Before you Begin

This 45-minute tutorial shows you how to use the Oracle Digital Assistant Native Client SDK for Web (also known as the ODA Web SDK) to add a chat window that communicates with an Oracle Digital Assistant skill to your website. You'll start with a basic web interface and then learn how to use the SDK to customize and enhance the look and feel.

This is the first tutorial in a series:

Background

While your customers can access your skills through many platforms, such as Android apps, Twilio SMS, Facebook Messenger, WeChat, and Slack, they also can access skills directly from your website through the ODA Web SDK. This feature gives your customers the opportunity to ask questions, complete transactions, resolve problems, and so on, as they browse your web pages.

The SDK's features include:

  • Configurable components, such as
    • Timestamp display
    • Chat bubble size and padding
    • Font color and size
    • Custom buttons and icons
    • Chat widget size
  • Multi-lingual chat
  • Draggable launch button
  • TTS support/Message narration
  • Autocompletion of text
  • JWT client authentication
  • Voice recognition/Voice messages

What Do You Need?

  • Oracle Digital Assistant Version 21.10 or higher
  • The pizza skill (included with this tutorial)
  • The chat-page ZIP file (included with this tutorial), which contains the starter files.
  • ODA Web SDK Version 21.10 or higher. You can access the client SDKs from your Digital Assistant instance by opening the side menu and then by clicking Downloads.
  • One of the following supported browsers:
    • Firefox
    • Chrome
    • Safari
    • Edge

    • Set Up Folders

    Let's start by creating a folder and adding the necessary artifacts.

    1. The chat-page.zip contains the starter files for the web page and chat widget. Download this file and then extract it to your local system. The downloaded chat-page file contains an index.html file and the css, images, and scripts subfolders.

      In this tutorial, your work will focus on the index.html file and the scripts subfolders.
    2. Download and extract the latest ODA Web SDK to your system. Copy the web-sdk.js file (located in native-client-sdk-js<version number>/native-client-sdk-js) to the scripts subfolder of chat-page (chat-page/scripts).
    3. In the top folder (chat-page), open index.html (chat-page/index.html) in a browser to take a look at the initial web page.

    Set Up Your Skill

    We've provided a pizza ordering skill. Follow these steps to import it into your instance and then train it.

    1. In your Oracle Digital Assistant instance, click the hamburger icon hamburger icon to open the side menu, and then click Development > Skills.
    2. Search for Web SDK Order Pizza.
    3. If your instance doesn't have this skill, then do these steps:
      • Download Web_Tutorial_1.zip to your system.
      • On the Skills page in your Digital Assistant instance, click Import Skill and import the downloaded ZIP file.

        Tip:

        If you don't see the imported skill, try reloading the page.
    4. On the Web SDK Order Pizza card, click Options Options icon and then click Clone.
    5. Give the cloned skill a unique name. For example, prepend WebSDKOrderPizza with your initials, such as ABWebSDKOrderPizza.

      Note:

      The name must begin with a letter and can contain only letters, numbers, periods, and underscores. It can't exceed 100 characters.
    6. Provide a display name, such as AB Web SDK Order Pizza, and leave all other values as they are.
    7. Select Open cloned skill afterwards.
    8. Click Clone.

      Image of Create Clone of skill with fields filled as specified in the steps.

    9. Click Train (located at the upper right). Ensure that Trainer Ht is selected, and then click Submit to train the skill.

      The Train icon

      Note:

      For this tutorial, you're going to use Trainer HT because of its fast training. For production work, use the deep learning model, Trainer Tm.

    10. Optionally, click Preview to open the Skill Tester.

      Preview icon

      Select the Oracle Web channel, and then enter hi to try out the skill.

      When you are done, close the tester.

      The Skill tester

    Create an Oracle Web Channel

    You use an Oracle Web channel to enable an Oracle Web chat widget to connect to your skill.

    For the purposes of this tutorial, you won't restrict the allowed domains, nor will you enable client authentication. For production work, however, you should do both, as described in Secure Your Web App Chat with Client Authentication.

    1. In the left Navigation menu, click Channels. Ensure that the Users page is open.
    2. Click + Add Channel.
    3. Set these values:
      • Name: A unique name for the channel, such as ABOracleWebChannel
      • Description: Channel for the Web SDK tutorial.
      • Channel Type: Oracle Web
      • Allowed Domains: * (asterisk)
      • Client Authentication Enabled: Switch to off.
      • Session Expiration (minutes): Use the default value.
    4. Click Create.
    5. Select the channel that you just created. Then, from the Route To drop-down list, choose the skill that you just created.

    6. Switch Channel Enabled to on.

      Image of Oracle Web channel with fields filled as specified in the steps.
    7. Make a note of the Channel ID because you will need it when you configure your chat widget.
    8. The chat widget configuration also needs the Oracle Digital Assistant host and domain name from the instance's URL, which appears in the browser's address field, so make a note of that too.

    Your instance is now set up for you to connect the pizza ordering skill to a web page's chat bot.

    Add a Chat Widget to the Web Page

    To get started, you'll create settings.js, add add a chat widget to the file, and set some essential configurations.

    1. In the scripts folder (chat-page/scripts), create a file named settings.js. Then open it in an editor.

    2. Copy and paste the following code into the settings.js file. 
      'use strict';


/**
 * Initializes the SDK and sets a global field with passed name for it the can
 * be referred later
 *
 * @param {string} name Name by which the chat widget should be referred
 */
function initSdk(name) {
    // Retry initialization later if WebSDK is not available yet
    if (!document || !WebSDK) {
        setTimeout(function () {
            initSdk(name);
        }, 2000);
        return;
    }

    if (!name) {
        name = 'Bots';          // Set default reference name to 'Bots'
    }
    var Bots;



    setTimeout(function () {

        var chatWidgetSettings = {
            URI: 'XXXXXXXXXXXXXXXXXXXXXXXXX', // ODA URI, pass the hostname. Do not include the protocol (https://).

            channelId: 'XXXXXXXXXXXXXXXXXXXXXXXXX', // Channel ID, available in channel settings in ODA UI

//Add settings here


        };


        Bots = new WebSDK(chatWidgetSettings);



        Bots.connect();

//Add Bots.setUserInputMessage('Order pizza'); here. Comment out for voice recognition.


        // Create global object to refer Bots
        window[name] = Bots;
    }, 0);
}

    3. Set these chatWidgetSettings properties:

      • URI: This is the host and domain name that you noted when you created the Oracle Web channel. For example: xxxx.xxx.digitalassistant.xxx.xxx.com.

        Note:

        Do not include the protocol (https://). Add the path from the host name to the .com extension.
        For example: xxxx.xxx.digitalassistant.xxx.xxx.com

      • channelId: This is the channel ID that you noted when you created the Oracle Web channel. For example: 12a45b92-2c85-88aa-810d-1dc0d1cfe472.

      For example:

              let chatWidgetSettings = {
            URI: 'xxxx.xxx.digitalassistant.xxx.xxx.com',
           
            channelId: '12a45b92-2c85-88aa-810d-1dc0d1cfe472'
        };

    4. Save your changes.

    5. In an editor, open the index.html file (chat-page/index.html).

    6. Add the references to the settings.js and web-sdk.js files by copying and pasting this code just before the </head> tag. 

          <script src="scripts/settings.js"></script>
    <script src="scripts/web-sdk.js" onload="initSdk('Bots')"></script>

    7. Save your changes.

    8. Let's try out the chat widget. Open your index.html file in a browser.

    9. Click the chat-widget icon to open the chat window.

      chat widget icon

    10. Type hi to start a conversation with the Web SDK Order Pizza skill.

      Chat widget

    11. Type order pizza to verify that you get a response.

    12. (Optional) Continue the conversation until you complete the order.

    Modify the Widget's Behavior

    Now that you've seen the widget's default behavior, let's modify settings.js to make some minor customizations, such as:

    • Automatically opening the chat window
    • Repositioning the window

    You'll also modify the behavior so that the user doesn't have to type anything to initiate the conversation.

    Note:

    Throughout this tutorial, you'll make changes by adding properties to the chatWidgetSettings object in the settings.js file.
    1. To show the connection status in the chat window, add this property to the chatWidgetSettings object: 
                  showConnectionStatus: true,
    2. To automatically open the chat window and to position the bottom right corner of the chat window 2 pixels from the bottom and 2 pixels from the right side of the browser window, add the following just below showConnectionStatus: true: 
                  openChatOnLoad: true,
            position: {bottom: '2px', right: '2px'},
    3. The user has to type something to initiate the conversation with the skill. You can avoid that by passing a hidden message to the skill. Add this property: 
                  initUserHiddenMessage: 'Hello',

      After this change, your chatWidgetSettings object should look like this (with your own URL and channel ID):

       var chatWidgetSettings = {

            URI: 'XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX',                                
            channelId: 'XXXXXXXXXXXXXXXXXXXXXXXXXXXXX',  
            showConnectionStatus: true,
            openChatOnLoad: true,
            position: { bottom: '2px', right: '2px' },
            initUserHiddenMessage: 'Hello',
    }

    4. Save your changes.
    5. To see your changes, Shift+Reload index.html in your browser.

      Notice that the chat window opens and displays the skill's first response automatically. Also notice that it shows the connected status.

      Chat window with connected status and welcome message.

    Embed the Widget in an HTML Container

    By default, the chat window appears as a fixed position element on the bottom-right corner of the page. To have more control over how the chat window appears, you'll activate the embedded mode and place the widget in a DOM element that serves as the container.

    The embedded widget occupies the full width and height of the container be default. To accommodate the widget, you'll need to specify the container's height and width.

    You'll also learn how to display the contents of a container at the top of the chat window.

    1. To add containers for your custom header and the text that you want to display at the top of the chat, insert this HTML just below the <body> tag: 
          <div id="custom-header" style="padding: 0; text-align: initial">
        <h1>Pizzabot</h1>
    </div>
    <div id="top-text" style="padding: 0; text-align: initial">
        <p>Talk to Pizzabot to order your pizza.</p>
    </div>
    2. Now let's add the container for your widget. In your index.html file, insert this HTML just before <footer class="footer">: 
              <div id="chat-container" class="chatbox" 
            style="height: 600px; width: 400px; padding: 0;
            text-align: initial">
        </div>

    3. Save your changes.

    4. Open scripts/settings.js if it isn't already open.

    5. Directly below initUserHiddenMessage: 'Hello', insert this code to activate the embedded mode and to tell the widget into which containers to put the embedded widget, the custom header, and the top text. 
                  embedded: true,
            targetElement: 'chat-container',    
            embedTopStickyId: 'top-text',
            customHeaderElementId: 'custom-header',
    6. Save your file, and then Shift+Reload index.html in your browser.

      As shown here, the chat window now appears within the beige box. The window has a title and has text at the top of the chat.

      Widget with Pittabot title and Talk to Pizzabot to order your pizza text at the top of the chat.
    7. Complete a pizza order. Notice how Talk to Pizzabot to order your pizza doesn't scroll up. You'll change this in the next step.
    8. Your customers don't need to keep seeing the Talk to Pizzabot... message, so change the embedTopStickyId property to embedTopScrollId as follows. 
          embedTopScrollId: 'top-text',
    9. Save index.html, and then refresh your browser (Shift+Reload).

      Now, when you converse with the skill, the text scrolls up.

    See embedded-widget.txt for the completed HTML for this section.

    Customize the Look and Feel of the Chat Window

    Typically, you'll want to add your own branding to the chat window, and change the fonts and colors to match your web site. We'll show you some of the ways that you can do that.

    1. We'll start by changing the icons by inserting the following definition for the icons property. Insert the following icons object just below customHeaderElementId: 'custom-header',:
      2.          icons: {
            avatarBot: 'images/bot-green.png',
            avatarUser: 'images/user-icon.png',
            logo: 'images/bot-white.png',
            launch: 'images/bot-button.png'
          },
    2. To change the text for the widget's chat title, connected text, input placeholder, and send icon, insert the il18n object below the icons object: 
                  i18n: {
                en: {
                    chatTitle: 'Order from',       // Replaces Chat
                    connected: 'Ready',            // Replaces Connected
                    inputPlaceholder: 'Type here', // Replaces Type a message
                    send: 'Send (Enter)'           // Replaces Send tool tip
                }
            },

      The text properties must be wrapped in sub-objects within an i18n object, and there must be a sub-object for the locale that's specified by the locale property, which is en by default.

      The chat widget first looks for text for the locale that's specified by the locale property (the default is en). If there isn't any, it then looks for text for the browser's locale. If that text doesn't exist, it then uses the en text. You can add sub-objects for additional languages, such as es, fr, and zh-cn in the sub-object for the browser's default locale.

    3. To change the font size and type for the messages, insert this property after the il8n object. 
                  fontFamily: 'Oracle Sans, Helvetica Neue, Helvetica, Arial, sans-serif',
    4. Save your changes.
    5. To see your changes, Shift+Reload index.html in your browser.

      Your chat window should look like this.

      Chat window with top bar now saying Order from Pizzabot and READY. There's a new icon for the bot response.
    6. (Optional) By default, the chat widget users a dark blue theme. To change it to a light blue theme, insert the following property below the fontFamily property: 
                  theme: 'classic',

      Save your changes and reload index.html to see the new theme. Then try the redwood-dark theme (theme: 'redwood-dark').

      Remove, or comment out (//theme: 'classic',), the theme property when done.

    Add Autocompletion, Profile Values, and Audio Response

    In this section, you'll learn about these advanced features:

    • Autocompletion: In your skill, you can enter some suggested utterances for each intent. Then, you can configure the widget to display autocompletion suggestions when the user starts typing one of those utterances.
    • Profile information settings: You can configure the widget to pass profile values to the skill.
    • Audio response: You can configure the widget to read the skill's responses aloud.

    Note that for audio response, you'll need to use a Chrome, Edge, Firefox, or Safari browser.

    1. To get started, you'll need to make some changes to your skill to test the passing of profile values. In your Digital Assistant instance, open your clone of the Web SDK Order Pizza skill.
    2. Click Flows Flows icon and scroll down to the welcome state (the state appears after # TUTORIAL START).
    3. Replace the text property with the following: 
            text: "Hello ${profile.firstName}. You last ordered ${profile.properties.value.lastOrderedItems}. Say 'order pizza' to start an order."

      These profile values will be set by the initUserProfile property that you'll soon add to the index.html file.

    4. Click Intents Intents icon, select the Order Pizza intent, and then scroll down to the Auto-Complete Suggestions section.
    5. Click Advanced input mode that's next to the Suggestions to Add field.

      A multi-line text box appears.

    6. Copy these suggestions, paste them into the text box, and then click + Create: 
      can I get a pizza
I want pizza
order pizza
what kinds of pizza do you have

      The Auto-Complete Suggestions section should look like this:

      Section with the pasted suggestions

    7. Click Train to add the autocompletion suggestions to the natural language parser.

      If you forget to retrain the skill, you won't see the suggestions when you next test the chat widget.

    8. In your scripts/settings.js file, just after the fontFamily property, insert the following properties to set the profile values and to turn on autocompletion and audio response: 
                  initUserProfile: {
                profile: {
                    firstName: 'First',
                    lastName: 'Last',
                    email: 'first.last@example.com',
                    properties: {
                        lastOrderedItems: '1 medium pepperoni'
                    }
                }
            },
            enableAutocomplete: true,
            enableAutocompleteClientCache: true, // Minimizes server calls when user uses autocomplete feature
            enableBotAudioResponse: true,

      You can substitute any name for 'First', 'Last' and 'first.last@example.com'.

      Note that while you can use either the initUserProfile property or the Bots.updateUser function to set profile values, you must use the initUserProfile property whenever you use initUserHiddenMessage. Otherwise, the profile values aren't set until after the welcome message displays.

    9. In the en sub-object in the i18n property, insert these properties just above the chatTitle property: 
      
                    audioResponseOff: 'Click to turn audio response on', // Tool tip for the speaker off button
                    audioResponseOn: 'Click to turn audio response off',  // Tool tip for the speaker on button
    10. To populate the message box with the most typically used utterance, add the following below Bots.connect(); 
              Bots.setUserInputMessage('Order pizza');
      When you're done, Your settings.js file should look like this: 
      'use strict';


/**
 * Initializes the SDK and sets a global field with passed name for it the can
 * be referred later
 *
 * @param {string} name Name by which the chat widget should be referred
 */
function initSdk(name) {
    // Retry initialization later if WebSDK is not available yet
    if (!document || !WebSDK) {
        setTimeout(function () {
            initSdk(name);
        }, 2000);
        return;
    }

    if (!name) {
        name = 'Bots';          // Set default reference name to 'Bots'
    }
    var Bots;



    setTimeout(function () {

        var chatWidgetSettings = {
            URI: 'XXXXXXXXXX', // ODA URI, only the hostname part should be passed, without the https://

            channelId: 'XXXXXXXX', // Channel ID, available in channel settings in ODA UI

            //Add settings here

            showConnectionStatus: true,
            openChatOnLoad: true,
            position: { bottom: '2px', right: '2px' },
            initUserHiddenMessage: 'Hello',
            embedded: true,
            targetElement: 'chat-container',
            embedTopScrollId: 'top-text',
            customHeaderElementId: 'custom-header',
            icons: {
                avatarBot: 'images/bot-green.png',
                avatarUser: 'images/user-icon.png',
                logo: 'images/bot-white.png',
                launch: 'images/bot-button.png'
            },
            i18n: {
                en: {

                    audioResponseOff: 'Click to turn audio response on', // Tool tip for the speaker off button
                    audioResponseOn: 'Click to turn audio response off',  // Tool tip for the speaker on button   
                    chatTitle: 'Order from',       // Replaces Chat
                    connected: 'Ready',            // Replaces Connected
                    inputPlaceholder: 'Type here', // Replaces Type a message
                    send: 'Send (Enter)'           // Replaces Send tool tip
                }
            },
            fontFamily: 'Oracle Sans, Helvetica Neue, Helvetica, Arial, sans-serif',
            //theme: 'redwood-dark',
            initUserProfile: {
                profile: {
                    firstName: 'Chris',
                    lastName: 'Smith',
                    email: 'chris.smith@example.com',
                    properties: {
                        lastOrderedItems: '1 medium pepperoni'
                    }
                }
            },
            enableAutocomplete: true,
            enableAutocompleteClientCache: true, // Minimizes server calls when user uses autocomplete feature
            enableBotAudioResponse: true,
        };


        Bots = new WebSDK(chatWidgetSettings);



        Bots.connect();

        //Add Bots.setUserInputMessage('Order pizza'); here 
     
        Bots.setUserInputMessage('Order pizza');



        // Create global object to refer Bots
        window[name] = Bots;
    }, 0);
}
    11. Save your changes.
    12. To see your changes, Shift+Reload index.html in your browser.

      Notice that the welcome message now displays some of the values that you set in the initUserProfile property. Also notice that it seeds the message box with Order pizza.

    13. Click the speaker icon Speaker icon to turn audio response on.
    14. Clear the message box, and then type I want and see the autocomplete suggestions that pop up.

      Autocompletion suggestions for I want

    15. Complete an order. The chat bot speaks its responses.

      Note: When the widget is expanded by default and initUserHiddenMessage is set, some browsers don't utter the first response.

    Add Voice Recognition

    Your skill can speak to you, but you can't speak to it...yet. In this section, you're going to enable voice recognition. But before you do this:

    1. Comment out Bots.setUserInputMessage('Order pizza'); to remove the seeded message to prevent it from interfering with your voice commands. 
      //Bots.setUserInputMessage('Order pizza');
    2. To quickly reset the conversation, add the following property below enableBotAudioResponse: true:
                  enableClearMessage: true,
    3. Add the following to enable voice recognition: 
                  enableSpeech: true,
    4. Refresh your browser and note that the widget now has a microphone Mic icon at the bottom right of the input field and the Clear Conversation Clear icon option in header.
    5. Click the microphone and say something like, "Order pizza" or "Get me a pizza". Complete the order using only voice commands. The voice indicator tells you if your voice commands are loud enough to be captured. The skill will also understand shortcuts -- the (CH) in (CH)eese Basic. Utter the two-letter shortcuts as separate letters ("C-H") to order Cheese Basic, for example.
      6. Description of voice_visualizer.png follows
      Description of the illustration

    Get User Feedback

    In this step, you're going to give users the option of rating the skill and leaving feedback by adding a System.Feedback component to the dialog flow.

    Description of sample-arch-image.png follows
    Description of the illustration

    1. Click Flows in the left navbar.
    2. In the confirmOrder state:
      • Set keepTurn to true (keepTurn: true) to allow the skill to retain control of the conversation.
      • In the transitions node, replace return: "done" with 
              next: "getUserFeedback"
        . When you're done, the confirmOrder state should look like this:
          confirmOrder:
    component: "System.Output"
    properties:
      text: "Your ${pizzaSize.value?lower_case} ${pizzaType.value?capitalize} pizza is on its way."
      keepTurn: true
      translate:
    transitions:
      next: "getUserFeedback"
    3. Add the following sequence after the confirmOrder state:
            
  getUserFeedback:
    component: "System.Feedback"
    properties: 
      threshold: 2
      maxRating: 5
      enableTextFeedback: true
    transitions:
      actions:
        above: "positiveFeedback"
        below: "negativeFeedback"
        cancel: "cancelFeedback"
  positiveFeedback:
    component: "System.Output"
    properties:
      text: "Thank you for your ${system.userFeedbackRating.value}-star rating."
    transitions:
      return: "done"
  negativeFeedback:
    component: "System.Output"
    properties:
      text: "Thank you for providing your feedback."
    transitions:
      return: "done"
  cancelFeedback:
    component: "System.Output"
    properties:
      text: "Feedback cancelled."
    transitions:
      return: "done"

      Note the values set for the maxRating and threshold properties, which set the highest rating that users can submit (five stars in this case) and the threshold for evaluating positive and negative feedback (two stars). When users give the skill a negative review, the skill prompts them for comments.

    4. Click Validate to check the indentation.
    5. Type, or say, "Order Pizza", then complete the order.
    6. Choose a positive rating (3, 4 or 5) using a voice command or text input.
    7. Click Reset.
    8. Order the pizza again, but this time, choose a negative rating (1 or 2).
    9. When prompted, enter your feedback. By default, all voice messages are sent automatically after the voice capture. However, if you misspoke, or if the skill's recognition is incorrect, you may want to edit a voice message before sending it. To disable the autosend for voice messages, add the following property to settings.js directly after enableSpeech: true: 
                  enableSpeechAutoSend: false,
      With autosend disabled, you need to click Send after each of your messages.

    You've completed the ODA Web SDK tutorial, which gives you an idea of the things that you can do with the SDK.