Main flow

更新时间:
复制 MD 格式

Create a main flow to configure multi-turn conversations between the Intelligent Contact Bot and users.

Create a canvas

  1. Log in to the Intelligent Contact Center console.

  2. In the left-side navigation pane, choose LLM-based Communication > Intelligent Contact Bot > Speech Management.

  3. On the Speech Management page, click a speech or click Configure in its upper-right corner to open the canvas page, where you can configure the main flow.

  4. The main flow page has two main areas: the flow list on the left and the editing area, known as the canvas, on the right.

    The top of the page contains tabs such as Main flow, Knowledge base, Global configuration, Intent library, Input/output parameters, Recognition capability, Speech dubbing, and Basic information. By default, the flow list on the left displays the Opening statement node with an Incomplete status. In the center, you can add three types of nodes: common node, jump node, and logic node. The upper-right corner contains the Flow simulation and Submit for review buttons. The lower-right corner contains the Validate flow and Save draft buttons.

  5. Click the + icon in the flow list to create a new flow.

Add nodes

Common node

A common node contains the bot's speech and the configuration for user reply branches. Drag a common node to the canvas and double-click the node or click the edit icon in its upper-right corner to open its edit page. In the Edit common node dialog box, the Bot speech field supports embedding variables in the format ${variableName}, for example, Hello, are you ${custName}?. The Allow interruption switch controls whether a user can interrupt the bot while it is playing the speech.

Parameter

Description

Bot speech

The content the bot plays. You can add one to four speeches. If the conversation returns to this node multiple times in a single call, the bot plays the speeches sequentially.

User reply branch

  • Supports seven types of user branches: any reply, affirmative, negative, rejection, mute, unrecognized, and custom.

  • You can select a semantic recognition model. The recognition result can then be used as a user branch type. To add a semantic recognition model, go to Speech Management > Configure > Recognition capability and click Add semantic model.

Note

The any reply branch is mutually exclusive with the other six user reply branches. The type of the matched branch is recorded in the detailed conversation history.

Tag tracking

When the conversation reaches this node, the system applies the configured tag to the user. You can view the matched tags in the detailed conversation history for data analysis.

  1. Click a branch to activate it; its color changes from gray to a highlighted state. After activating a branch, you must add an intent that defines its matching condition. You can select a system-preset general intent from the drop-down list. In the branch configuration, the Allow interruption switch controls whether users can interrupt the bot's speech. This is turned off by default.

  2. If no intent meets your requirements, you can click Create Intent. A new intent can match user intent by using keywords or training phrases. Both methods support adding items in batches, with each item separated by a line break. Keywords support the use of regular expressions, and training phrases use natural language understanding technology. In the Edit common node panel, enter the Intent Name (up to 30 characters) and Intent ID. In the Keywords field, enter one or more regular expressions, separated by line breaks. In the Training Phrases field, enter multiple phrases, separated by line breaks. We recommend that each phrase does not exceed 30 characters. When you are finished, click Save and Return.

Jump node

A jump node primarily links main flows or configures conversation-ending actions, such as hanging up or transferring the call. The Bot speech field supports embedding variables in the format ${variableName}. To enable this feature, submit a ticket.

Parameter

Description

Bot speech

The content of the speech the bot plays before a jump. You can add up to four speeches.

Action

The action performed when the conversation flow reaches the jump node, such as hanging up, transferring the call, or jumping to another main flow.

Tag tracking

When the conversation reaches this node, any matched tags are automatically recorded in the call details for personalized data analysis.

Connecting lines

Connecting lines from user reply branches link nodes on the canvas, defining the path of the conversation flow. To delete a connecting line, select it and press the Delete key. In the upper-left corner of the canvas, there are two buttons: + Common node and + Jump node. A flow begins at the start point. For example, the first node might say, "Hello, are you interested in buying a commercial property?" The trigger condition could be set to any reply, which connects to a second node that says, "Thank you for your time," and then performs the hang up action.