Icon

04 Agentic AI with Databricks AI

<p>This workflow demonstrates how to create a <strong>simple agent with Databricks AI</strong> that can process <strong>a table </strong>containing product feedback. The agent is able to send the table to various tools and return the updated table in its second output port. The first output port contains the <strong>conversation</strong>, where you can see the reasoning behind the tool choices and the results generated by each tool.</p>

04 Agentic AI with Databricks AI

This workflow demonstrates how to create a simple agent with Databricks AI that can process a table containing product feedback. The agent is able to send the table to various tools and return the updated table in its second output port. The first output port contains the conversation, where you can see the reasoning behind the tool choices and the results generated by each tool.

List the tools in the tools folder

Table of product feedback

Note: the data is not sent to the LLM provider. The agent can only access table specs (table description, column names and column types) and send it to the tools, without accessing its content.

Authenticate and connect to a Databricks Workspace

Connect to LLMs served on Databricks and integrate them to KNIME workflows


Connect to yourDatabricks workspace
Databricks Workspace Connector
Tools
List Files/Folders
Product feedback
Table Creator
Databricks LLM Selector
Workflow to Tool
First outpu: conversationSecond output: feedback with discount code
Agent Prompter

Nodes

Extensions

Links