Types

_AvailFilters

Bases: TypedDict

Internal class for the available parameter set.

Source code in parameter_search/custom_types.py

class _AvailFilters(TypedDict):
    """Internal class for the available parameter set."""

    loader: list[str]
    chunking_config: list[str]
    embedding_model: list[str]
    vector_store: list[str]
    similarity_top_k: list[int]
    llm: list[str]
    mode: list[str]

GitDataFileConfig

Bases: TypedDict

Git data instance for a file.

Attributes:

Name	Type	Description
filename	str	The file (paper) to associate this github repository info with.
git_info	GitData	The github repository information for document ingestion.

Source code in parameter_search/custom_types.py

class GitDataFileConfig(TypedDict):
    """Git data instance for a file.

    Attributes
    ----------
    filename : str
        The file (paper) to associate this github repository info with.
    git_info : GitData
        The github repository information for document ingestion.
    """

    filename: str
    git_info: GitData

SearchSpace

Bases: TypedDict

Search space used for hyperparameter search.

Attributes:

Name	Type	Description
filenames	list[str]	The file (paper) name's to process.
loader	list[str]	The list of available data loaders to test.
chunking_config	list[str]	The chunking strategies to test.
embedding_model	list[str]	The embedding models to test.
vector_store	list[str]	The vector stores to test.
similarity_top_k	list[int]	The similarity top k values to test.
llm	list[str]	The LLMs to test.
git_data	Optional[list[GitDataFileConfig]]	The git data information.
other_docs	Optional[dict[str, list[str]]]	Any other documents to include (keys are the paper name to associate with).

Source code in parameter_search/custom_types.py

class SearchSpace(TypedDict):
    """Search space used for hyperparameter search.

    Attributes
    ----------
    filenames : list[str]
        The file (paper) name's to process.
    loader : list[str]
        The list of available data loaders to test.
    chunking_config : list[str]
        The chunking strategies to test.
    embedding_model : list[str]
        The embedding models to test.
    vector_store : list[str]
        The vector stores to test.
    similarity_top_k : list[int]
        The similarity top k values to test.
    llm : list[str]
        The LLMs to test.
    git_data : Optional[list[GitDataFileConfig]]
        The git data information.
    other_docs : Optional[dict[str, list[str]]]
        Any other documents to include (keys are the paper name to
        associate with).
    """

    filenames: list[str]
    loader: list[str]
    chunking_config: list[str]
    embedding_model: list[str]
    vector_store: list[str]
    similarity_top_k: list[int]
    llm: list[str]
    git_data: Optional[list[GitDataFileConfig]]
    other_docs: Optional[dict[str, list[str]]]

create_git_data_file_config(filename, git_info)

Constructor for the GitDataFileConfig TypedDict.

Parameters:

Name	Type	Description	Default
filename	str	The file (paper) to associate this github repository info with.	required
git_info	GitData	The github repository information for document ingestion.	required

Returns:

Type	Description
GitDataFileConfig

Source code in parameter_search/custom_types.py

def create_git_data_file_config(filename: str, git_info: GitData) -> GitDataFileConfig:
    """Constructor for the GitDataFileConfig TypedDict.

    Parameters
    ----------
    filename : str
        The file (paper) to associate this github repository info with.
    git_info : GitData
        The github repository information for document ingestion.

    Returns
    -------
    GitDataFileConfig
    """
    return_data: GitDataFileConfig = {"filename": filename, "git_info": git_info}
    return return_data

init_search_space(filenames=None, loader=None, chunking_config=None, embedding_model=None, vector_store=None, similarity_top_k=None, llm=None, git_data=None, other_docs=None)

Creates a search space instance.

Parameters:

Name	Type	Description	Default
filenames	list[str] | str | None	The filenames to test over for the search space (if None, defaults to all the filenames in the bcorag/test_papers/ directory). Note, many files can increase run time significantly as a full parameter search will be executed on each paper sequentially.	None
loader	list[str] | str | None	The data loaders for the search space (if None, defaults to the full list as defined in the conf.json list).	None
chunking_config	list[str] | str | or None	The chunking strategies for the search space (if None, defaults to the full list as defined in the conf.json list).	None
embedding_model	list[str] | str | or None	The embedding model for the search space (if None, defaults to the full list as defined in the conf.json list).	None
vector_store	list[str] | str | or None	The vector store for the search space (if None, defaults to the full list as defined in the conf.json list).	None
similarity_top_k	list[int] | int | or None	The similarity top k for the search space (if None, defaults to the full list as defined in the conf.json list).	None
llm	list[str] | str | or None	The llm for the search space (if None, defaults to the full list as defined in the conf.json list).	None
git_data	list[GitDataFileConfig] | GitDataFileConfig | None	The git data for each file (if None, assumes no git data for any files).	None
other_docs	dict[str, list[str]] | None	The other documents to include (if None, assumes no other docs for any files).	None

Returns:

Type	Description
SearchSpace	The search space grid.

Source code in parameter_search/custom_types.py

def init_search_space(
    filenames: Optional[list[str] | str] = None,
    loader: Optional[list[str] | str] = None,
    chunking_config: Optional[list[str] | str] = None,
    embedding_model: Optional[list[str] | str] = None,
    vector_store: Optional[list[str] | str] = None,
    similarity_top_k: Optional[list[int] | int] = None,
    llm: Optional[list[str] | str] = None,
    git_data: Optional[list[GitDataFileConfig]] = None,
    other_docs: Optional[dict[str, list[str]]] = None,
) -> SearchSpace:
    """Creates a search space instance.

    Parameters
    ----------
    filenames : list[str] | str | None, optional
        The filenames to test over for the search space (if `None`,
        defaults to all the filenames in the `bcorag/test_papers/`
        directory). Note, many files can increase run time
        significantly as a full parameter search will be executed
        on each paper sequentially.
    loader : list[str] | str | None, optional
        The data loaders for the search space (if `None`, defaults to
        the full list as defined in the `conf.json` list).
    chunking_config : list[str] | str | or None, optional
        The chunking strategies for the search space (if `None`, defaults
        to the full list as defined in the `conf.json` list).
    embedding_model : list[str] | str | or None, optional
        The embedding model for the search space (if `None`, defaults
        to the full list as defined in the `conf.json` list).
    vector_store : list[str] | str | or None, optional
        The vector store for the search space (if `None`, defaults
        to the full list as defined in the `conf.json` list).
    similarity_top_k : list[int] | int | or None, optional
        The similarity top k for the search space (if `None`, defaults
        to the full list as defined in the `conf.json` list).
    llm : list[str] | str | or None, optional
        The llm for the search space (if `None`, defaults
        to the full list as defined in the `conf.json` list).
    git_data : list[GitDataFileConfig] | GitDataFileConfig | None, optional
        The git data for each file (if `None`, assumes no git data for
        any files).
    other_docs : dict[str, list[str]] | None, optional
        The other documents to include (if `None`, assumes no other docs
        for any files).

    Returns
    -------
    SearchSpace
        The search space grid.
    """

    def _validate_options(
        option: OptionKey, option_list: list[str] | list[int]
    ) -> bool:
        if not set(option_list) <= set(_avail_options[option]):
            return False
        return True

    match filenames:
        case list():
            filenames_space: list[str] = filenames
        case str():
            filenames_space = [filenames]
        case None:
            filenames_space = get_file_list("./bcorag/test_papers", "*.pdf")
        case _:
            graceful_exit(1, "Invalid type for filenames")
    for file in filenames_space:
        if not os.path.isfile(file):
            graceful_exit(1, f"Invalid file `{file}`")

    match loader:
        case list():
            loader_space: list[str] = loader
            if not _validate_options("loader", loader_space):
                graceful_exit(1, "Invalid or undefined loader in search space")
        case str():
            loader_space = [loader]
            if not _validate_options("loader", loader_space):
                graceful_exit(1, "Invalid or undefined loader in search space")
        case None:
            loader_space = _avail_options["loader"]
        case _:
            graceful_exit(1, "Invalid type specified for loader")

    match chunking_config:
        case list():
            chunking_space: list[str] = chunking_config
            if not _validate_options("chunking_config", chunking_space):
                graceful_exit(
                    1, "Invalid or undefined chunking strategy in search space"
                )
        case str():
            chunking_space = [chunking_config]
            if not _validate_options("chunking_config", chunking_space):
                graceful_exit(
                    1, "Invalid or undefined chunking strategy in search space"
                )
        case None:
            chunking_space = _avail_options["chunking_config"]
        case _:
            graceful_exit(1, "Invalid type specified for chunking_config")

    match embedding_model:
        case list():
            embedding_model_space: list[str] = embedding_model
            if not _validate_options("embedding_model", embedding_model_space):
                graceful_exit(1, "Invalid or undefined embedding model in search space")
        case str():
            embedding_model_space = [embedding_model]
            if not _validate_options("embedding_model", embedding_model_space):
                graceful_exit(1, "Invalid or undefined embedding model in search space")
        case None:
            embedding_model_space = _avail_options["embedding_model"]
        case _:
            graceful_exit(1, "Invalid type specified for embedding_model")

    match vector_store:
        case list():
            vector_store_space: list[str] = vector_store
            if not _validate_options("vector_store", vector_store_space):
                graceful_exit(1, "Invalid or undefined vector store in search space")
        case str():
            vector_store_space = [vector_store]
            if not _validate_options("vector_store", vector_store_space):
                graceful_exit(1, "Invalid or undefined vector store in search space")
        case None:
            vector_store_space = _avail_options["vector_store"]
        case _:
            graceful_exit(1, "Invalid type specified for vector_store")

    match similarity_top_k:
        case list():
            similarity_top_k_space: list[int] = similarity_top_k
        case int():
            similarity_top_k_space = [similarity_top_k]
        case None:
            similarity_top_k_space = _avail_options["similarity_top_k"]
        case _:
            graceful_exit(1, "Invalid type for similarity top k")

    match llm:
        case list():
            llm_space: list[str] = llm
            if not _validate_options("llm", llm_space):
                graceful_exit(1, "Invalid or undefined llm in search space")
        case str():
            llm_space = [llm]
            if not _validate_options("llm", llm_space):
                graceful_exit(1, "Invalid or undefined llm in search space")
        case None:
            llm_space = _avail_options["llm"]
        case _:
            graceful_exit(1, "Invalid type for llm")

    match git_data:
        case list():
            git_data_space: list[GitDataFileConfig] | None = git_data
        case None:
            git_data_space = None

    match other_docs:
        case None:
            other_docs_space = None
        case _:
            if all(isinstance(item, list) for item in list(other_docs.values())):
                other_docs_space = other_docs
            else:
                graceful_exit(1, "Invalid other docs search space foramt.")

    return_data: SearchSpace = {
        "filenames": filenames_space,
        "loader": loader_space,
        "chunking_config": chunking_space,
        "embedding_model": embedding_model_space,
        "vector_store": vector_store_space,
        "similarity_top_k": similarity_top_k_space,
        "llm": llm_space,
        "git_data": git_data_space,
        "other_docs": other_docs_space,
    }

    return return_data