LLMViewSelector

dbally.view_selection.LLMViewSelector

LLMViewSelector(llm: LLM, prompt_template: Optional[PromptTemplate[ViewSelectionPromptFormat]] = None)

Bases: ViewSelector

The LLMViewSelector utilises LLMs to select the most suitable view to answer the user question.

Its primary function is to determine the optimal view that can effectively be used to answer a user's question.

The method used to select the most relevant view is self.select_view. It formats views using view.name: view.description format and then calls LLM Client, ultimately returning the name of the most suitable view.

Constructs a new LLMViewSelector instance.

PARAMETER	DESCRIPTION
llm	LLM used to generate IQL TYPE: LLM
prompt_template	template for the prompt used for the view selection TYPE: Optional[PromptTemplate[ViewSelectionPromptFormat]] DEFAULT: None

Source code in src/dbally/view_selection/llm_view_selector.py

def __init__(self, llm: LLM, prompt_template: Optional[PromptTemplate[ViewSelectionPromptFormat]] = None) -> None:
    """
    Constructs a new LLMViewSelector instance.

    Args:
        llm: LLM used to generate IQL
        prompt_template: template for the prompt used for the view selection
    """
    self._llm = llm
    self._prompt_template = prompt_template or VIEW_SELECTION_TEMPLATE

select_view async

select_view(question: str, views: Dict[str, str], event_tracker: EventTracker, llm_options: Optional[LLMOptions] = None) -> str

Based on user question and list of available views select the most relevant one by prompting LLM.

PARAMETER	DESCRIPTION
question	user question asked in the natural language e.g "Do we have any data scientists?" TYPE: str
views	dictionary of available view names with corresponding descriptions. TYPE: Dict[str, str]
event_tracker	event tracker used to audit the selection process. TYPE: EventTracker
llm_options	options to use for the LLM client. TYPE: Optional[LLMOptions] DEFAULT: None

RETURNS	DESCRIPTION
str	The most relevant view name.

RAISES	DESCRIPTION
LLMError	If LLM text generation fails.

Source code in src/dbally/view_selection/llm_view_selector.py

async def select_view(
    self,
    question: str,
    views: Dict[str, str],
    event_tracker: EventTracker,
    llm_options: Optional[LLMOptions] = None,
) -> str:
    """
    Based on user question and list of available views select the most relevant one by prompting LLM.

    Args:
        question: user question asked in the natural language e.g "Do we have any data scientists?"
        views: dictionary of available view names with corresponding descriptions.
        event_tracker: event tracker used to audit the selection process.
        llm_options: options to use for the LLM client.

    Returns:
        The most relevant view name.

    Raises:
        LLMError: If LLM text generation fails.
    """
    prompt_format = ViewSelectionPromptFormat(question=question, views=views)
    formatted_prompt = self._prompt_template.format_prompt(prompt_format)

    llm_response = await self._llm.generate_text(
        prompt=formatted_prompt,
        event_tracker=event_tracker,
        options=llm_options,
    )
    selected_view = self._prompt_template.response_parser(llm_response)
    return selected_view