Skip to content

Files

Latest commit

cb1c54b · Dec 9, 2024

History

History
83 lines (40 loc) · 2.57 KB

Exercise05.md

File metadata and controls

83 lines (40 loc) · 2.57 KB

ChatGPT and Flow

Exercise 5 - Classify text

  • Using the same in-private browser instance, navigate to the Power Automate site if you are not already there.

  • In the top right of the screen, ensure the Environment is set to Dev One.

  • Choose + Create from the navigation on the left.

  • Next click on the Automated cloud flow button.

  • In the following dialog select When a file is created for OneDrive for Business then click Create

  • Click the When a file is created action.

  • Similar to the previous exercises we need to configure this trigger step. So select the step and choose a folder to use (eg. Root)

  • Next click on the plus sign and select Add an action

  • Search for prompt and click on the See more link under AI Builder

  • Choose AI Classify

  • Set the Input Text to File content

    inputtext

  • Set the Input Categories to the following:

    positive, negative
    
  • Add another Action

  • Search for notification and then select Send me an email notification

  • Click in the Subject text box and click on the lightning bolt and set to Finish reason

  • For Body select Text from the options

    notificationgpt

  • Click on Save and wait a few moments until you get a confirmation that the flow has been saved

  • Then click on Test and choose the radio button for Manually and finally click the Test button. Similar to the previous exercise you need to kickoff the test, switch the OneDrive and upload one of the files in the AIBuilderLabFiles\Reviews folder.

Note: You will get a notification about adding an Approval process. For this example it is not required but definitely a best practice with any AI generated content.

  • Now upload one of the review documents to OneDrive for Business

  • Review the emails sent in Outlook Web Access to see their classification.

Note: Sometimes more than one notification will come across.

Summary

In this exercise, you created a flow that leveraged a pre-created AI Builder model to parse and classify sentiment data out of uploaded review data all without writing any code.