Config

Configuration schemas for LLMRouter backends.

KNNConfig

Bases: BaseModel

Configuration for the KNN classifier.

Attributes:

Name	Type	Description
num_classes	int	Number of output classes (models to route to).

MLPConfig

Bases: BaseModel

Configuration for the MLP classifier.

Attributes:

Name	Type	Description
num_classes	int	Number of output classes (models to route to).
input_dim	int	Dimensionality of the input embeddings.
idx_to_model	dict[int, str]	Mapping from predicted class index to model name.
hidden_layer_sizes	list[int]	Sizes of hidden layers.
activation	str	Activation function name. Defaults to "relu".
hyperparameters	dict[str, Any]	Additional hyperparameters. Defaults to {}.

build_hyperparameter(overrides=None)

Build final hyperparameter dict with safe merging.

Merges core config params, stored hyperparameters, and runtime overrides with protection against overriding reserved keys.

Parameters:

Name	Type	Description	Default
overrides	dict[str, Any]	Runtime hyperparameter overrides. Cannot override reserved keys (hidden_layer_sizes, activation). Defaults to None.	None

Returns:

Type	Description
dict[str, Any]	dict[str, Any]: Final merged hyperparameter dictionary.

Raises:

Type	Description
ValueError	If overrides attempt to modify reserved keys.

SVMConfig

Bases: BaseModel

Configuration for the SVM classifier.

Attributes:

Name	Type	Description
num_classes	int	Number of output classes (models to route to).