Model configuration

Look up a field below for what it accepts and the limit enforced when you save or run it. For which models are available in your workspace, see AI Providers in Admin & Settings.

You set these from the Playground’s editor header or the SDK.

Generation parameters

Field	Type	Valid range
temperature	number	0.0 to 2.0
frequency_penalty	number	-2.0 to 2.0
presence_penalty	number	-2.0 to 2.0
top_p	number	0.0 to 1.0
max_tokens	integer	1 to 65536

A value outside its range is rejected before the model runs. These ranges are what the API and SDK enforce.

Output and tool settings

Field	Type	Valid values	Default
output_format	string	array, string, number, object, audio, image	string
tool_choice	string	auto, required, or unset	Unset
tools	array	A list of tool IDs	Unset
response_format	object	A JSON schema object, a string, or the ID of a saved response schema	Unset

Prompt structure limits

Item	Limit
Messages	An ordered list of {role, content} objects. role must be system, user, or assistant; content must be a string. Any other role is rejected
Prompt template name	Up to 2000 characters
Base template name	Up to 255 characters
Folder name	Up to 255 characters
Version name	Must match v followed by digits, for example v1, v12. Anything else is rejected
Version comparison	Up to 3 versions at once. A fourth is rejected

Run behavior limits

Limits on how many calls run at once from a Run Prompt column across a dataset.

Item	Limit
Concurrent calls in a dataset run	1 to 10, default 5. A value over 10 is rejected

Response schemas

A response schema is a saved shape that response_format can point at by ID instead of an inline JSON schema object.

Field	What it means
name	Unique within your organization and workspace
schema_type	json or yaml

Keep exploring