library) and Conversation Attributes (Silence-triggered model, Silence duration, Hang up on silence configuration, Smart answering detection, Maximum call duration).

Interruption events

During a conversation with the large model application, an interruption event may occur. When an interruption event occurs, the application predicts the point of interruption and passes the content, truncated at the point of interruption, to the large model.

For example, if the agent is saying This is a test opening prompt and you say Who are you? after it has said This is a, an interruption event occurs. The application predicts the interruption point. The content passed to the large model is not:

[
  {
    "role": "assistant",
    "content": "This is a test opening prompt"
  },
  {
    "role": "user",
    "content": "Who are you?"
  }
]

but instead:

[
  {
    "role": "assistant",
    "content": "This is a"
  },
  {
    "role": "user",
    "content": "Who are you?"
  }
]