Virtual Agent action scripts

  • Release version: Washingtondc
  • Updated February 1, 2024
  • 9 minutes to read
    • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of Virtual Agent Action Scripts

    The Virtual Agent action scripts are essential for integrating ServiceNow's AI capabilities with various chat platforms. These scripts facilitate the transformation and handling of user data, enabling seamless communication between the Virtual Agent and chat interfaces. Utilizing Flow Designer, customers can create mandatory action and subflow scripts that manage user interactions effectively.

    Show full answer Show less

    Key Features

    • Provider and Action Scripts: These scripts interact with chat providers like Twilio to manage both inbound and outbound messages. Key scripts include:
      • Provider Attribute: Extracts user data like authentication tokens and message content.
      • Sender Action: Sends responses asynchronously to chat providers.
      • Response Processor: Updates message status based on responses from external providers.
      • Transformers: Both outbound and inbound transformers convert messages between Virtual Agent and chat provider formats.
    • Contextual Actions: These allow for dynamic conversations, such as switching to live agents or ending conversations based on user input.

    Key Outcomes

    By implementing these scripts, ServiceNow customers can:

    • Enhance user engagement through customized conversational experiences.
    • Streamline interactions by managing message flows and responses effectively.
    • Ensure compatibility with various chat providers while maintaining the integrity of user data.
    • Quickly adapt to user requests and contextual commands, improving service efficiency.

    You need to write a set of mandatory Flow Designer action and subflow scripts to pass the user data to VACS, apply transformations and send the response back to your chat interface endpoint.

    Provider and inbound/outbound scripts

    A provider, such as Twilio, provides the chat capability. User action scripts perform provider actions, as well as conversational custom chat integration configuration for inbound and outbound transformer actions for your chatbot or custom conversations. Examples of each script are provided. They include a description of what each script does, as well as the action input and action output. You also see examples of the script JSON.

    You use Flow Designer to build these scripts to transform server and client messages. The provider, sender, and inbound/outbound transformer scripts are required. You can use the additional scripts to add more functionality to your chat.

    Table 1. Scripts
    Script name Purpose Required (Y/N)
    Provider attribute Contains a Provider Auth token, a user identifier, user input, and context variables. Y
    Sender action

    Bundles the request, and sends a response asynchronously via Flow Designer or Integration Hub.

    		 Y
    Response processor Performs platform actions, such as an update message status, that's based on a response received for an outbound message.​ N
    Contextual action Supports contextual actions. N
    Link account action Provides a callback about whether account linking was successful or not​. N
    Outbound transformer Transforms a chat server's supported control to chat interface-specific rendering. Y
    Inbound transformer Optional. You can use the default rich controls if they apply to your chat provider. You do not need to provide a separate inbound transform for each input control. If a default behavior does not work, you can override the default with your own inbound transform for a specific Virtual Agent server control. Y

    Provider attributes action script

    The provider attributes action script extracts data from the incoming message. The script knows about the conversational custom chat integration protocol and outputs the content of the incoming message, the user ID of the sender, and an optional authentication token.
    Note:
    If your provider attributes action script has a contextual action, your script must have both the request_context.contextual_action and request_context.typed_value so that the user input (which is matched to a contextual action) is included in the transcript. If you do not set the request_context.typed_value, that message content does not appear in the transcript.
    Table 2. Provider attributes action script input/output
    Input Output
    • Headers (JSON) - Request headers
    • Payload (JSON) - Request body​
    • Token (String) - Authentication token.
    • provider_user_id (String) - Name of the sender, for example, a user name in a Slack message.
    • request_context (JSON) - Actual content received.
      • typed_value - Typed content such as an answer to a question.
      • attachment_value.url - URL of the attachment.
      • attachment_value.content_type - Content type of the attachment.
      • attachment_value.name - File name of the attachment.
      • contextual action - Action to perform, such as END_CONVERSATION, START_CONVERSATION, AGENT. This script looks at the incoming message. If there is a recognized keyword, then one of the actions can be set, such as outputs.request_context.contextual_action = "END_CONVERSATION";. The framework recognizes this keyword and invokes the contextual action script.
      • context_vars - Contextual variables to pass into a conversation, such as `outputs.request_context.context_vars = {language: "en"};`.
    Figure 1. Example: Flow Designer provider attributes action script input, including headers and payload
    An Action Input in Flow Designer with Headers and Payload set to use JSON.
    Figure 2. Example: Flow Designer provider attributes action script output, including token, request_context, and provider_user_id
    An Action Output in Flow Designer specifies the token, request_context, and provider_user_id values.

    Example: Provider attributes action script.

    (function execute(inputs, outputs) {
    var headers = (inputs.headers);
    var payload = (inputs.payload);

    // check if an authentication token was included. this is optional.
    var smsUtil = new VASMSTwilioUtil();
    outputs.token = JSON.stringify(smsUtil.getToken(headers, payload));

    var data = payload.data;
    var request_context = {}; 
    // inspect the actual message. it could be an MMS (attachment) or raw text
    if (data['MediaUrl0']) {
        var attachment_value = {};
        attachment_value.url = data['MediaUrl0'];
        attachment_value.content_type = data['MediaContentType0'];
        attachment_value.name = smsUtil.getFileName(attachment_value.url , attachment_value.content_type);
        request_context.attachment_value = attachment_value;
    } else {
       request_context.typed_value = data['Body'];
    }

    // set the mandatory outputs
    outputs.request_context = request_context;
    outputs.provider_user_id = data.From;
})(inputs, outputs);

    Sender subflow script

    The sender subflow script knows how to send a message to the provider. The script should asynchronously invoke a Flow Designer or Integration Hub subflow. The Virtual Agent is freed to continue processing messages rather than trying to send external requests. When creating a subflow do not have an action call an action within the subflow.

    }else{
  outputs.sendmessage = true;
  messages = inputs.results.results;
  outputs.from = from;
  outputs.to = to;
  outputs.messages = getMessages(messages);
}

    Contextual action script

    The contextual action script performs special keyword actions such as typing agent while inside a Virtual Agent topic to get immediately handed off to a live agent.

    Table 3. Contextual action script Input
    Input Output
    • conversation_id (String) - The conversation ID.
    • request_context (JSON) - Request passed from the provider attribute script that contains the action to perform, such as (END_CONVERSATION, START_CONVERSATION, AGENT).
    		 None

    Example: Contextual action script.

    (function execute(inputs, outputs) {
  var contextual_action = inputs.request_context.contextual_action;
  if (contextual_action === "END_CONVERSATION") {
    sn_cs.VASystemObject.endConversation(inputs.conversation_id);
  } else if (contextual_action === "AGENT") {
    sn_cs.VASystemObject.switchToLiveAgent(inputs.conversation_id);
  } else if (contextual_action === "START_CONVERSATION") {
    sn_cs.VASystemObject.startConversation(inputs.conversation_id);
  }
})(inputs, outputs);

    Link account action script

    The link account action script contains identifiers for the user who is linking. The script also contains the account that it is linked to.

    Table 4. Link account action script Input
    Input Output
    • provider_user_id (String) - Identifier of the user who is linking, such as a phone number in an SMS case.
    • status (String) - Success/failure.
    • user_id (String) - ServiceNow user_sys_id for the account that was linked to.
    		 None

    Example: Link account action script.

    (function execute(inputs, outputs) {
  
  var richControl = (inputs.rich_control);
  var value = richControl.value;
  outputs.result = richControl.header + ': ' + value.action;
  
})(inputs, outputs);

    Response processor action script

    The response processor action script performs specialty actions that are based on the response to a send message request. The framework already does basic message status tracking.

    Table 5. Response processor account action script input
    Input Output
    • message_id (String) - Identifier of sys_cs_message record that is associated with this response.
    • headers (JSON) - Headers that are received.
    • body (String) - Body that is received.
    • status code (Integer) - The HTTP status code that is received.
    		 None

    Example: Response processor action script.

    (function execute(inputs, outputs) {
    gs.debug("Response from provider: message_id = " + inputs.message_id + ", status_code = " +
      inputs.status_code + ", headers = " + JSON.stringify(inputs.headers) + ", body = " + inputs.body);
})(inputs, outputs);

    Outbound transformer scripts

    The outbound transformer scripts convert outgoing Virtual Agent messages to a protocol that is related to the conversational custom chat integration.

    Input.

    • rich_control (JSON) - Virtual Agent server rich control component that must be transformed.
    • payload (JSON) - Last response that was received from a requester.

    Output.

    result (String) - Provider-appropriate component message to send.

    Table 6. Examples: Outbound transformer scripts
    Outbound transformer name Script example
    Virtual Agent - SMS Twilio Adapter Default Text Outbound Transformer 
    (function execute(inputs, outputs) {
  var rich_control = inputs.rich_control;
  outputs.result = rich_control['label'];
})(inputs, outputs);
    Virtual Agent - SMS Twilio Adapter Default Link Outbound Transformer 
    (function execute(inputs, outputs) {
  var richControl = (inputs.rich_control);
  var value = richControl.value;
  outputs.result = richControl.header + ': ' + value.action;
})(inputs, outputs);
    Virtual Agent - SMS Twilio Adapter Default Picker Outbound Transformer 
    (function execute(inputs, outputs) {
  var rich_control = inputs.rich_control;
  var options = rich_control['options'];
  if (options && options.length > 0) {
      var optionsLength = options.length;
      var picker = rich_control['label'] || "";
      for (var x = 0; x < optionsLength; x++) {
          picker += "\n" + (x+1) +  ": " + options[x].label;
      }
  }
  outputs.text_message = picker;
})(inputs, outputs);
    Virtual Agent - SMS Twilio Adapter Default Multi Link Outbound Transformer 
    (function execute(inputs, outputs) {
  var rich_control = inputs.rich_control;
  var linkResult = rich_control['header'];
  var totalValues = rich_control['values'].length;
  for (var i = 0; i < totalValues; i++) {
    linkResult = linkResult + '\n';
    var linkValue = rich_control['values'][i];
    linkResult = linkResult + '\n' + linkValue['description'] + ': ' + linkValue['action'];
  }
  outputs.result  = linkResult;
})(inputs, outputs);
    Virtual Agent - SMS Twilio Adapter Default Card Outbound Transformer 
    (function execute(inputs, outputs) {
  var richControl = (inputs.rich_control);
  var data = JSON.parse(richControl.data);    
  var tableTitle = data.title;
  var fields = data.fields;
  var subtitle = data.subtitle;
  var url = data.url;    
  var message = subtitle ? getCardLine(tableTitle, subtitle) : tableTitle + "\n";
  for (var i= 0; i<fields.length; i++) {
     message += getCardLine(fields[i].fieldLabel, fields[i].fieldValue);
  }
  message += url;
  outputs.result = message;

  function getCardLine(key, value) {
      if (value == '' || value == null || key == '' || key == null)
          return;
      return key + ": " + value + "\n";
  }
})(inputs, outputs);

    Inbound transformer scripts

    The inbound transformer scripts convert incoming conversational custom chat integration messages to a protocol that the Virtual Agent recognizes. It determines the required value to be injected into the rich control that is presented to the user.

    Input.

    • request_context (JSON) - The request passed from the Provider Attribute script.
    • rich_control (JSON) - The last rich control sent to the requester that contains the original "question".

    Output.

    result (JSON) - The result expected by the Virtual Agent server. If the requester is responding to a topic picker, it is the sys_id of the topic. If the requester is responding to an input text, it is the text response.
    • value - sys_id or text value.
    • search_text - Text that is used if the requester selected something that was unexpected. For example, a user typed something that the NLU could assist with, such as, "None of these options match what I want" in response to a topic picker component.
    Table 7. Examples: Inbound transformer scripts
    Inbound transformer name Script example
    Virtual Agent - SMS Twilio Adapter Default Text Inbound Transformer

    Purpose: Find the text response to the InputText question posed

    		 
    (function execute(inputs, outputs) {
  var requestContext = inputs.request_context;
  var richControl = inputs.rich_control;
  var typedValue = requestContext['typed_value'];
  var attachmentValue = requestContext['attachment_value'];
  var smsUtil = new VASMSTwilioUtil();
  var result = {};
  if(richControl['itemType'] == 'image' || richControl['itemType'] == 'file')	{
    if (attachmentValue && attachmentValue.content_type.includes('image')) {
      result.url = attachmentValue.url;
      result.content_type = attachmentValue.content_type;
      result.name = attachmentValue.name;
      result.headers = attachmentValue.headers;
    } else if (typedValue) {
      result.url = typedValue;
      var contentType = smsUtil.getContentType(typedValue);
      if (typeof contentType === 'undefined')
        gs.error('Please enter a valid image link. Inbound image link: {0}', typedValue);
      else
        result.content_type = contentType;
    } else {
      gs.error('Please upload an image or enter a valid image link.');
    }
  }
  result['value'] = typedValue;
  result['search_text'] = '';
  outputs.result = result;
})(inputs, outputs);
    Virtual Agent - SMS Twilio Adapter Default Picker Inbound Transformer

    Purpose: Find the sys_id of the option that the user selected for this picker question. This script uses the same algorithm of the outbound transformer. If no options are selected, and the user typed something instead, then the "search_text" is set and fed into NLU if it is enabled on this instance.

    		 
    (function execute(inputs, outputs) {
  var request_context = inputs.request_context;
  var rich_control = inputs.rich_control;
  var selectedValue = request_context["typed_value"];
  var result = {};
  result["value"] = "";
  result["search_text"] = "";
  if (rich_control['options']) {
    var options = rich_control['options'];
    var optionsLength = options.length;
    selectedValue === Number(selectedValue)
    if (selectedValue > 0 && selectedValue <= optionsLength) {
        var selectedOption = options[selectedValue - 1];
        result["value"] = selectedOption.value;
    } else {
        result["search_text"] = request_context["typed_value"];
    }
  }
  outputs.result = result;
})(inputs, outputs);