Accessing AI Engine Configuration

To access the AI Engine Configuration page, follow these steps:

1. In the main menu, under Application Setup, click AI Engine Configuration. The AI Engine Configuration page opens.

   

   Column header descriptions

   COLUMN NAME	DESCRIPTION
   AI Engine	Displays the AI engine name.
   Action	Link to configure the AI engine.

    

2. Under Action column, click Configure link associated with the desired AI engine. The AI Engine name: OpenAI window opens.

   

Provide values for the displayed parameters based on the descriptions given in the below table. (For detailed description, refer Azure OpenAI documentation.)

PARAMETER	DESCRIPTION
API Key*	Enter the Azure API service key for establishing connection.
Deployment Name*	Enter the model deployment name of Azure Open AI model. Note Ensure that the Deployment Name specified during deployment matches the exact name registered at Azure OpenAI. This specific name is crucial for seamless integration, as any deviation will result in a 404 error when making API calls.
Endpoint*	Enter the Azure API service endpoint URL.
Frequency Penalty*	Frequency penalty controls the repetition of same tokens in the generated response. Higher penalty reduces token repetition. Enter the frequency penalty.
Generation Sample Count*	Determine the number of completions generated before forming a response. Enter the generation sample count.
Max Tokens*	Set the maximum number of tokens in a single request.
Nucleus Sampling Factor*	Provide an alternative to temperature for token selection. Enter the nucleus sampling factor.
Presence Penalty*	Reduce the probability of repeating words in predicted text. Enter the presence penalty.
Temperature*	Control output randomness; lower values favor high-probability words. Enter the temperature.
*Mandatory field

3. Click Save to save the details.

 

Extracting data on MI screen

Once an AI engine is selected for a subtype, a new Prompt text field is visible, and a new Suggestion field type is available when you set up ontology. For more details, refer to Setting up ontology.

Prompt text field

Once an AI engine is selected for a subtype, a new Prompt text field is visible (see below image). Enter specific text prompts (e.g., "Total amount") for AI data extraction. This text field is available for all the field types.

Note: You need to scroll down to see the Prompt field.

Suggestion field type

In addition a new Suggestion field type (see below image) is available when setting up ontology. For more details, refer to Setting up ontology. This field type allows you to input a statement prompt (e.g., "What is invoice total amount") for AI-driven data extraction.

Ways to extract data

Depending on your selection, two methods are available for data extraction using AI engines during MI & QC:

1. For Suggestion field type: Enter a question prompt, and the extracted data will be in answer format (See the image above).

   If you want to change the prompt for the suggestion field, click on the icon next to the suggestion field (this icon is only visible if the Suggestion field type is selected). A field name prompt window opens (see the image below). Change the prompt and click Extract to extract the data.

   

2. For other field types: Use the Extract Value icon on the quick access toolbar to extract data with entered prompts.