Skip to main content
Version: 10.2.8

Migrate AI Agent to Control Tower

The instruction provides guidance on how to make automations you develop appear in Control Tower as an instance of a new AI Agent entity.

I'm developing a brand new AI Agent to run on Work.AI​

Basic scenario​

ODF 2 is the recommended framework for the implementation of AI Agent if you create your project based on this guide. To proceed, do as follows:

  1. Go to the [ARTIFACT]-package module of your project.

  2. Open the meta-info.json file.

  3. Specify the following attributes in the file:

    • USE_CASE_CODE: set a unique identifier of the AI Agent entity; it is also used to identify the set of metrics to compare different versions of the AI Agent.

      Use a unique code that won't overlap with any other codes in the customer's environment. Maintain the same AI Agent code as you develop your AI Agent—don't change it from version to version.

    • USE_CASE_VERSION: specify the version number of your AI Agent, for instance, 10.8.18.

    • NAME: populate it with the AI Agent name.

  4. If you have analytics dashboards, place their assets into the project's bundle. For details, see the instruction.

  5. Build your project's Asset Bundle and deploy it to the environment. For detailed instructions, refer to Deploy Asset Bundle.

  6. To check that everything is done correctly, go to Control Tower and open Digital Workers > View All. You should see something like this.

Project-specific adjustments​

My project is more complex than the one provided in the guide, for example, I have multiple package modules that generate multiple Asset Bundles during the build phase. What am I supposed to do?​

Modify meta-info.json in every module that generates an Asset Bundle with the same attributes. We don't provide any tools to do it in one action.

Will this work for AI Agents based on the IDP template?​

Yes, it will. Since an IDP project template is an example of a project with multiple modules that generate an Asset Bundle, the advice above works. You need to modify respective meta-info.json in every module that generates the Asset Bundle specifying the same values of the USE_CASE_CODE, USE_CASE_VERSION, and NAME attributes.

We don't use the Asset Bundle Import functionality in our project. It's overkill for us. We deploy BCB to the development environment, build a Business Process there, and migrate the Business Process to higher environments. How do we make this work for us?​

The only way to have an AI Agent entity created in Control Tower is by uploading an Asset Bundle. The deployment via the Asset Bundle isn't overkill, it's an easier way to do the same deployment procedure. All you need is to put your Business Process into the business-process folder of your [ARTIFACT]-package module and build the Asset Bundle that will deploy both BCB and the Business Process to the target environment. See Deploy Asset Bundle for more details.

I have a skill deployed to earlier 10.x version of WorkFusion platform​

The guide below is applicable for the following use case:

  • You have a pre-10.2 WorkFusion platform installation.
  • You have one or more Business Processes that logically belong to the same AI Agent that you want to view when you dive into the AI Agent version details.

This scenario is similar to creating a new AI Agent as you can modify the package module of your project, build an Asset Bundle, and deploy it as if it's a new AI Agent.

  1. Go to the [ARTIFACT]-package module of your project.

  2. Open the /datastore folder and rename the CSV files in the folder. This is needed not to have your Data Stores content overwritten on the Asset Bundle import, which you'll have to do next. If you have any custom solution that allows you to version your project's assets, follow your project guidelines.

  3. Open the meta-info.json file.

  4. Specify the following attributes in the file:

    • USE_CASE_CODE: set a unique identifier of the AI Agent entity; it is also used for identifying the set of metrics to compare different AI Agent versions.

      Use a unique code that won't overlap with any other codes in the customer's environment. Maintain the same AI Agent code as you develop your AI Agent—don't change it from version to version.

    • USE_CASE_VERSION: specify the version number of your AI Agent, for instance, 10.8.18.

    • NAME: populate it with the AI Agent name.

  5. If you have analytics dashboards, place their assets into the project's bundle. For details, see the instruction.

  6. Build your project's Asset Bundle and deploy it to the environment. Refer to Deploy Asset Bundle. To ensure correct deployment, your AI Agent's Business Processes shouldn't be running at the target environment.

  7. To check that everything is done correctly, go to Control Tower and open Digital Workers > View All. You should see something like this:

  8. To continue maintaining the project, rename the CSV files in the /datastore folder to the original state.

Workaround: I don't have access to my AI Agent project, but I have an Asset Bundle​

If you have an Asset Bundle file but don't have access to the AI Agent project, you can modify your Asset Bundle to let the platform know about your AI Agent. Note that if your AI Agent project development continues and new versions are going to be deployed, you have to follow the recommended approach or repeat this workaround for every new version of the AI Agent.

  1. Unzip your Asset Bundle.

  2. Open the meta-info.json file in the root of the folder.

  3. Modify the file specifying the following attributes. The order doesn't matter, but you have to maintain a valid JSON file structure:

    • USE_CASE_CODE: set a unique identifier of the AI Agent entity; it is also used for identifying the set of metrics to compare different AI Agent versions.

      Use a unique code that won't overlap with any other codes in the customer's environment. Maintain the same AI Agent code as you develop your AI Agent—don't change it from version to version.

    • USE_CASE_VERSION: specify the version number of your AI Agent, for instance, 10.8.18.

    • NAME: populate it with the AI Agent name.

  4. If you have the /datastore folder, open it and remove all CSV files.

  5. Archive the assets into an Asset Bundle. Make sure it has the same structure and meta-info.json is still at the root level of the Asset Bundle structure.

  6. Open Control Tower and go to Digital Workers > View All.

  7. Click the Upload New button.

    info

    To view the Digital Workers > View All page, you should have the View Digital Workers permission. To view the Upload New button, you should have Export/Import and Advanced Package Import permissions. For more information on permissions, read the instruction.

  8. Click the Add button in the popup window and select your Asset Bundle file in the file selection dialog.

  9. After you confirm the file selection, the Asset Bundle file uploads to your environment, and you can cancel the process until the upload finishes. After that, the import process begins. You can track the import progress on the screen.

    If you see a disabled Finish and show Digital Worker button at the bottom of the screen, it indicates that your meta-info.json file is modified correctly. In the Asset bundle details section, you can see the attributes you provided: the AI Agent name, code, and version.

  10. After the successful installation, the imported Asset Bundle appears as an AI Agent in the corresponding group. You can also access the import log to see if there were any warnings during the import procedure. For details, click View import log file.

Although the approach might seem the easiest, it's not recommended as it doesn't tie AI Agent assets, except Business Processes, to the respective AI Agent. It can cause some new functionality that will use the internal AI-Agent-Asset relation to malfunction.

However, there might be some reasons to follow this approach. For example, you know your next version is almost ready and packaged in a way that supports its display on the Digital Workers page, and the current one is getting obsolete soon.

  1. Export all Business Processes related to your AI Agent using the Business Process Package Export feature.

  2. Create a folder named business-process.

  3. Put all BP packages into the business-process folder.

  4. Create an empty meta-info.json file in the same place where the business-process folder resides. The structure should look like this:

  5. Edit the meta-info.json file, specifying the following attributes:

    • USE_CASE_CODE: set a unique identifier of the AI Agent entity; it is also used for identifying the set of metrics to compare different AI Agent versions.

      Use a unique code that won't overlap with any other codes in the customer's environment. Maintain the same AI Agent code as you develop your AI Agent—don't change it from version to version.

    • USE_CASE_VERSION: specify the version number of your AI Agent, for instance, 10.8.18.

    • NAME: populate it with the AI Agent name.

    • WF_VERSION: specify your platform version, for example, 10.2 or 10.2.1.

  6. Archive the folder and file into the Asset Bundle. Make sure it has the correct structure and meta-info.json is at the Asset Bundle structure's root level.

  7. Open Control Tower and go to Digital Workers > View All.

  8. Click the Upload New button.

    info

    To view the Digital Workers > View All page, you should have the View Digital Workers permission. To view the Upload New button, you should have Export/Import and Advanced Package Import permissions. For more information on permissions, read the instruction.

  9. Click the Add button in the popup window and select your Asset Bundle file in the file selection dialog.

  10. After you confirm the file selection, the Asset Bundle file uploads to your environment, and you can cancel the process until the upload finishes. After that, the import process begins. You can track the import progress on the screen.

    If you see a disabled Finish and show Digital Worker button at the bottom of the screen, it indicates that your meta-info.json file is modified correctly. In the Asset bundle details section, you can see the attributes you provided: the AI Agent name, code, and version.

  11. After the successful installation, the imported Asset Bundle appears as an AI Agent in the corresponding group. You can also access the import log to see if there were any warnings during the import procedure. For details, click View import log file.