SELECT.GPT

  • Updated

It selects content from a neutral table format and generates text based on a prompt. For each row in the table, it uses the prompt to create text and stores the result in the corresponding cell. If the input data table contains a file (as binary id), you can specify in the command that this file (such as an image or a pdf) should be used along with the prompt so that the GPT analyses and gives a response based on the file.

Properties

FromName of the Input Data Source. The input source must contain the neutral table format.
ToName of the Output Data Source.
The output-source will always be a neutral table format with an extra column for the generated text.
ScriptThe Prompt for the GPT. Use {ColumnName} to reference a column in the table 

Parameters

@ResultColumn Specifies the target column where the result will be written. If the column already exists in the input data table, its values will be overwritten.
@MaxRowLimit (optional)Defines the maximum number of rows GPT can process to prevent excessive or unintended calls. The default is 100.
@Temperature (optional)Sets the sampling temperature for the GPT call, controlling output randomness. Higher values (e.g., 1.0) generate more diverse results, while lower values (e.g., 0.2) produce more deterministic output. If not set, the model’s default temperature is used.
@MaxTokens (optional)Sets the maximum number of tokens the GPT model can generate in a response, limiting the output length. If not specified, the model’s default value is used.
@Model (optional) Specifies which model to use. The default is gpt-4-turbo.
@ApiKey (optional) Set the API key. If not specified, the key from the AI settings will be used by default.

image-20250812-125525.png

Example
Image

FromToCommandScript
.ItemSELECT.PERFION<Query>
<Select languages="EN" maxCount='5'>
<Feature id="Image" />
</Select>
<From id="Product" />
<Where>
<Clause id='Image' operator='HAS' />
</Where>
</Query>
ItemItemEnrichedSELECT.GPTWrite a 1 paragraph description based on this image {Image}

SELECT.PERFION retrieves products that have an image, including the Image column as a binary id. SELECT.GPT then references {Image} in the Script, since that column holds a file rather than text, the image itself is sent to the model, and the generated description is written to Description_EN via @ResultColumn.

Was this article helpful?

0 out of 0 found this helpful

Comments

0 comments

Please sign in to leave a comment.