Overview

Schema package for modality transformer operations.

This package re-exports the data models used by modality transformers.

TransformResult -- Result of a modality transformation operation.
ConverterConfig -- Configuration for a modality converter within a transformer.
RouterConfig -- Configuration for the routing layer of a transformer.

`ConverterConfig`

Bases: BaseModel

Configuration for building a single modality converter instance.

Attributes:

Name	Type	Description
`source_modality`	`Modality`	Source modality for conversion. Defaults to Modality.IMAGE.
`target_modality`	`Modality`	Target modality for conversion. Defaults to Modality.TEXT.
`task_type`	`ModalityConverterTask`	Converter task type. Defaults to ModalityConverterTask.AUTO.
`approach_type`	`ModalityConverterApproach \| None`	Converter approach type. Defaults to None.
`preset`	`str \| None`	Preset name for the converter. Required for ModalityConverterBuildStrategy.PRESET. Defaults to None.
`strategy`	`ModalityConverterBuildStrategy \| None`	Build strategy for the converter. When None, the strategy is auto-determined from the provided fields (preset → PRESET, lmrp_config → LMRP, otherwise → KWARGS). Defaults to None.
`lmrp_config`	`dict[str, Any] \| None`	Configuration for custom LMRP. Required for ModalityConverterBuildStrategy.LMRP. Follows build_lm_request_processor signature. Defaults to None.

Pydantic model configuration.

This class defines the Pydantic model configuration for the ConverterConfig model.

Attributes:

Name	Type	Description
`extra`	`str`	Allow extra fields.

Bases: BaseModel

Configuration for building a router instance.

Attributes:

Name	Type	Description
`modality`	`Modality`	Modality type. Defaults to Modality.IMAGE.
`preset`	`ModalityTransformerRouterPreset`	Router preset type. Defaults to MULTIMODAL.
`model_id`	`str`	Embedding model ID for multimodal preset. Defaults to DEFAULT_ROUTER_EMBEDDING_MODEL.
`model_params`	`dict[str, Any]`	Additional parameters for the embedding model.
`es_url`	`str \| None`	Elasticsearch URL for vector store. Defaults to None.
`es_index_name`	`str \| None`	Elasticsearch index name. Defaults to None.
`route_mapping`	`dict[str, str]`	Mapping from route labels (as returned by the router) to converter names (keys in converter_config). Defaults to an empty dict.
`batch_size`	`int`	Batch size for indexing utterances. Defaults to DEFAULT_BATCH_SIZE.

Bases: BaseModel

Base class for all transformation operation results.

This class provides the foundation for structured results from any transformation operation, including: - Image-to-Caption - Image-to-Mermaid

Attributes:

Name	Type	Description
`result`	`str \| bytes`	The transformed result.
`tag`	`str`	The tag of the conversion result.
`route`	`str`	The selected route from the router.
`captions`	`str \| list[str]`	The captions generated during conversion.
`text_repr`	`str \| list[str] \| dict[str, Any]`	The text representation of the result.
`metadata`	`dict[str, Any] \| BaseModel`	Additional metadata generated from the transformation.