Configuration

This section explains each of the properties in the configuration blade.

Fig 1: Azure OpenAI Configuration

General

Property	Description
Collection	This associates the Agent to a specific Collection (default to that of the current Data Stream).

Refer to Collections and Stream Hosts to understand more about Collections.

Connection

The Endpoint and Key Credential can be found in the Keys and Endpoint menu on the Azure Portal page.

Property	Description
Use Variables	Tick to select saved variables; or enter the values manually for the Endpoint, Key Credential, and Deployment Name.
Endpoint	The Endpoint Url of the Azure OpenAI instance.
Key Credential	A valid key for the Azure OpenAI instance.
Deployment Name	The Deployment Name of a deployed model.

Message

The Messages list contains the messages sent to the OpenAI model. The response is generated once every message has been sent, allowing you to split a command into separate messages.

Property	Description
Description	A unique, friendly name to describe the message that appears in the message list.
Order	Adjust this automatically generated value to rearrange the order in which the messages are sent.
Type	The options for the message originator are: Assistant: messages that establish a history of the AI's responses for it to draw from. System: messages that inform the AI of its role. User: messages that prompt a response. Read here for more on prompt engineering.
Message	The message to send. Insert columns from the input payload by enclosing them in double curly braces. For instance, to insert the value of the property ReadingNo, use the text "{{ReadingNo}}".

Fig 2: Azure OpenAI Message Configuration

Before configuring the AI Agent, please ensure that its input endpoint is connected to a parent Agent which will be sending data to it.

Endpoints

Endpoint	Description
Input	This endpoint is used to receive data from the parent Agent.
Output	Events received from the parent Agent are made available to this endpoint, with the OpenAI model's response appended in a new Response attribute (string).
Error	Errors encountered during the execution of this Agent will be published on this endpoint.