Citation

CitationQueryEngine

Bases: BaseQueryEngine

Citation query engine.

Parameters:

Name	Type	Description	Default
retriever	BaseRetriever	A retriever object.	required
response_synthesizer	Optional[BaseSynthesizer]	A BaseSynthesizer object.	None
citation_chunk_size	int	Size of citation chunks, default=512. Useful for controlling granularity of sources.	DEFAULT_CITATION_CHUNK_SIZE
citation_chunk_overlap	int	Overlap of citation nodes, default=20.	DEFAULT_CITATION_CHUNK_OVERLAP
text_splitter	Optional[TextSplitter]	A text splitter for creating citation source nodes. Default is a SentenceSplitter.	None
callback_manager	Optional[CallbackManager]	A callback manager.	None
metadata_mode	MetadataMode	A MetadataMode object that controls how metadata is included in the citation prompt.	NONE

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

class CitationQueryEngine(BaseQueryEngine):
    """Citation query engine.

    Args:
        retriever (BaseRetriever): A retriever object.
        response_synthesizer (Optional[BaseSynthesizer]):
            A BaseSynthesizer object.
        citation_chunk_size (int):
            Size of citation chunks, default=512. Useful for controlling
            granularity of sources.
        citation_chunk_overlap (int): Overlap of citation nodes, default=20.
        text_splitter (Optional[TextSplitter]):
            A text splitter for creating citation source nodes. Default is
            a SentenceSplitter.
        callback_manager (Optional[CallbackManager]): A callback manager.
        metadata_mode (MetadataMode): A MetadataMode object that controls how
            metadata is included in the citation prompt.
    """

    def __init__(
        self,
        retriever: BaseRetriever,
        llm: Optional[LLM] = None,
        response_synthesizer: Optional[BaseSynthesizer] = None,
        citation_chunk_size: int = DEFAULT_CITATION_CHUNK_SIZE,
        citation_chunk_overlap: int = DEFAULT_CITATION_CHUNK_OVERLAP,
        text_splitter: Optional[TextSplitter] = None,
        node_postprocessors: Optional[List[BaseNodePostprocessor]] = None,
        callback_manager: Optional[CallbackManager] = None,
        metadata_mode: MetadataMode = MetadataMode.NONE,
    ) -> None:
        self.text_splitter = text_splitter or SentenceSplitter(
            chunk_size=citation_chunk_size, chunk_overlap=citation_chunk_overlap
        )
        self._retriever = retriever

        callback_manager = callback_manager or Settings.callback_manager
        llm = llm or Settings.llm

        self._response_synthesizer = response_synthesizer or get_response_synthesizer(
            llm=llm,
            callback_manager=callback_manager,
        )
        self._node_postprocessors = node_postprocessors or []
        self._metadata_mode = metadata_mode

        for node_postprocessor in self._node_postprocessors:
            node_postprocessor.callback_manager = callback_manager

        super().__init__(callback_manager=callback_manager)

    @classmethod
    def from_args(
        cls,
        index: BaseGPTIndex,
        llm: Optional[LLM] = None,
        response_synthesizer: Optional[BaseSynthesizer] = None,
        citation_chunk_size: int = DEFAULT_CITATION_CHUNK_SIZE,
        citation_chunk_overlap: int = DEFAULT_CITATION_CHUNK_OVERLAP,
        text_splitter: Optional[TextSplitter] = None,
        citation_qa_template: BasePromptTemplate = CITATION_QA_TEMPLATE,
        citation_refine_template: BasePromptTemplate = CITATION_REFINE_TEMPLATE,
        retriever: Optional[BaseRetriever] = None,
        node_postprocessors: Optional[List[BaseNodePostprocessor]] = None,
        # response synthesizer args
        response_mode: ResponseMode = ResponseMode.COMPACT,
        use_async: bool = False,
        streaming: bool = False,
        # class-specific args
        metadata_mode: MetadataMode = MetadataMode.NONE,
        **kwargs: Any,
    ) -> "CitationQueryEngine":
        """Initialize a CitationQueryEngine object.".

        Args:
            index: (BastGPTIndex): index to use for querying
            llm: (Optional[LLM]): LLM object to use for response generation.
            citation_chunk_size (int):
                Size of citation chunks, default=512. Useful for controlling
                granularity of sources.
            citation_chunk_overlap (int): Overlap of citation nodes, default=20.
            text_splitter (Optional[TextSplitter]):
                A text splitter for creating citation source nodes. Default is
                a SentenceSplitter.
            citation_qa_template (BasePromptTemplate): Template for initial citation QA
            citation_refine_template (BasePromptTemplate):
                Template for citation refinement.
            retriever (BaseRetriever): A retriever object.
            node_postprocessors (Optional[List[BaseNodePostprocessor]]): A list of
                node postprocessors.
            verbose (bool): Whether to print out debug info.
            response_mode (ResponseMode): A ResponseMode object.
            use_async (bool): Whether to use async.
            streaming (bool): Whether to use streaming.
            optimizer (Optional[BaseTokenUsageOptimizer]): A BaseTokenUsageOptimizer
                object.

        """
        retriever = retriever or index.as_retriever(**kwargs)

        response_synthesizer = response_synthesizer or get_response_synthesizer(
            llm=llm,
            text_qa_template=citation_qa_template,
            refine_template=citation_refine_template,
            response_mode=response_mode,
            use_async=use_async,
            streaming=streaming,
        )

        return cls(
            retriever=retriever,
            llm=llm,
            response_synthesizer=response_synthesizer,
            callback_manager=Settings.callback_manager,
            citation_chunk_size=citation_chunk_size,
            citation_chunk_overlap=citation_chunk_overlap,
            text_splitter=text_splitter,
            node_postprocessors=node_postprocessors,
            metadata_mode=metadata_mode,
        )

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

    def _create_citation_nodes(self, nodes: List[NodeWithScore]) -> List[NodeWithScore]:
        """Modify retrieved nodes to be granular sources."""
        new_nodes: List[NodeWithScore] = []
        for node in nodes:
            text_chunks = self.text_splitter.split_text(
                node.node.get_content(metadata_mode=self._metadata_mode)
            )

            for text_chunk in text_chunks:
                text = f"Source {len(new_nodes) + 1}:\n{text_chunk}\n"

                new_node = NodeWithScore(
                    node=TextNode.model_validate(node.node), score=node.score
                )
                new_node.node.set_content(text)
                new_nodes.append(new_node)
        return new_nodes

    def retrieve(self, query_bundle: QueryBundle) -> List[NodeWithScore]:
        nodes = self._retriever.retrieve(query_bundle)

        for postprocessor in self._node_postprocessors:
            nodes = postprocessor.postprocess_nodes(nodes, query_bundle=query_bundle)

        return nodes

    async def aretrieve(self, query_bundle: QueryBundle) -> List[NodeWithScore]:
        nodes = await self._retriever.aretrieve(query_bundle)

        for postprocessor in self._node_postprocessors:
            nodes = postprocessor.postprocess_nodes(nodes, query_bundle=query_bundle)

        return nodes

    @property
    def retriever(self) -> BaseRetriever:
        """Get the retriever object."""
        return self._retriever

    def synthesize(
        self,
        query_bundle: QueryBundle,
        nodes: List[NodeWithScore],
        additional_source_nodes: Optional[Sequence[NodeWithScore]] = None,
    ) -> RESPONSE_TYPE:
        nodes = self._create_citation_nodes(nodes)
        return self._response_synthesizer.synthesize(
            query=query_bundle,
            nodes=nodes,
            additional_source_nodes=additional_source_nodes,
        )

    async def asynthesize(
        self,
        query_bundle: QueryBundle,
        nodes: List[NodeWithScore],
        additional_source_nodes: Optional[Sequence[NodeWithScore]] = None,
    ) -> RESPONSE_TYPE:
        nodes = self._create_citation_nodes(nodes)
        return await self._response_synthesizer.asynthesize(
            query=query_bundle,
            nodes=nodes,
            additional_source_nodes=additional_source_nodes,
        )

    def _query(self, query_bundle: QueryBundle) -> RESPONSE_TYPE:
        """Answer a query."""
        with self.callback_manager.event(
            CBEventType.QUERY, payload={EventPayload.QUERY_STR: query_bundle.query_str}
        ) as query_event:
            with self.callback_manager.event(
                CBEventType.RETRIEVE,
                payload={EventPayload.QUERY_STR: query_bundle.query_str},
            ) as retrieve_event:
                nodes = self.retrieve(query_bundle)
                nodes = self._create_citation_nodes(nodes)

                retrieve_event.on_end(payload={EventPayload.NODES: nodes})

            response = self._response_synthesizer.synthesize(
                query=query_bundle,
                nodes=nodes,
            )

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

        return response

    async def _aquery(self, query_bundle: QueryBundle) -> RESPONSE_TYPE:
        """Answer a query."""
        with self.callback_manager.event(
            CBEventType.QUERY, payload={EventPayload.QUERY_STR: query_bundle.query_str}
        ) as query_event:
            with self.callback_manager.event(
                CBEventType.RETRIEVE,
                payload={EventPayload.QUERY_STR: query_bundle.query_str},
            ) as retrieve_event:
                nodes = await self.aretrieve(query_bundle)
                nodes = self._create_citation_nodes(nodes)

                retrieve_event.on_end(payload={EventPayload.NODES: nodes})

            response = await self._response_synthesizer.asynthesize(
                query=query_bundle,
                nodes=nodes,
            )

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

        return response

retriever property

retriever: BaseRetriever

Get the retriever object.

from_args classmethod

from_args(index: BaseGPTIndex, llm: Optional[LLM] = None, response_synthesizer: Optional[BaseSynthesizer] = None, citation_chunk_size: int = DEFAULT_CITATION_CHUNK_SIZE, citation_chunk_overlap: int = DEFAULT_CITATION_CHUNK_OVERLAP, text_splitter: Optional[TextSplitter] = None, citation_qa_template: BasePromptTemplate = CITATION_QA_TEMPLATE, citation_refine_template: BasePromptTemplate = CITATION_REFINE_TEMPLATE, retriever: Optional[BaseRetriever] = None, node_postprocessors: Optional[List[BaseNodePostprocessor]] = None, response_mode: ResponseMode = COMPACT, use_async: bool = False, streaming: bool = False, metadata_mode: MetadataMode = NONE, **kwargs: Any) -> CitationQueryEngine

Initialize a CitationQueryEngine object.".

Parameters:

Name	Type	Description	Default
index	BaseGPTIndex	(BastGPTIndex): index to use for querying	required
llm	Optional[LLM]	(Optional[LLM]): LLM object to use for response generation.	None
citation_chunk_size	int	Size of citation chunks, default=512. Useful for controlling granularity of sources.	DEFAULT_CITATION_CHUNK_SIZE
citation_chunk_overlap	int	Overlap of citation nodes, default=20.	DEFAULT_CITATION_CHUNK_OVERLAP
text_splitter	Optional[TextSplitter]	A text splitter for creating citation source nodes. Default is a SentenceSplitter.	None
citation_qa_template	BasePromptTemplate	Template for initial citation QA	CITATION_QA_TEMPLATE
citation_refine_template	BasePromptTemplate	Template for citation refinement.	CITATION_REFINE_TEMPLATE
retriever	BaseRetriever	A retriever object.	None
node_postprocessors	Optional[List[BaseNodePostprocessor]]	A list of node postprocessors.	None
verbose	bool	Whether to print out debug info.	required
response_mode	ResponseMode	A ResponseMode object.	COMPACT
use_async	bool	Whether to use async.	False
streaming	bool	Whether to use streaming.	False
optimizer	Optional[BaseTokenUsageOptimizer]	A BaseTokenUsageOptimizer object.	required

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

@classmethod
def from_args(
    cls,
    index: BaseGPTIndex,
    llm: Optional[LLM] = None,
    response_synthesizer: Optional[BaseSynthesizer] = None,
    citation_chunk_size: int = DEFAULT_CITATION_CHUNK_SIZE,
    citation_chunk_overlap: int = DEFAULT_CITATION_CHUNK_OVERLAP,
    text_splitter: Optional[TextSplitter] = None,
    citation_qa_template: BasePromptTemplate = CITATION_QA_TEMPLATE,
    citation_refine_template: BasePromptTemplate = CITATION_REFINE_TEMPLATE,
    retriever: Optional[BaseRetriever] = None,
    node_postprocessors: Optional[List[BaseNodePostprocessor]] = None,
    # response synthesizer args
    response_mode: ResponseMode = ResponseMode.COMPACT,
    use_async: bool = False,
    streaming: bool = False,
    # class-specific args
    metadata_mode: MetadataMode = MetadataMode.NONE,
    **kwargs: Any,
) -> "CitationQueryEngine":
    """Initialize a CitationQueryEngine object.".

    Args:
        index: (BastGPTIndex): index to use for querying
        llm: (Optional[LLM]): LLM object to use for response generation.
        citation_chunk_size (int):
            Size of citation chunks, default=512. Useful for controlling
            granularity of sources.
        citation_chunk_overlap (int): Overlap of citation nodes, default=20.
        text_splitter (Optional[TextSplitter]):
            A text splitter for creating citation source nodes. Default is
            a SentenceSplitter.
        citation_qa_template (BasePromptTemplate): Template for initial citation QA
        citation_refine_template (BasePromptTemplate):
            Template for citation refinement.
        retriever (BaseRetriever): A retriever object.
        node_postprocessors (Optional[List[BaseNodePostprocessor]]): A list of
            node postprocessors.
        verbose (bool): Whether to print out debug info.
        response_mode (ResponseMode): A ResponseMode object.
        use_async (bool): Whether to use async.
        streaming (bool): Whether to use streaming.
        optimizer (Optional[BaseTokenUsageOptimizer]): A BaseTokenUsageOptimizer
            object.

    """
    retriever = retriever or index.as_retriever(**kwargs)

    response_synthesizer = response_synthesizer or get_response_synthesizer(
        llm=llm,
        text_qa_template=citation_qa_template,
        refine_template=citation_refine_template,
        response_mode=response_mode,
        use_async=use_async,
        streaming=streaming,
    )

    return cls(
        retriever=retriever,
        llm=llm,
        response_synthesizer=response_synthesizer,
        callback_manager=Settings.callback_manager,
        citation_chunk_size=citation_chunk_size,
        citation_chunk_overlap=citation_chunk_overlap,
        text_splitter=text_splitter,
        node_postprocessors=node_postprocessors,
        metadata_mode=metadata_mode,
    )