This is a type used in Stencila Schema for selecting and configuring generative AI models.

It exists to make model choice and inference settings part of the document model, so prompts, chats, and instructions can be reproduced and inspected rather than depending on hidden runtime configuration. This supports transparent and portable AI-assisted workflows.

Key properties include model identity, provider-specific settings, and inference controls such as temperature and token limits.

Properties

The ModelParameters type has these properties:

Name	Description	Type	Inherited from
`modelIds`	The ids of the models to select.	`String`*	-
`replicates`	The number of replicate inferences to run per model id.	`UnsignedInteger`	-
`qualityWeight`	The relative weighting given to model quality (0-100).	`UnsignedInteger`	-
`costWeight`	The relative weighting given to model cost (0-100).	`UnsignedInteger`	-
`speedWeight`	The relative weighting given to model speed (0-100).	`UnsignedInteger`	-
`minimumScore`	The minimum score for models to be selected (0-100).	`UnsignedInteger`	-
`temperature`	The temperature option for model inference (0-100).	`UnsignedInteger`	-
`randomSeed`	The random seed used for the model (if possible)	`Integer`	-
`executeContent`	Automatically execute generated content.	`Boolean`	-
`executionBounds`	The environment in which code should be executed.	`ExecutionBounds`	-
`maximumRetries`	When executing content, the maximum number of retries.	`UnsignedInteger`	-
`id`	The identifier for this item.	`String`	`Entity`

The ModelParameters type is related to these types:

Parents: Entity

Children: none

Bindings

The ModelParameters type is represented in:

JSON-LD

JSON Schema

Python class ModelParameters

Rust struct ModelParameters

TypeScript class ModelParameters

This documentation was generated from ModelParameters.yaml by docs_types.rs.

Properties

Related

Bindings