Skip to main content

How to Use AI Flow in Bots

Updated

This article introduces how to use the AI Flow block in MSPbots.

Background information

Langflow is a new visual framework for building multi-agent and RAG applications. Its intuitive interface allows for easy manipulation of AI building blocks, enabling users to prototype and turn ideas into powerful real-world solutions.

Langflow is integrated into the MSPbots system as the Flow integration. 

You can use the AI Flow block in the bot to invoke the workflows you created in Langflow and complete tasks within the workflow.

Prerequisites

  • A successful AI Flow integration with MSPbots. You can refer to AI Flow Integration Setup for instructions.
  • A working AI flow in the Langflow app. Get more information on the Langflow visual flow builder from the Langflow documentation.
  • Only users with the AI Admin or AI Trial role can edit the AI ​​Flow block.

Execute the AI flow in the bot

The following steps serve as an example to guide you in creating a bot that analyzes data from the "Open AI Analysis For Zendesk Manage Notes" widget by executing the "Sentiment Analysis" AI flow. It then saves the executed results to the "Zendesk Ticket Notes Sentiment Analysis" dataset, filters the executed results, and sends the filtered results as messages to the configured recipients.

  1. Create a new bot, refer to 1. Set up a new bot.
  2. Add and configure the following bot blocks.

    AI Flow.jpg

    1. Trigger - Configure the data source and the schedule for running the bot, refer to 2. Set up the bot trigger. 

      Langflow (21).jpg

    2. AI Flow - Select the Flow to be executed from the workflows dropdown list.

      AI_flow.jpg

      If you want to edit the selected workflow in our AI platform, click Edit AI Flow in the Workspace. This will launch our AI platform and open the corresponding workflow. This option is currently available only to users with the AI trial role.
    3. Save Prompt Result - Select a dataset to save the execution results.
      For example, selecting "Zendesk Ticket Notes Sentiment Analysis" means the results will be saved in the "Zendesk Ticket Notes Sentiment Analysis" dataset.

      Langflow (13).jpg

    4. DataFilter - Configure the filtering conditions for the executed results.
      For example, "Langflow_Result_sentiment Equals negative and Type=text" means filtering the negative text results.

      Langflow (14).jpg

    5. Card Alert - Edit the message content to be sent and specify the recipients of the message.

      Langflow (15).jpg

  3. Click Finish to complete the configuration.

    Langflow (18).jpg

  4. Click Save to keep your settings.

    AI Flow-3.jpg

  5. Activate the bot, refer to Activating a Bot on the MSPbots App.
  6. Verify activation, refer to How to Check if a Bot Is Activated.
  7. When this bot is triggered, it will send a message.
    For example, if the sentiment analysis AI flow results in negative, the bot will send a message to the recipients you have set.

    Langflow (20).jpg