Breadcrumbs

Microsoft Foundry AI Service Provider Reference

Prerequisites

The AI agent feature is only available in a Kubernetes setup.
To use PAS AI agents, you need an account with an AI service provider that is not provided by Scheer PAS. Currently, OpenAI and Mistral are supported as providers. You can also use other LLMs if they are hosted by Microsoft Foundry.

You need to create an AI Service Provider alias (see Aliases for more information) to use PAS AI agents in a Designer service. You have two options:

  • Microsoft Foundry AI Service Provider for models hosted in an Azure Environment (explained on this page)

  • AI Service Provider Alias for models of OpenAI or Mistral (refer to AI Service Provider Reference for details)

Azure Foundry AI Service Provider Alias

Azure AI Foundry is an AI development and operations platform in the Azure cloud that provides companies with access to a wide range of pre-trained models, tools, and features for developing AI agents.

Using an LLM hosted by Microsoft Foundry

If you are using an LLM hosted by Microsoft Foundry, ensure that it supports function calling, structured output, and has the Chat completions API at its disposal.

Attribute
(Name of Setting)

Description

Allowed Values / Examples

Name

Specify a name for the alias.

any string

Standard

API Endpoint

Please refer to the display in Microsoft Foundry for this information.

any string

Model

Select the LLM hosted by Microsoft Foundry you want to use.

Ensure that the model supports function calling, structured output, and has the Chat completions API at its disposal.

API Key

Enter the API key from the Microsoft Foundry project where the model was deployed.

any string

API Version

Please refer to the display in Microsoft Foundry for this information.

any string

Advanced

Parallel Tool Calls

Allows that tools used by the agent can be called up in parallel. Prevents tools from having unwanted side effects on each other when running simultaneously. Default is true.

Recommendation

Disabling this option can slow down the AI agent.
Do not disable this option unless you know what you are doing.

true

false

Temperature

Controls the randomness of the generated text. A higher temperature will result in more random and diverse text, while a lower temperature will result in more predictable text.
The default value is set by the model itself.

Recommendation

Do not explicitly set a temperature unless you know what you are doing.

any float between 0 and 1
⬇️

AIAgent_InsuranceCase_Example

Click here to download a simple example model that shows how use an AI agent to process unstructured data with Scheer PAS Designer.
The AI Agent feature is only available on Kubernetes. To execute this example, you need your own API key from OpenAI or Mistral.
You can also use other LLMs if they are hosted by Microsoft Foundry.

Related Content

📗

Related Pages:

📘

Related Documentation: