Configuring Your Assistants

Users can interact with an assistant by asking questions about their data. Each assistant is connected to one or more Knowledge Stores and makes calls to a configured large language model (LLM) vendor to generate responses.

Using semantic search, assistants identify information relevant to a user’s question based on meaning and similarity, not exact matches. This makes them well suited for tasks such as summarizing information, comparing data across sources, and exploring trends or patterns.

Assistants generally perform best when they are designed for a focused workflow or use case. For example, one assistant might focus on sales performance data, while another specializes in analyzing contracts. You can tailor each assistant’s Knowledge Stores, capabilities, and user access to match its intended purpose.

Understanding the Assistants Interface

Select Management > AI Assistants >Assistants, and you will see a table of your existing assistants.

Name

The assistant's display name used across the system.

LLM Vendor

The configured LLM vendor used by this assistant.

Status

Indicates whether the assistant is active and available for use.

Agents

Capabilities available to this assistant. Some agents are provided by the system, while others are created from Knowledge Stores. Agents created from Knowledge Stores provide access to the data in those Knowledge Stores.

Filtering the Display

To filter the display, use the drop-down menus and type-in boxes above the table.

Default Available Agents

Agents provided by default give assistants capabilities such as working with database metadata, generating visualizations and data models, and generating and executing SQL queries. The following agents are available by default:

Data Catalog Agent—Answers general and informational questions about the makeup of tables, columns, and schemas available in the database.
Graphing Agent—Creates data visualizations from SQL queries or data available in the conversation. It can also summarize and explain generated visualizations.
Data Model Agent—Generates relational data models from natural-language descriptions, including tables, primary and foreign keys, relationships, and CREATE TABLE SQL statements.
Text-to-SQL Agent —Handles SQL query generation, editing, and execution for user questions that request information to be retrieved from the database.

Creating an Assistant

Click the Management icon in the application menu.
Under AI Assistants, click Assistants.
Click +New Assistant.
Provide the following information:

Name

Required— Unique identifier for the Assistant.

Vendor

Required— The vendor you want to use with this assistant. You must set this up in advance in the Management >AI Assistants>Vendors.

Memory

Optional— Provide information that will be kept in the assistant's memory when answering questions.

User Roles

Optional— Select one or more roles allowed to interact with this assistant.

Agents

Optional— Select the agents this assistant can use. Agents provide specialized capabilities or access to data in configured Knowledge Stores. See Default Available Agents for a list of built-in agents and their capabilities.

Important:

To use an assistant in the SQL Editor, you must enable the Text-to-SQL agent when configuring the assistant.
Click Submit.

Editing an Existing Assistant

Click the Management icon in the application menu.
Under AI Assistants, click Assistants.
Click the Edit menu. A drop-down menu appears.
Click Edit Assistant.
Modify the fields you want to change.
Click Submit.