Use Bing Search as a fallback in Copilot Studio

  • Article

Note

This article applies to classic chatbots only. To extend copilots created in Copilot Studio, you don't need to use Bot Framework Composer. All features described in this article are available for copilots, directly in Copilot Studio.

Enhance a classic chatbot by developing custom dialogs with Bot Framework Composer, and then adding them to your chatbot in Copilot Studio.

In this example, you learn how to set Bing search as a fallback answer in Copilot Studio by using Composer.

Important

Composer integration isn't available to users who only have the Teams Microsoft Copilot Studio license. You must have a trial or full Microsoft Copilot Studio license.

Prerequisites

Create a fallback trigger

  1. In Copilot Studio, open the chatbot from Example 2.

  2. Open your chatbot in Composer. For instructions on how to do so, see Get started with Bot Framework Composer.

  3. In the bot explorer, go to the main dialog. Select More options (...) then select Add new trigger.

  4. In the Create a trigger window, for What is the type of this trigger?, select Unknown intent. Select Submit.

  5. Go to the Bot Responses page, select Contoso Meal Delivery Service, then select Show code.

  6. Copy and paste the following code into the code view:

    # adaptivecardjson_BingSearch(user_utterance)
- ```
{    "$schema": "http://adaptivecards.io/schemas/adaptive-card.json",
    "type": "AdaptiveCard",
    "version": "1.2",
    "fallbackText": "This card requires Media to be viewed. Ask your platform to update to Adaptive Cards v1.1 for this and more!",
    "actions": [
        {
            "type": "Action.OpenUrl",
            "title": "Search with Bing",
            "url": "https://www.bing.com/search?q=${user_utterance}",
            "style": "positive"
        }
    ],
    "body": [
        {
            "type": "Image",
            "url": "https://img-prod-cms-rt-microsoft-com.akamaized.net/cms/api/am/imageFileData/RWe65Z?ver=2d4e&q=90&m=6&h=201&w=358&b=%23FFFFFFFF&l=f&o=t&aim=true"
        },
        {
            "type": "TextBlock",
            "text": "${user_utterance}",
            "wrap": true,
            "separator": true,
            "horizontalAlignment": "Center",
            "size": "Medium"

        }
    ]
}
```

    Screenshot of the Adaptive Card JSON added to bot responses.

  7. Copy and paste the following code into the same code view:

    # AdaptiveCardBingSearch(user_utterance)
[Activity
    Attachments = ${json(adaptivecardjson_BingSearch(user_utterance))}
]

    Screenshot of the activity added to the bot responses.

  8. Go to the Create page.

  9. In the main dialog, select the Unknown intent trigger.

  10. On the authoring canvas, select Add (+) then Send a response.

  11. In the properties pane, under Bot responses, and select Show code.

    Warning

    Adding the expression in the next step to the response editor instead of the code editor will result in the chatbot responding with raw JSON instead of an Adaptive Card.

  12. Copy and paste the following expression:

    - ${AdaptiveCardBingSearch(turn.activity.text)}

    Screenshot of the response for the send a response action.

Test your chatbot

  1. Publish your Composer content to make it available in your chatbot in Copilot Studio.

    Important

    Selecting Publish in Composer makes the changes available for testing, but doesn't automatically publish your bot.

    Publish your bot in Microsoft Copilot Studio to update your bot across all the channels it's connected to.

  2. In Copilot Studio, go to the Topics page to see your new OnUnknownIntent topic.

  3. Open the Test bot pane and make sure Track between topics is turned on. Enter "Is tofu vegan?" to start your conversation.

Next step

Display a form with a Submit button in Copilot Studio.