AI Universal Bot

The Aisera Gen AI platform allows you to create applications/bots in different domains and link them together with a single parent application/bot.

The specific use case is to serve relevant content from multiple domains to application/bot users based on their language and country, content in multiple languages, and content relevant to different countries.

Users interact with the Universal bot and ask question across multiple domains, multiple countries and in different languages. The universal bot and child bots together serve relevant content to the user based on the bot domains and the user's language, country.

High-Level Universal Bot Interactions

• The User sends a Request to the Universal bot

• The Universal bot detects the language in which user has asked the question

• The Universal Bot translates the user query to English. This is done via the language translation processor of Universal bot.

• The Universal bot domain classifier processor detects the domain

• The Request is routed to the domain specific Child bot

• The domain-specific Child bot interacts with various fulfillment engines like ICM, Neural Search and RAG.

• The Neural Search service within the Aisera platform runs the inference and finds documents that are served to the users.

  • Neural Search calls the Access Control Service to filter out content that the application/bot user is restricted from viewing, based on the user's permission level as well as the user’s language, country, department, region, or other metadata.

• The Access Control Service will applies two broad categories of filters that you can use to control which content is served to the user.

  • One is RBAC or Role-Based-Access-Control. For role-based control, you assign a custom or default role to each user and then configured which content a user with each specific role can view. For instance, in the Aisera Gen AI platform, the default roles can view Aisera Admin UI windows based on the Aisera Entities that are included within each window. In the computer industry, this is referred to as a hard filter because it is a highly restrictive and explicit access control rule that prevents access for specified users or groups, even if they have broader, inherited permissions.

  • The other category is related to user settings, such as a user’s location or language. The computer industry term for these settings is soft filter because it refers to a discretionary and flexible form of access control that can be easily overridden or modified. A simple example of content that contains metadata as a soft files is a Knowledge Base Article that lists public holidays based on the user's country or region.

• After the application/bot receives content from any fulfillment service, the content will be translated to the language in which the user submitted the Request.

Child Bots with Different Versions

The Universal Bot supports a mix of Conversational AI 1.0 and Conversational AI 2.0 Child bots.

While it is a best practice at this time to migrate your Conversational AI 1.0 Bots to Conversational AI 2.0 bots, if you're using a Universal Bot you can choose to independently migrate each Conv 1.0 Child bot app to Conv 2.0 app (and not wait for all child bots to complete migration before switching).

This section contains the following topics:

Create a Universal Bot

Multi-Language Support

Universal Bot Analytics

See Universal Operation Mode if you want all of your bots to share the same parameters.