flaskAI Agent (Preview)

An agent feature that autonomously executes analysis and creates/edits Notebooks based on natural language instructions

circle-exclamation

AI Agent is a feature that autonomously performs data analysis and creates or edits Notebooks based on natural language instructions.

It can reference catalog and Notebook contents to carry out analysis aligned with your intent.

Key Features

  • Natural language instruction execution:

    • Simply enter instructions in the chat, and AI Agent will autonomously perform the necessary operations.

  • Content reference:

    • Can reference catalog and Notebook contents during execution.

  • Table exploration:

    • Automatically explores tables suitable for analysis based on catalog information and table column details.

  • Notebook creation and editing:

    • In addition to returning analysis results in the chat, AI Agent can also create new Notebooks or edit existing ones.

How to Use

Where to Launch

AI Agent is available from the following screens:

  • Chat panel: Launch from the Agent icon in the UI. When viewing a Notebook, it appears as a side panel on the right.

  • Full-screen: Launch from the side panel in the UI.

Modes

AI Agent has two modes:

  • ASK mode: Performs data analysis including query execution, but does not create or edit Notebooks. Use this when you want to ask questions about your data without modifying any Notebooks.

  • Agent mode: Executes all operations including Notebook creation, editing, and chart additions. Use this when you want to create Notebooks or add charts.

Specifying Context

Use @ to specify the context for AI Agent to reference.

View
Available Context

While viewing a Notebook

Tables, Notebooks, Saved Queries, SQL Blocks

Full-screen and other views

Tables, Notebooks, Saved Queries

Settings

Auto-approve Tool Execution

You can configure whether SQL queries are automatically executed per connection. For BigQuery, you can also set an execution size limit based on dry run results.

Auto-load Agent Context

You can configure whether agent context files are automatically loaded.

AI Profile

You can select which of your configured AI profiles to use.

Agent Context File

You can create an agent context folder at the following levels:

  • Workspace

  • Team space

  • Private folder

You can also create an agent context page for individual Notebooks using .agent.

Unsupported Pages

AI Agent is currently not available on the following pages:

  • Notebook Home

  • Query

  • Catalog

  • Workflow

  • Report

Data Privacy

These requests are sent to each LLM platform you choose but are not used for training purposes.

circle-info

For inquiries about the AI Agent feature, please contact us via chat or [email protected].

Last updated

Was this helpful?