LLM General

Add an LLM General node to your data flow. This node is used to send a natural language "prompt" to the default LLM to generate values for a new column, based on the values in existing columns.

Note: It is important to note that any of the Generative Artificial Intelligence (Gen-AI) nodes could potentially create a performance overhead, depending on the size of your model or the complexity of your prompt. You should, therefore, consider the inclusion of these nodes carefully, and perhaps configure your Sync Model settings with reference to your requirements. For more information, see Data Flow Ribbon.

Add an LLM General node

This task is part of a "data preparation" or "transformation" stage that lets you cleanse and improve your data before you write it to your target database.

Before you begin

You should have a Data Flow that includes at least one datasource and one or more Tables. You are going to generate some new values for a new column based on a natural language prompt.

If you do not have permission to add this type of node, you will not see the Gen-AI options in the Data Flow > Elements list.

Procedure

From the Data Flow page in Model Pro:

  1. Click Gen-AI in the left-hand menu and, from the sub-menu, drag-and-drop LLM General onto the canvas.
  2. Connect the Table to the LLM General node.

    3. Note: The connection is created by selecting the "end connector" (red arrow) on the Table node and then dragging your cursor to the "start connector" (blue box) on the LLM General node.

  3. Select LLM General and, from the Properties panel, set:
    • Select Columns: Use the multi-select tool to select the names of the Columns that you want your prompt to take into account. For example, if you are going to write a prompt that combines values from multiple columns, you need to select all those columns here.
    • New Column Name: Specify the name that you want to give to the column that will store your generated data. By default, the name LLM Result is supplied for the new column.
    • Prompt: Type the prompt that you want to use to generate the value to add to your new column. Note: The length of your prompt could cause you to reach a token limit if you are not careful.
  4. Optionally, if you don't want to use the default LLM for this action, select an alternative LLM from the LLM Provider fields.

    5. For more information about this option, see Using an alternative LLM Provider below.

  5. Click Preview (the two buttons are shown with green arrows in the following example) and check that your new column is created as you expect in the Preview panel.

    6. Tip: The preview creates a fixed number of rows (50 by default). You can increase the number of rows by changing the Preview Size option. For more information, see Datasource Preview.

  6. Click Save at any time to save your changes.

You can now continue building your Data Flow as usual.

Using an alternative LLM Provider

Typically your default LLM (as selected in the Admin Console's AI Settings page) is used for all generative activities in Pyramid. However, if you have some specialist task to perform and you know that one of your other LLMs performs that type of action better than the default, then you can use the LLM Provider options to select your preferred LLM just for this task:

To use an alternative LLM for the task represented by the selected Gen-AI node:

  1. Select the Alternative LLM Provider checkbox.
  2. Select the LLM Provider you want to use from the drop-down.

Note: The list of LLM Providers on your drop-down is the list of LLM Providers in the Admin Console's LLM Manager page.

What next?

You may either want to continue processing your data or add a Target to your Data Flow at this point. The Target is the database that you want to use to store your processed data.

Example: LLM General