Generative AI in Z Understand

Z Understand includes Generative AI capabilities to assist architects and developers when analyzing applications.

Generative AI capabilities include responding to AI chat queries about the application, such as identifying the business functions in the code and relationships between application resources. Additionally, it assists in generating descriptions and names when identifying business rules, which helps to populate a data dictionary for your application.

Semantic Insights using Generative AI

IBM® watsonx Code Assistant™ for Z Semantic Insights service in IBM Cloud® assists to generate natural language descriptions and business terms for code variables. Also, Semantic Insights uses Generative AI to help identify business rules and populate data dictionary.

AI agents

To answer questions about the application and provide insights, the chat uses a set of AI agents. These agents run in IBM Cloud and access data from on-prem databases to generate responses.

Note: Your application data is not stored in IBM Cloud.

The following agents are:

Orchestrator – Handles requests from the chat and directs to the right agent for answering, formulates the response, and sends back the answer to the chat.
App Insights – An agent that can query the database where the metadata about your application is saved and provides responses about relationships between application resources.
Code search – An agent that can find business functions in your application code by querying a Milvus database that has been populated by a semantic indexer.

Semantic indexer

The semantic indexer is part of the build utility for scanning application code. When semantic indexer runs, it uses Generative AI to generate explanations for every paragraph in the application and stores the output in a Milvus Database. This database is used by the Code search agent to support the Chat experience in the web UI.

Semantic indexer

Data sources

In addition to the analysis project that contains your application data and the Milvus database with code explanations, you can also upload files to further refine the AI output in the web UI, including a glossary that includes the terms and definitions relevant to your business, along with the naming conventions (such as acronyms) used in your code. This information is used when identifying business rules to provide better AI suggestions.

Note: There is no limit to the number of glossaries that you can upload for your application.