Skip to content

LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows

3.0 KB
Raw Blame
import logging

from executor.executors.exceptions import RetrievalError
from executor.executors.retrievers.base_retriever import BaseRetriever
from llama_index.core import VectorStoreIndex
from llama_index.core.retrievers import RecursiveRetriever
from llama_index.core.vector_stores import ExactMatchFilter, MetadataFilters

logger = logging.getLogger(__name__)


class RecursiveRetrieval(BaseRetriever):
    """Recursive retrieval using LlamaIndex's native RecursiveRetriever.

    This retriever performs recursive retrieval by breaking down queries
    and refining results through multiple retrieval steps.
    """

    def retrieve(self) -> set[str]:
        """Retrieve text chunks using LlamaIndex's native RecursiveRetriever.

        Returns:
            set[str]: A set of text chunks retrieved from the database.
        """
        try:
            logger.info(
                f"Retrieving chunks for {self.doc_id} using LlamaIndex RecursiveRetriever."
            )

            # Get the vector store index
            vector_store_index: VectorStoreIndex = self.vector_db.get_vector_store_index()

            # Create base retriever with metadata filters
            base_retriever = vector_store_index.as_retriever(
                similarity_top_k=self.top_k,
                filters=MetadataFilters(
                    filters=[
                        ExactMatchFilter(key="doc_id", value=self.doc_id),
                    ],
                ),
            )

            # Create RecursiveRetriever
            recursive_retriever = RecursiveRetriever(
                "vector",  # root retriever key
                retriever_dict={"vector": base_retriever},
                verbose=True,
            )

            # Retrieve nodes using RecursiveRetriever
            nodes = recursive_retriever.retrieve(self.prompt)

            # Extract unique text chunks
            chunks: set[str] = set()
            for node in nodes:
                if node.score > 0:
                    chunks.add(node.get_content())
                else:
                    logger.info(
                        f"Node score is less than 0. "
                        f"Ignored: {node.node_id} with score {node.score}"
                    )

            logger.info(
                f"Successfully retrieved {len(chunks)} chunks using RecursiveRetriever."
            )
            return chunks

        except (ValueError, AttributeError, KeyError, ImportError) as e:
            logger.error(
                "Error during recursive retrieval for %s: %s: %s",
                self.doc_id,
                type(e).__name__,
                e,
                exc_info=True,
            )
            raise RetrievalError(f"{type(e).__name__}: {e}") from e
        except Exception as e:
            logger.error(
                "Unexpected error during recursive retrieval for %s: %s: %s",
                self.doc_id,
                type(e).__name__,
                e,
                exc_info=True,
            )
            raise RetrievalError(f"Unexpected error: {type(e).__name__}: {e}") from e