Which GenAI Autopilot URL do we use?

Pega-as-a-Service
, , , , ,
1

For using the autopilot function in app studio, we have to configure an endpoint first.

In de community edition, Pega uses an internal url which is retrieved using “D_pyGetPegaGenAIGatewayServiceBaseURL”. For connector “pyConnectToGenerativeAI”.

It is recommended to use Azure OpenAI, and the configuration would be as follows:

https://[YOUR-RESOURCE-NAME].openai.azure.com/openai/deployments/[DEPLOYMENT-NAME]/chat/completions?api-version=[API-VERSION]

  • Replace placeholders:
    • [YOUR-RESOURCE-NAME] → Your Azure OpenAI resource name
    • [DEPLOYMENT-NAME] → Your model deployment name (e.g., gpt-4-turbo)
    • [API-VERSION] → Latest supported version (e.g., 2024-02-15-preview)

Example: https://my-ai-resource.openai.azure.com/openai/deployments/pega-autopilot-gpt4/chat/completions?api-version=2024-02-15-preview

Now the question arrises, can we only use autopilot if we have a paid version in place of OpenAI which provides access to the api.

OR can we try out the autopilot functionality using any other LLM service. If so, are there any mocks or free services we can connect to?

2

@Bob Reawaruw please clarify - are you using the Community Edition (this is the free trial software you mention)?

Have you already followed all of the instructions for Enabling Pega GenAI Coach ?

Don’t miss this part of the instructions, the enable the toggle bit Connecting to the generative AI provider in Pega Cloud

Pega GenAI Autopilot is primarily designed to work with Microsoft Azure OpenAI Service. Pega explicitly states that “Pegasystems Inc. only authorizes using Pega GenAI in Pega Cloud or by using Pega GenAI PremBridge to connect to Pega Cloud.” While the documentation mentions that you can “customize Connect Generative AI Rules to fit your development needs,” it also clearly states that “any unauthorized configuration of Pega GenAI is not supported by Pega and should be expected to stop working in future versions of Pega Platform.”

Important Considerations

  1. Pega GenAI features (including Autopilot) are available exclusively in:
    • Pega Cloud, or
    • Through Pega GenAI PremBridge for client-managed deployments
  2. Before using Pega GenAI in Pega Cloud, you must:
    • Connect to the required AI provider and model
    • Choose an appropriate scope for the Pega GenAI toggle
    • Complete the required configuration

hvc.png
hvc.png1231×239 44.8 KB

ea.png
ea.png1750×347 61.1 KB

References:

3

@MarijeSchillernThanks for the elaborated reply!

So for the Pega GenAI Coach, that could be something to implement in the future. As of now, we aim for the implementation of Autopilot as a first base of information for new ant current developers.

As we ar using a very recent on premise installation, PreMBridge would be a solution to incorporate.

After reading this documentation, it seems that for a premise instance to use GenAI, it should have a companion application in the cloud. This companion application would handle all GenAI request, or rather tunnel it.

Are we understanding the documentation correctly?

4

Yes, your understanding of the documentation is correct. For an on-premise Pega instance to use GenAI capabilities (including Autopilot), it does require a companion application in the cloud that handles all GenAI requests through tunneling.

The companion application is officially called “Pega GenAI PremBridge” and serves as the essential connection between your on-premise installation and the Pega GenAI services in the cloud. This architecture ensures that even client-managed deployments can leverage the powerful GenAI capabilities while maintaining your existing on-premise infrastructure.

Here’s how the process works in more detail:

  1. Your on-premise Pega Platform (which must be running release 24.1 or later) connects to the PremBridge application hosted in Pega Cloud.
  2. The PremBridge application then forwards all GenAI requests to the appropriate GenAI services and returns the responses back to your on-premise installation.
  3. To set this up, you’ll need to:
    • Work with your Pega representative to request a Pega Cloud instance with the PremBridge application installed
    • Configure the GenAI Gateway URL in your on-premise installation (using a Dynamic System Setting called GenAIPremBridgeURL)
    • Set up the required OAuth authentication credentials

This approach ensures that your on-premise installation can securely access all GenAI capabilities, including Autopilot, while maintaining your existing deployment architecture.

References:
Integrating client-managed deployments with Pega GenAI
Implementation of Pega GenAI

@bansa3 @piekd can one of you comment further?