Forum Widgets
Latest Discussions
Use Video Cards in Health Bot using Web Chat
There are several ways to display video content inside a Health Bot conversation using WebChat client. One way is to use Adaptive Cards. But if you don't need all the layout features of an adaptive card, you can also use a much simpler "VideoCard" attachment as part of a prompt or a statement in your Health Bot scenario flow. Based on the video source url, it will host YouTube or Vimeo video clips inside an iframe in the WebChat widget. Create a prompt or a statement step, click on adding a "Dynamic Card" and paste code snippet such as this: (function(){ return [ { contentType: "application/vnd.microsoft.card.video", content: { media: [{ url: "https://youtu.be/rwe2291YT8Rxw?si=t5m40pOGou5XdFqm" }], } } ] })() This will generate an attachment JSON that will be added to the prompt/statement activity. Obviously, you can use an expression instead of the hard codes URL link. When running the bot scenario, in WebChat widget, you will see the prompt hosting the native player with all the features.Understanding Health Bot's Logging Custom Dimensions
Microsoft Health Bot has the ability to emit custom logging into customer supplied Application Insights resource using customer's instrumentation key. See here, for more details. There is also the ability to emit events programmatically inside the scenario's "action" step. For Example: When this code runs, it will emit a custom event that can be viewed using Application Insights viewer. It will contain the payload of this specific call inside the Custom Dimensions property of the event. Event name: As specified in the session.logCustomEvent call Custom Dimensions properties: callstack: The full stack of the scenarios in case there are scenarios that call other scenarios channelId: Chat channel in which this event happened. eg. Web chat, Teams, IVR etc... conv_id: Conversation Id of this conversation session. This is a hashed value, but it can allow the user to track the history of a specific session. Note however that the SALT value creating this hash is being replaced daily. correlation_id: It's an internal id that uniquely identifies this API call as it goes through our APIM (API Management) resource. The customer can give this id to the support engineer to assist debugging by the service team. dialogName: Scenario Id that emitted this event. eventId: Unique identifier of this logging event. locale: The client's locale while emitting this event. Client can switch locales mid conversation. offerid: azurehealthbot payload: The custom payload (JSON format) passed by this specific call. region: The deployment region of this Health Bot speaker: The speaker of this event. Can be either "bot" or "user" stepid: Internal unique id of this "action" step that emitted the event. If you have several such action step in one scenario, it can be a bit difficult to tell which made this call. To solve this, you can select this scenario and press the "export" button in the toolbar. This will download the scenario in JSON format where you can locate the "action" step and retrieve its id field. stepType: The step type that emitted this event. Any step with JavaScript editor can be used to emit the custom event. tenantId/tenantName: Unique name and id you the customer user_id: Hash value of the end user emitting the event. You can track the events of a specific user throughout the conversation using this id. Note however that the SALT value creating this hash is being replaced daily.
Getting Started with Healthbot using Customer Generated Sources
I am trying to get our customer generated data from AI Search integrated with the Healthbot. Here are the steps: 1. Using OpenAI Studio, file upload in new storage container and specified AI Search instance and created a new Index. Validated the index could be searched based on the context of the single file upload. 2. Using OpenAI Studio, tested the chat function and it returned the right results and referenced the file 3. When entering Azure AI Search parameters in the data connection, I entered in the index name and key then chose Vector as the search type. I also verified my index fields is using contentVector. The built scenario always falls back to medline and never returns the customer data, is there any recommendations on where to check? This use to work when using the import data wizard in AI Search but even that index fails to return results in healthbot but works just fine in AI Search.Configuring WhatsApp channel for Microsoft Health Bot
Microsoft Health Bot leverages Twilio’s WhatsApp API to enable chatting with WhatsApp users. To enable this feature, you first need to set up a Twilio account and a WhatsApp Business account (WABA). In this guide, we will use the WhatsApp Sandbox environment to configure the channel. 1. Under Integration/Channels, enable the WhatsApp channel 2. Paste in the Phone number you have obtained from Twilio. This phone number should be registered as WhatsApp Business number. 3. Paste your Twilio Account SID and you Auth Token taken from Twilio's account page. 4. Save these settings. 5. Re-Open the WhatsApp channel and notice that the "Service URL endpoint" now appears. Copy this URL. You will need it to paste into the WhatsApp configuration in Twilio. 6. Navigate to Messaging/Send WhatsApp Message page and paste the URL you copied from Health Bot into the "Sandbox Configuration when a Message comes in" URL. Method should ne "POST". 7. Test it by sending the specified message on this page to be added to the Sandbox participants list. From this point on, every message you send to this number will be routed to the Health Bot via the service URL. Any response from the Health Bot will be sent to your WhatsApp client.Upcoming Backup Encryption Upgrade for Microsoft Healthcare Bot Users
To improve security, we will be upgrading our backup encryption system on August 1, 2024. With this upgrade any Microsoft Healthcare Bot backup made before Sep 28th 2023 will not work. Important If you haven't made any backups after September 28 2023, please create a new backup before August 1, 2024. After this date, backups older than a year will no longer be usable for restoring. If you have made backups in the last year, no action is needed. Feel free to reach out if you have any questions.Azure AI Health Bot – now supports Microsoft Entra Access Management
We are excited to announce the introduction of Microsoft Entra Access Management support in the Azure AI Health Bot. This enhancement increases security by leveraging the robust and proven capabilities of Microsoft Entra. Customers interested in this feature can opt-in by navigating to the User Management page and enabling the Microsoft Entra Access Management feature. This feature can only be enabled for users who have the Health Bot Admin role in the Azure access control identity-access-management (IAM) pane. When Microsoft Entra Access Management is enabled, all users and roles should be managed through Azure Access control identity-access-management (IAM) pane. The Access Control (IAM) now contains the same Azure AI Health Bot roles in Azure, such as Health Bot Admin, Health Bot Editor and Health Bot Reader. When the Microsoft Entra Access Management feature is enabled, the User Management page will be read-only. All users in the Management Portal page will need to be manually added with the right roles through the Azure Access Control (IAM) page in the Azure Portal. You can read more on the Microsoft Entra Access Management features on our public documentation pageFailed fetching CLU intents, please try again.
I have created a CLU language model with intents, variables etc. I have trained the language model and deployed it. I was able to successfully test the deployment aswell. I am trying to attach this CLU model in Healthbot. I have given all the details necessary to add the language to bot. The details included subscription key, deployment name, project name etc. When I click fetch CLU intents, I get 'failed fetching CLU intents, please try again'. I double checked on the configurations and they are looked good but I am not sure why I get this error. Could someone please help resolve this issue.Accessing "Conversation", "User" or "Scenario" entire objects is not allowed
Health bot uses three scopes of variables: Conversation - Variables that store data throughout the entire conversation. They are removed when the conversation ends. User - Variables that are stored in the context of the User Id that interacts with the bot. They are stored until the user asks to remove them. They are used so that we can retrieve user specific data, like birthdate, without asking it each time the same user interacts with the bot. Scenario - Variables that are stored in the context of the scenario and kept until the scenario ends. In analogy to normal programming language, they are like function variables that are allocated on the stack. If scenario A calls scenario B, each of the scenarios can have the same variable "foo" and each scenario will hold a different value. This allows us to write contained and reusable scenarios without the fear of "side effects". The syntax for accessing each variable scope is as follows: "This is a welcome message " + conversation.welcomeMessage "conversation" is not a real object, it's just a scope of the welcomeMessage variable. Therefore, we don't allow accessing the "conversation" as plain JavaScript object since it's really a reserved word. To access specific variables within the conversation scope, you can use the assign step and "pluck" the variables within the conversation scope into a new object. For example: We define a new object on the scenario scope called "dumpObject", we then "cherry pick", variables from scenario or conversation scopes. We later can use this object whenever we like, for example use it to dump onto custom telemetry.
Azure Health Bot using Azure Container Instance (ACI) for enhanced security.
We are pleased to announce our latest security improvement we have introduced in our platform. Our customers who used the production ready (Standard plan) instances will now enjoy the use of their own dedicated Azure Container Instance (ACI) that will execute their JavaScript code in their own isolated container. All ACIs operate within a separate subnet in our AKS cluster's virtual network (VNet), and a unified network security group (NSG) is applied to all ACIs to prevent any outgoing communication. As part of this security enhancement JavaScript functions, you create will not be directly shared among steps in the scenario. To enable this, we have created a Global Action step you can utilize to have JavaScript code that is sharable with every other step in the scenario: For more information about the available objects and JS packages please visit our documentation. To learn more about ACI, visit Azure Container Instances | Microsoft Azure
