类class agentscope.embedding

EmbeddingModelCard

class EmbeddingModelCard(BaseModel)

源码Source agentscope/embedding/_embedding_model_card.py:12–179

向量化模型卡片，与模型服务通信并返回生成或向量化结果。位于 agentscope.embedding 包。核心成员包括 from_yaml。

Mirrors :class:~agentscope.model.ModelCard but tailored for embedding models. Uses input_types / output_types to describe model capabilities, and parameter_schema (built from the embedding class's Parameters + YAML parameter_overrides) to tell the frontend which knobs the user can adjust.

The output type application/x-embedding indicates that the model produces dense vector embeddings.

属性摘要Attribute Summary

属性Attribute	类型Type	说明Description	值Value
`type`	Literal['embedding_model']	类型字段。	`'embedding_model'`
`name`	str	名称字段。	`Field(description='The model name used in API calls.')`
`label`	str	标签字段。	`Field(description='Human-readable label for the frontend.')`
`status`	Literal['active', 'deprecated', 'sunset']	模型生命周期状态。	`Field(default='active', description='The model lifecycle status.')`
`input_types`	list[str]	输入类型字段。	`Field(default=['text/plain'], description='Supported input media types.')`
`output_types`	list[str]	输出类型字段。	`Field(default=['application/x-embedding'], description='Supported output media types.')`
`dimensions`	int	用于此模型的默认输出向量维度。	`Field(..., description='Default output vector dimensions for this model.', gt=0)`
`supported_dimensions`	list[int] \| None	`supported_dimensions` 字段。	Field(default=None, description='If set, the only dimensions this model can produce. ``None`` means dimensions are fixed at :attr:`dimensions` and cannot be overridden.')
`context_size`	int \| None	上下文大小字段。	`Field(default=None, description='Maximum input length (in tokens) per request.', gt=0)`
`parameter_schema`	dict	参数模式字段。	`Field(default_factory=dict, description='JSON Schema for user-configurable parameters (built from the Parameters class + YAML overrides).')`
`parameter_overrides`	dict[str, dict]	`parameter_overrides` 字段。	`Field(default_factory=dict, description='Raw parameter overrides from the YAML file.')`

方法摘要Method Summary

方法Method	说明Description
`def from_yaml(yaml_path: str, parameter_class: Type[BaseModel]) -> Self`	执行 `from_yaml` 方法。 Load an embedding model card from a YAML file.

属性详情Attribute Detail

type

Literal['embedding_model'] = 'embedding_model'

类型字段。声明为 Literal['embedding_model']。

The type field.

源码Source agentscope/embedding/_embedding_model_card.py:25

name

str = Field(description='The model name used in API calls.')

名称字段。声明为 str。

The model name used in API calls.

源码Source agentscope/embedding/_embedding_model_card.py:28

label

str = Field(description='Human-readable label for the frontend.')

标签字段。声明为 str。

Human-readable label for the frontend.

源码Source agentscope/embedding/_embedding_model_card.py:31

status

Literal['active', 'deprecated', 'sunset'] = Field(default='active', description='The model lifecycle status.')

模型生命周期状态。

The model lifecycle status.

源码Source agentscope/embedding/_embedding_model_card.py:34

input_types

list[str] = Field(default=['text/plain'], description='Supported input media types.')

输入类型字段。声明为 list[str]。

Supported input media types.

源码Source agentscope/embedding/_embedding_model_card.py:40

output_types

list[str] = Field(default=['application/x-embedding'], description='Supported output media types.')

输出类型字段。声明为 list[str]。

Supported output media types.

源码Source agentscope/embedding/_embedding_model_card.py:47

dimensions

int = Field(..., description='Default output vector dimensions for this model.', gt=0)

用于此模型的默认输出向量维度。

Default output vector dimensions for this model.

源码Source agentscope/embedding/_embedding_model_card.py:53

supported_dimensions

list[int] | None = Field(default=None, description='If set, the only dimensions this model can produce. ``None`` means dimensions are fixed at :attr:`dimensions` and cannot be overridden.')

supported_dimensions 字段。声明为 list[int] | None。

If set, the only dimensions this model can produce.

源码Source agentscope/embedding/_embedding_model_card.py:66

context_size

int | None = Field(default=None, description='Maximum input length (in tokens) per request.', gt=0)

上下文大小字段。声明为 int | None。

Maximum input length (in tokens) per request.

源码Source agentscope/embedding/_embedding_model_card.py:81

parameter_schema

dict = Field(default_factory=dict, description='JSON Schema for user-configurable parameters (built from the Parameters class + YAML overrides).')

参数模式字段。声明为 dict。

JSON Schema for user-configurable parameters (built from the Parameters class + YAML overrides).

源码Source agentscope/embedding/_embedding_model_card.py:88

parameter_overrides

dict[str, dict] = Field(default_factory=dict, description='Raw parameter overrides from the YAML file.')

parameter_overrides 字段。声明为 dict[str, dict]。

Raw parameter overrides from the YAML file.

源码Source agentscope/embedding/_embedding_model_card.py:99

方法详情Method Detail

from_yaml

def from_yaml(yaml_path: str, parameter_class: Type[BaseModel]) -> Self

执行 from_yaml 方法。

Merges the base parameter_class JSON Schema with parameter_overrides from the YAML — identical to the approach used by :meth:~agentscope.model.ModelCard.from_yaml.

参数Parameters

名称Name	类型Type	说明Description
`yaml_path`	str	yaml 路径（类型 `str`）。Path to the YAML file.
`parameter_class`	Type[BaseModel]	参数类（类型 `Type[BaseModel]`）。The `Parameters` class from the embedding model subclass.

返回Returns Self

返回自身。`EmbeddingModelCard`: The loaded model card.

源码Source agentscope/embedding/_embedding_model_card.py:106–179