AI Presets

info

The interface for managing ai command presets is available starting with Search Anywhere Framework 6.2.

Purpose of Presets

A preset is a named set of parameters for invoking a language model with the ai SAFL command.

A preset lets you avoid specifying the same command parameters in every query.

Preset parameters apply by default. An argument explicitly specified in a query takes precedence over the preset value. The preset itself remains unchanged.

Preset List

To manage presets, go to Main Menu - Settings - Module Settings - AI Presets.

Access Permissions

Cluster permissions for presets are described in AI Preset Permissions.

A preset can be read and edited by its owner and by users with roles that have the corresponding permissions for that preset.

A user with permission to edit a preset can also delete it.

The list displays Name, Model, Data Processing Mode, and System Prompt.

Use the search field to find a preset by name, model, or system prompt. To filter presets by tag, use the Tags list.

To create a preset, click Create Preset in the upper-right corner of the page.

To configure access permissions or tags, or to delete a preset, open the menu at the end of its row and select Permissions, Edit Tags, or Delete.

Important

When you delete a preset, the system does not check whether it is used in saved queries. Before deleting it, check queries that specify its name in the preset argument.

Creating and Editing a Preset

1. Click Create Preset in the list, or select the name of an existing preset to edit it
2. When creating a preset, enter its Preset Name
3. Fill in the required parameters
4. Click Save. To close the form without saving changes, click Cancel

The name is set when the preset is created. All other fields are optional: if a field is empty, the corresponding parameter is not set in the preset.

Preset fields correspond to the arguments of the ai SAFL command:

Field	Command Argument	Description
Preset Name	preset	Name used to select the preset in a query
Model Name	model	Name of the language model registered in the model registry
Data Processing Mode	mode	summary transforms all input rows into one result row; per_row enriches each input row with result fields
Temperature	temperature	Model response generation temperature
Timeout, ms	timeout	Provider response timeout in milliseconds
Response Token Limit	max_tokens	Maximum number of tokens in the model response
Prompt Token Limit	max_prompt_tokens	Maximum number of tokens in the prompt
Context Fields (input)	input	Input row fields passed to the model as context; * means all fields
Result Fields (output)	output	Names of the fields where processing results are written
System Prompt	system_message	General instruction for the model: its role, data processing rules, and response requirements
User Prompt	user_message	Specific task for the model to perform using the provided data

In Context Fields (input) and Result Fields (output), enter each item separately and press Enter. To remove an item, click the cross next to it.

If a preset field is empty and the argument is not specified in the query, the default value from the ai command's optional arguments table applies. The model and result fields are required: if neither the preset nor the query specifies model or output, the command fails.