Page tree

Setting Up AI

The first time you enable AI in your  Agiloft system, you need to do some initial setup to enable the feature and configure the models you want to use. After initial setup, you still use the same process to change which AI Capabilities are running, and to add or remove models.  


AI features require either an Advanced or Premium package, with the AI Platform license add-on.

Initial Setup

To enable AI in your system, access the AI Credentials table. You can search for this table directly, or access it through the following steps:

  1. Go to Setup > Integration.
  2. Under AI, you will see a button for either Deploy or Configure. If you see Deploy, click it, and then click Configure. Otherwise, just click Configure.
  3. Wait until the AI Credentials page loads.
  4. Choose an account. The accounts are listed below.
    1. Local Agiloft Server Models: Holds models that are hosted on the same server as the KB.
    2. Agiloft Shared Hosting Models: Holds models that are hosted on Agiloft's dedicated servers.
    3. Default Amazon SageMaker Account: Holds models that are hosted on AWS. This allows you to host additional machine learning models that are either custom built or imported from an outside source.

When you first set up AI, you must configure either the Agiloft Shared Hosting Model credential or the Default Amazon SageMaker Account credential so your system can access machine learning models. This process is described in Configuring AI Credentials. AI Models hosted on the two Agiloft servers are not trainable, whereas models hosted on SageMaker are trainable. Once your credentials have been activated, you can then move models onto the Local Agiloft Server Models credential if you would like. For more information about moving models, visit the Move to Local section.

Amazon SageMaker is accessed through an AWS Account. If you're interested in using Amazon SageMaker with the Agiloft AI Core and do not have the option of creating your own AWS account, Agiloft can create and maintain an AWS account for you. However, it is recommended that you use your own AWS accounts whenever possible. If you choose to have Agiloft create and maintain an AWS account for you, the cost of maintaining this account is passed on as part of your regular invoice.

Configuring AI Credentials

To connect to Agiloft Shared Hosting Models, contact Sales to have a set of credentials and bucket name created for your organization. When you have the credentials ready:

  1. Navigate to the AI Credentials table.
  2. Click Agiloft Shared Hosting Models account to open it.
  3. Click Edit.
  4. Enter the credentials you received in the Access Key and Secret Key fields. If you did not already receive credentials, click the Get Keys button to generate Access Key and Secret Key values.
  5. If the record doesn't automatically save, click Save and navigate to the Models tab to see the models you can access.

If you have a SageMaker account you want to connect:

  1. Click Default Amazon SageMaker Account to open it.
  2. Click Edit.
  3. Enter the credentials for your account in the Access Key and Secret Key fields in Agiloft. To find this information in AWS, refer to Understanding and Getting Your Security Credentials.
  4. Click Save and Proceed to Models to see your SageMaker models.

Moving to Local

You might want to move a model to your local server. Hosting models locally can improve security if you run inference on your local server, and saves the cost of hosting models on SageMaker. However, keep in mind that models might run slower if they are moved onto a local server, and that models based on SageMaker algorithms might not allow you to run inference locally. You do not need to enter any Secret or Access Keys for local credentials.

To add models to the local server, convert a model from either the Agiloft Shared Hosting Models credential or the Default Amazon SageMaker Account credential. To convert a model:

  1. Open the AI Model record.
  2. Click Convert to Local Model. A validation message says that the download has begun. It might take a few minutes.
  3. When the download is complete, you receive a pop-up notification that the new record has been created.
  4. Go to the new AI Model record under the Local Agiloft Server Models credential. It has a Status of Disconnected.
  5. To activate the model, click Start Model.

Starting and Stopping Models

If you've already set up some models and you just want to start or stop them from running:

  1. Go to Setup > Integration, click Configure under AI, open the account hosting the model, and go to the Models tab.
  2. Open the model you want to run.
  3. Click Start Model or Stop Model to toggle the model on or off.