Multi step

MultiStepQueryEngine

Bases: BaseQueryEngine

Multi-step query engine.

This query engine can operate over an existing base query engine, along with the multi-step query transform.

Parameters:

Name	Type	Description	Default
query_engine	BaseQueryEngine	A BaseQueryEngine object.	required
query_transform	StepDecomposeQueryTransform	A StepDecomposeQueryTransform object.	required
response_synthesizer	Optional[BaseSynthesizer]	A BaseSynthesizer object.	None
num_steps	Optional[int]	Number of steps to run the multi-step query.	3
early_stopping	bool	Whether to stop early if the stop function returns True.	True
index_summary	str	A string summary of the index.	'None'
stop_fn	Optional[Callable[[Dict], bool]]	A stop function that takes in a dictionary of information and returns a boolean.	None

Source code in llama-index-core/llama_index/core/query_engine/multistep_query_engine.py

class MultiStepQueryEngine(BaseQueryEngine):
    """Multi-step query engine.

    This query engine can operate over an existing base query engine,
    along with the multi-step query transform.

    Args:
        query_engine (BaseQueryEngine): A BaseQueryEngine object.
        query_transform (StepDecomposeQueryTransform): A StepDecomposeQueryTransform
            object.
        response_synthesizer (Optional[BaseSynthesizer]): A BaseSynthesizer
            object.
        num_steps (Optional[int]): Number of steps to run the multi-step query.
        early_stopping (bool): Whether to stop early if the stop function returns True.
        index_summary (str): A string summary of the index.
        stop_fn (Optional[Callable[[Dict], bool]]): A stop function that takes in a
            dictionary of information and returns a boolean.

    """

    def __init__(
        self,
        query_engine: BaseQueryEngine,
        query_transform: StepDecomposeQueryTransform,
        response_synthesizer: Optional[BaseSynthesizer] = None,
        num_steps: Optional[int] = 3,
        early_stopping: bool = True,
        index_summary: str = "None",
        stop_fn: Optional[Callable[[Dict], bool]] = None,
    ) -> None:
        self._query_engine = query_engine
        self._query_transform = query_transform
        self._response_synthesizer = response_synthesizer or get_response_synthesizer(
            callback_manager=self._query_engine.callback_manager
        )

        self._index_summary = index_summary
        self._num_steps = num_steps
        self._early_stopping = early_stopping
        # TODO: make interface to stop function better
        self._stop_fn = stop_fn or default_stop_fn
        # num_steps must be provided if early_stopping is False
        if not self._early_stopping and self._num_steps is None:
            raise ValueError("Must specify num_steps if early_stopping is False.")

        callback_manager = self._query_engine.callback_manager
        super().__init__(callback_manager)

    def _get_prompt_modules(self) -> PromptMixinType:
        """Get prompt sub-modules."""
        return {
            "response_synthesizer": self._response_synthesizer,
            "query_transform": self._query_transform,
        }

    def _query(self, query_bundle: QueryBundle) -> RESPONSE_TYPE:
        with self.callback_manager.event(
            CBEventType.QUERY, payload={EventPayload.QUERY_STR: query_bundle.query_str}
        ) as query_event:
            nodes, source_nodes, metadata = self._query_multistep(query_bundle)

            final_response = self._response_synthesizer.synthesize(
                query=query_bundle,
                nodes=nodes,
                additional_source_nodes=source_nodes,
            )
            final_response.metadata = metadata

            query_event.on_end(payload={EventPayload.RESPONSE: final_response})

        return final_response

    async def _aquery(self, query_bundle: QueryBundle) -> RESPONSE_TYPE:
        with self.callback_manager.event(
            CBEventType.QUERY, payload={EventPayload.QUERY_STR: query_bundle.query_str}
        ) as query_event:
            nodes, source_nodes, metadata = self._query_multistep(query_bundle)

            final_response = await self._response_synthesizer.asynthesize(
                query=query_bundle,
                nodes=nodes,
                additional_source_nodes=source_nodes,
            )
            final_response.metadata = metadata

            query_event.on_end(payload={EventPayload.RESPONSE: final_response})

        return final_response

    def _combine_queries(
        self, query_bundle: QueryBundle, prev_reasoning: str
    ) -> QueryBundle:
        """Combine queries."""
        transform_metadata = {
            "prev_reasoning": prev_reasoning,
            "index_summary": self._index_summary,
        }
        return self._query_transform(query_bundle, metadata=transform_metadata)

    def _query_multistep(
        self, query_bundle: QueryBundle
    ) -> Tuple[List[NodeWithScore], List[NodeWithScore], Dict[str, Any]]:
        """Run query combiner."""
        prev_reasoning = ""
        cur_response = None
        should_stop = False
        cur_steps = 0

        # use response
        final_response_metadata: Dict[str, Any] = {"sub_qa": []}

        text_chunks = []
        source_nodes = []
        while not should_stop:
            if self._num_steps is not None and cur_steps >= self._num_steps:
                should_stop = True
                break
            elif should_stop:
                break

            updated_query_bundle = self._combine_queries(query_bundle, prev_reasoning)

            # TODO: make stop logic better
            stop_dict = {"query_bundle": updated_query_bundle}
            if self._stop_fn(stop_dict):
                should_stop = True
                break

            cur_response = self._query_engine.query(updated_query_bundle)

            # append to response builder
            cur_qa_text = (
                f"\nQuestion: {updated_query_bundle.query_str}\n"
                f"Answer: {cur_response!s}"
            )
            text_chunks.append(cur_qa_text)
            for source_node in cur_response.source_nodes:
                source_nodes.append(source_node)
            # update metadata
            final_response_metadata["sub_qa"].append(
                (updated_query_bundle.query_str, cur_response)
            )

            prev_reasoning += (
                f"- {updated_query_bundle.query_str}\n" f"- {cur_response!s}\n"
            )
            cur_steps += 1

        nodes = [
            NodeWithScore(node=TextNode(text=text_chunk)) for text_chunk in text_chunks
        ]
        return nodes, source_nodes, final_response_metadata