End users can interact with a VA only when it is published. The Kore.ai Experience Optimization Platform provides a publishing flow to allow administrators to review new VAs and updates before they are made available to end users. The publishing workflow applies to any component that impacts the end user.
Before you can publish your assistant, you need to configure at least one channel. Learn more about Channel Enablement.
Moving forward, we will discuss the publication statuses your assistant and its components go through, the publication process, excluded components, as well as the approval process.Â
Publication Statuses
Every assistant created in the Platform has two states:
- In Development: Developers can make changes to any of the components in this state. So, when the VA is first created it is by default in the In Development state. When you publish any components, a published state of these components is created and these components are also available in the In Development version of the VA so that you can continue to make any necessary changes to them.
- Published: The components that have been approved for publishing are available in the Published state of the VA. You cannot make changes to the Published version of the components. The developer can pick and choose what changes are to be published during the publishing process. Admins can review the changes and approve or reject the publish request.
The Publish Status dock item checks if the publishing is completed within the specific time period and shows the ongoing progress of the publishing process. Once the publishing is completed it shows whether the publish request is successful or failed with details.
You can switch between the two states using the drop-down on the top right side of any Virtual Assistant window.
If your bot is trained in one or more languages that use NLP Version 2, you will see a banner notifying you about the upcoming auto-upgrade to NLP Version 3.
Component Statuses
Bot tasks and flows pass through the following stages:
- In Progress: Developers have begun configuring the task but haven’t yet defined all the required configurations. This status applies only to Alert, Action, and Information tasks.
- Configured – The task configuration is complete but the Owner hasn’t yet published the task.
- Awaiting Approval – The Owner has published the task thus initiating a request to the Admin, but the Admin hasn’t yet approved the publishing request.
- Published – The task is published for personal, enterprise, or public use.
- Upgrade in Progress – Developers have created an upgraded version of the published task to change any configurations, but the configuration is not yet complete.
- Rejected – The Admin rejects the publishing request. An email is sent to the developer with comments from the Admin.
- Suspended – The Admin suspended use of a deployed VA in an enterprise. An email is sent to the developer with comments from the Admin. While the end-users cannot access the suspended VAs/tasks, developers can work with the In-development copy of the VA.
Notes:
|
Publishing Components
You can only publish a Virtual Assistant or its components if you are the owner. If you are a developer with access to editing the VA, you cannot publish it’ only the owner can.
To publish a VA, please follow the steps below:
- Open the VA whose updates you want to publish.
- Select Deploy from the top menu
- From the left menu select  Publish.
- The Publish page opens with all the components selected by default.
- Clear the checkboxes of any components you do not want to publish, and then click Next.
- Enter any useful comments for the Admin and click Confirm.
Note: If you select to publish components without including the NL Model, then any unpublished training data will not be published and will be associated with the auto-created in-development version of the respective tasks. |
Publishable Components
Every component that impacts end-user interactions or experience goes through the publishing workflow in the Platform. These components are categorized into sections. The following is the description and the list of components under these sections:
Tasks & Languages | ||
---|---|---|
Tasks (Dialog, Alert,, Panels, Widgets, Digital Forms, and Small Talk) |
|
|
Knowledge Graph (per language) |
|
Natural Language | |
---|---|
NL Model | The NL Model lets you publish only the NLP training data of specific intents without publishing the task definitions. When multiple users are working on a virtual assistant (VA), Bot developers or NLP trainers can independently publish their work without impacting the bot definitions of other users.
For example, an NLP trainer might modify an utterance to correct some issue in a dialog task. In this case, the NLP trainer would like to publish only the training data without including the task definition. Whenever an utterance is added to an already published task, you can select that task under the ML Utterances, Patterns, and Rules modules under the NL Model to publish only the training data of selected intents without including the task definitions. You can also publish the Trait Groups and other NL Model configurations independently. Note: If you have selected one or more Dialog Tasks from Publish -> Tasks -> Dialog Tasks in the Publish request, the training data of these selected tasks is published and you cannot customize the ML Utterances, Patterns and Rules in the NL Model. If you want to publish the only training data of specific intents, you must first deselect tasks from the Publish -> Tasks -> Dialog Tasks section. Follow the steps below to select the intents for which you want to publish the ML Utterances, Patterns and Rules. When you have already selected the NL Model components for specific intents customization and then try to select the Dialog Tasks a warning message appears with notification to Reset the publishable components. Steps to publish only the training data for specific intents using the NL Model:
|
Settings |
|
Channels | |
---|---|
Channels |
|
Extensions | |
---|---|
Event Handlers | The full set of selected events in the Published Version get replaced with the ones in the In Development version. |
BotKit | The BotKit configuration in the Published Version gets replaced with the configuration in the In Development version. |
Agent Transfer | The Agent Transfer configuration in the Published Version gets replaced with the configuration in the In Development version. |
Web / Mobile SDK | The Web / Mobile SDK Configuration configuration in the Published Version gets replaced with the configuration in the In Development version. |
Settings | |
---|---|
General Settings | Replaces General Settings in the Published version with the ones in the In Development version. |
Bot Variables | Replaces the Bot Variables in the Published version with the ones in the In Development version. |
PII Settings | Replaces the PII settings in the Published version with the ones in the In Development version. |
IVR Settings | Replaces the IVR Settings in the Published version with the ones in the In Development version. |
Hold & Resume Settings | Replaces the Hold & Resume Settings in the Published version with the ones in the In Development version. |
Custom Script | Replaces the Custom Script file in the Published version with the ones in the In Development version. |
Advanced Settings | Replaces the following in the Published version with the ones in the In Development version:
– Language Detection Preference – Subdomain – Show link for task setup/execution – Error Messages |
Components Excluded from the Publication WorkflowÂ
The publishing workflow doesn’t apply to the following components as they either do not impact end customers or do only when used as part of a task definition. When you access these components from either the In Development or Published version, they show the same information.
- Batch Testing (including custom suites and test reports)
- Manage Developer Access
- App Creation, Resetting, and Deletion
- Language enablement: When you enable a new language in the In Development version of the VA, the publishing workflow works similarly to publishing tasks – you need to publish all the components all over again by selecting the checkbox for the newly enabled language.
Publication Approval
When you publish a VA or any of its components, the Platform initiates a request to the Admin to approve it. Depending on the Purpose defined for the VA during its initial set up, the following happens:
- For an Employee VA: The Admin needs to select users from the enterprise Kore.ai account who can interact with the VA , and then approve the publishing request. Once that’s done, the selected users can use the VA in the selected channels.
- For a Consumer VA: Any user can use the VA in the selected channels soon after the Admin approves the publishing request.
Auto Approvals
Admins can set up auto-approval for any updates made to Enterprise or Consumer VAs using the following steps:
- Open the Xo Platform Admin Console at: https://bots.kore.ai/admin.
- Go to Bots Management > Consumer Bots or Bots Management > Enterprise Bots.
- Click the more icon for the required bot and select Bot Settings.
- On the Bot Settings window, select Auto Approve.
- Select Auto Approve all publish requests for this bot.