General

1. Select New Search Flow from the main page.

General Settings

1. Search Flow ID - Enter a unique ID for this search flow.

2. Search flow name - Enter a user facing name for this search flow.

3. Description - Enter a description for this search flow.

   • This is displayed below the name in the search flow for users, it should help them understand when to use it.

4. Type - Select the type of search flow from the dropdown.

   • Search, Search then model or Result.

5. History Source ID - Select a source to store previous generative chats. This allows users to search for previous chats.

   • To use this feature the 'Aiimi Insight Engine AI History' data model and an 'AI History' source must exist.

     • The 'Aiimi Insight Engine AI History' data models are created using Index Utilities.

     • The 'AI History' source can be created under source configurations. Select AI History as the type, and make sure you have selected the 'Aiimi Insight Engine AI History' data models.

6. Sources - Select the sources used from the dropdown to limit the sources used in this search flow.

7. Access Control Groups - Enter the access groups who can use this to limit the users who can see and use this search flow.

8. Default - If checked users with access to this search flow will automatically use it when they search.

   • If a user has more than one default then due to access groups the first 'default' in alphabetical order will be used.

Search Flow Features

The settings available within Search Flow Features will vary depending on the type selected.

• Advanced Settings - Check this to enable the models advanced settings. For example, a generative AI temperature settings will be shown to the user.

  • If your model doesn't have any advanced settings it may seem like it didn't work.

• Lenses - If checked the search flow option will be available on all lenses.

• ChatBot - Check this to include this 'search then model' flow in the ChatBot.

• Verify - If checked, users will have access to an experimental feature that highlights facts in generative responses. It then cross checks that across the results that were used to generate the response.

Shared Model Features

The settings available within Shared Model Features will vary depending on the type selected.

• Max Number of Results for Model - Enter the maximum number of results sent to a generative model. For example, if your search returns 100 and the max is 10, the first 10 will be sent.

• Extractive - Check this to use an extractive model in this flow.

  • An extractive model must be configured in the model steps tab.

• Generative - Check this to use a generative model in the flow.

  • A generative model must be configured in the model steps tab.

• Open Book - If checked, the model assume search results will be passed to it.

• Multi Result Execution - If checked, the model will allow you to pass more then one result.

• Closed Book - If checked, the models can be used without search results or any context.

  • Any answer will come from the models memory (i.e. its training data).

• URL - If checked, URLs can be passed to the model as input.

  • For example, the contents of a web page can be passed to an AI model.

• Free Text - If checked, free text can be passed as input.

• Risk Scan - If checked, the risk scanning feature will be enabled. This allows users to check generated content for sentiment and toxicity.