Skip to content

config ¤

This module manages loading and validating configuration settings for the application from TOML files.

It also provides utility functions for importing classes and instantiating objects based on their dotted paths.

settings module-attribute ¤

settings: LazySettings = Dynaconf(
    envvar_prefix="BOOKACLE",
    root_path=ROOT_PATH,
    settings_files=["settings.toml", "prompts.toml"],
)

Settings for the application.

The following validators are registered on the settings object.

from dynaconf import Validator

[
Validator(
    "custom_loaders_dir",
    "clustering_func",
    "stream_output",
    "embedding_model.model_class",
    "embedding_model.model_arguments",
    "summarization_model.model_class",
    "summarization_model.model_arguments",
    "document_splitter.splitter_class",
    "document_splitter.splitter_arguments",
    "retriever.retriever_class",
    "retriever.retriever_arguments",
    "clustering_backend.backend_class",
    "clustering_backend.backend_arguments",
    "qa_model.model_class",
    "qa_model.model_arguments",
    must_exist=True,
),
Validator(
    "embedding_model",
    cast=lambda value: cast_class_path_to_instance(
        value["model_class"], value["model_arguments"]
    ),
),
Validator(
    "summarization_model",
    cast=lambda value: cast_class_path_to_instance(
        class_path=value["model_class"],
        arguments=value["model_arguments"],
    ),
),
Validator("document_splitter", cast=cast_document_splitter),
Validator("retriever_config", cast=cast_retriever_config),
Validator(
    "retriever",
    cast=lambda value: cast_class_path_to_instance(
        class_path=value["retriever_class"],
        arguments=value["retriever_arguments"],
    ),
),
Validator("clustering_func", cast=import_attribute_from_module),
Validator(
    "clustering_backend",
    cast=lambda value: cast_class_path_to_instance(
        class_path=value["backend_class"],
        arguments=value["backend_arguments"],
    ),
),
Validator(
    "tree_builder_config",
    cast=lambda value: cast_class_path_to_instance(
        class_path=value["config_class"], arguments=value["config_arguments"]
    ),
),
Validator(
    "tree_builder",
    cast=lambda value: cast_class_path_to_instance(
        class_path=value["builder_class"],
        arguments=value["builder_arguments"],
    ),
),
Validator(
    "qa_model",
    cast=lambda value: cast_class_path_to_instance(
        class_path=value["model_class"], arguments=value["model_arguments"]
    ),
),
Validator("chunk_size", default=None),
Validator("chunk_overlap", default=None),
]

Note that the validators are not applied automatically since there is overhead (for example, in casting the embedding models and summarization models).

To apply the validators, call settings.validators.validate().

For more details, see the Dynaconf documentation on validators: https://www.dynaconf.com/validation/.

The default settings are loaded from the settings.toml and prompts.toml.

  • settings.toml contains the main configuration settings.
  • prompts.toml contains the prompts for the user interface.

The settings can be accessed as a dictionary or as attributes. For example:

from bookacle.conf import settings
# Validate the settings
settings.validators.validate()
# Access as attribute
print(f"Default embedding model: {settings.EMBEDDING_MODEL}")
# Access as dictionary
print(f"Default QA model: {settings['qa_model']}")

Default embedding model: SentenceTransformerEmbeddingModel(model_name='paraphrase-albert-small-v2', use_gpu=False)
Default QA model: OllamaQAModel(model_name='qwen2:0.5b')

For more details on managing settings, see [Configuration][].

import_attribute_from_module ¤

import_attribute_from_module(dotted_path: str) -> Type[Any]

Imports an attribute (class or function) from a dotted module path.

Parameters:

  • dotted_path (str) –

    The full path to the attribute in the form of ‘module.submodule.ClassName’.

Returns:

  • Type[Any] –

    The imported attribute (e.g., a class or function).

Raises:

cast_class_path_to_instance ¤

cast_class_path_to_instance(
    class_path: str, arguments: dict[str, Any]
) -> object

Instantiates a class by its dotted path and passes the provided arguments.

Parameters:

  • class_path (str) –

    The full path to the class in the form ‘module.submodule.ClassName’.

  • arguments (dict[str, Any]) –

    A dictionary of arguments to pass to the class constructor.

Returns:

  • object –

    An instance of the class.

Raises:

cast_document_splitter ¤

cast_document_splitter(
    value: dict[str, Any]
) -> DocumentSplitterLike

Casts a dictionary containing document splitter configuration into an instance of DocumentSplitterLike.

Parameters:

  • value (dict[str, Any]) –

    Dictionary containing the class path of the document splitter and its arguments.

Returns:

Raises:

cast_retriever_config ¤

cast_retriever_config(
    value: dict[str, Any]
) -> RetrieverLike

Casts a dictionary containing retriever configuration into an instance of RetrieverLike.

Parameters:

  • value (dict[str, Any]) –

    Dictionary with the class path of the retriever and its arguments.

Returns:

Raises: