Publishing and deploying your assistant#

To this point, acting as an Assistant Builder, you built the assistant, configured conversational search, and added skills and automations. You tested your assistant by using the preview capability of AI Assistant Builder. The preview capability is a closed environment for experimenting with prompts.

After your assistant is finalized, you can publish it to make it available to users. Each assistant that you create comes with two environments: draft and live. You configured your assistant in the draft environment. Each environment has its own set of IDs, URLs, and service credentials referenced by external services.

The Environments page in the AI assistant builder has tabs for managing both the Draft environment and the Live environment:

The Draft environment contains all your in-progress work in the Actions, Preview, and Publish pages. Use the Draft environment tab to build out your assistant and use for internal testing before deployment. Any integrations (channels) that you use for the Draft environment are unique to that environment, and changes to draft integrations don’t affect the Live environment.

Publish the assistant#

Each time that you publish, you’re creating a new version of the assistant, for example V1. When you publish your content, you’re creating a snapshot of the draft content, resulting in a version.

Versions do not contain integration configurations or environment settings

Published versions contain all of the content from actions, including settings and variables. However, versions do not contain integration configurations or environment settings. Integration configurations and environment settings must be configured manually in each environment.

For managing quality-control and versioning, the Live environment is the version of the assistant to give to users.

Follow these steps to publish the first version of your assistant by using Assistant Builder:

Hover over the Home icon () and click Publish.
Click Publish.
Enter a description of the changes (a), set the environment to Live (b), and then click Publish (c).

Configure the live environment#

Important: When the live environment is created, the settings from the draft environment are not carried over (for example, the configuration of the OpenSearch instance used for conversational search).

Hover over the Home icon () and click Environments.
Click Live.
Click Web chat.
Customize the live assistant as you see fit.

On the Style tab, you’re able to set the Assistant name that is displayed on the chat window when users are interacting with the assistant. For pilots or demos, consider personalizing this name for the client. Also in the Style tab, you can set the themes and display settings of the chat windows, including the ability to enable the IBM Watermark and enable streaming.

On the Home tab, you enable and customize the greeting message from the assistant when the user accesses the assistant chat. You are also able to set Conversation starters that are displayed in the chat window. When selected by the user, the text of the conversation starter is sent as a prompt, so it is important that your assistant is trained and tested to answer correctly. It is highly recommended to remove the default conversation starters and create your own. The ability to add a Background style for the assistant chat window is on the home screen tab.

Explore all the other tabs.

Customize your live environment.

For this lab, toggle Streaming on and turn Suggestions off on the Suggestions tab. You may also want to change the theme to Dark to differentiate your draft and live environments.
Click Save and exit.
Click Add in the Search tile.
Click Custom service.
Enter the URL for your bring-your-own-search (BYOS) engine (a), select Basic authentication for the authentication type (b), enter admin for the Username (c), enter the password that you specified in the wrapper-creds.yaml file for the Password (d), and then click Next (e).

Use the correct URL and authentication type!

Use the URL and credentials for your BYOS OpenSearch engine created earlier here.
Verify conversational search is enabled and click Save.
Update the Custom service settings (a-f), click Save (g), and then click Close (h).
Customize the settings.

This is your assistant. Feel free to customize the settings. The settings shown below reflect the changes made earlier in the lab guide to the draft version of the assistant. This includes the Metadata field to weigh ingested client documents higher using:
```
{"doc_weight":
{"product_docs":0.2,
"customer_docs":0.8},
"ibm_indices":"*_ibm_docs_slate,*_ibm_redbooks_slate",
"standardize":true,
"customer_indices":"customer_*"
}
```

Connect the skills to the live environment#

Click Skill sets in the main menu.
Select your assistant's live instance in the Skill sets list.
Click Connections.
Search for the application name you specified earlier.
Click the ellipses () for your app and then click Connect app.
Click Connect app.
Enter the username (a) and password (b) using the username (admin) and password for your IBM Technology Zone (ITZ) watsonx Assistant for Z Pilot - AAP & z/OS reservation, and then click Connect app (c).

Learn more about publishing your assistant and creating live environments here.

Deploy the assistant#

After configuring your assistant’s settings and publishing, the final step is to deploy your assistant, which can be done across various channels depending on the use case.

Several options exist for deploying your assistant through channels and integrations to satisfy the use cases that you might encounter. Learn more about all the deployment options here.

For this lab, deploy the assistant by using the web chat integration. The web chat integration provides an assistant interface that can integrate with a website. Learn more about the web chat integration here.

Open the Environments page in the AI assistant builder.
Click Web chat for the Live environment.
Click the Embed tab.
Copy and record the integrationID, region, and serviceInstanceID values.

In a text editor, create a file that is named Watson Assistant Chat.html and paste the following text in the file.

File name:

Watson Assistant Chat.html

File contents:

<html lang="en">
<head>
<title>Watson Assistant Chat</title>
<meta name="viewport" content="width=device-width, initial-scale=1">

<style>
  .WebChatContainer {
    position: absolute;
    left: 0;
    right: 0;
    top: 0;
    bottom: 0;
  }
</style>
</head>
<body>

<div class="WebChatContainer"/>

<script>
const element = document.querySelector('.WebChatContainer');

window.watsonAssistantChatOptions = {
    integrationID: "<YOUR INTEGRATION ID>", // The ID of this integration.
    region: "<YOUR REGION>", // The region your integration is hosted in.
    serviceInstanceID: "<YOUR SERVICE INSTANCE ID>", // The ID of your service instance.
    element,

    openChatByDefault: true,
    hideCloseButton: true,

    layout: {
      showFrame: false,
      hasContentMaxWidth: true,
    },

    onLoad: async (instance) => {
      window.WACInstance = instance;
      await instance.render();
    }
};

setTimeout(function() {
    const t = document.createElement('script');
    t.src = 'https://web-chat.global.assistant.test.watson.appdomain.cloud/versions/' + (window.watsonAssistantChatOptions.clientVersion || 'latest') + '/WatsonAssistantChatEntry.js';
    document.head.appendChild(t);
});
</script>

</body>
</html>

Before modification:

After modification:

Open the Watson Assistant Chat.html file in a web browser.

Your assistant is now live. Explore some of the earlier prompts to verify that the assistant is accessing the ingested documents and your skills and skill flows are active.

Wait 5-10 seconds before clicking apply on skill actions.

Prompts to try:

What is z/OS continuous delivery?

Get z/OS facts

Show me z/OS facts

Gather and display z/OS facts