from __future__ import annotations import contextlib import enum import logging import uuid from typing import ( Any, Callable, Dict, Generator, Iterable, List, Optional, Tuple, Type, Union, ) import numpy as np import sqlalchemy from sqlalchemy import delete, func from sqlalchemy.dialects.postgresql import JSON, UUID from sqlalchemy.exc import ProgrammingError from sqlalchemy.orm import Session from sqlalchemy.sql import quoted_name from langchain_community.vectorstores.utils import maximal_marginal_relevance try: from sqlalchemy.orm import declarative_base except ImportError: from sqlalchemy.ext.declarative import declarative_base from langchain_core.documents import Document from langchain_core.embeddings import Embeddings from langchain_core.utils import get_from_dict_or_env from langchain_core.vectorstores import VectorStore ADA_TOKEN_COUNT = 1536 _LANGCHAIN_DEFAULT_COLLECTION_NAME = "langchain" def _results_to_docs(docs_and_scores: Any) -> List[Document]: """Return docs from docs and scores.""" return [doc for doc, _ in docs_and_scores] class BaseEmbeddingStore: """Embedding store.""" def get_embedding_store( distance_strategy: DistanceStrategy, collection_name: str ) -> Any: embedding_type = None if distance_strategy == DistanceStrategy.HAMMING: embedding_type = sqlalchemy.INTEGER # type: ignore else: embedding_type = sqlalchemy.REAL # type: ignore DynamicBase = declarative_base(class_registry=dict()) # type: Any class EmbeddingStore(DynamicBase, BaseEmbeddingStore): __tablename__ = collection_name uuid = sqlalchemy.Column( UUID(as_uuid=True), primary_key=True, default=uuid.uuid4 ) __table_args__ = {"extend_existing": True} document = sqlalchemy.Column(sqlalchemy.String, nullable=True) cmetadata = sqlalchemy.Column(JSON, nullable=True) # custom_id : any user defined id custom_id = sqlalchemy.Column(sqlalchemy.String, nullable=True) embedding = sqlalchemy.Column(sqlalchemy.ARRAY(embedding_type)) # type: ignore return EmbeddingStore class QueryResult: """Result from a query.""" EmbeddingStore: BaseEmbeddingStore distance: float class DistanceStrategy(str, enum.Enum): """Enumerator of the Distance strategies.""" EUCLIDEAN = "l2sq" COSINE = "cosine" HAMMING = "hamming" DEFAULT_DISTANCE_STRATEGY = DistanceStrategy.COSINE class Lantern(VectorStore): """`Postgres` with the `lantern` extension as a vector store. lantern uses sequential scan by default. but you can create a HNSW index using the create_hnsw_index method. - `connection_string` is a postgres connection string. - `embedding_function` any embedding function implementing `langchain.embeddings.base.Embeddings` interface. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is the name of the table in which embedding data will be stored The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `distance_strategy` is the distance strategy to use. (default: EUCLIDEAN) - `EUCLIDEAN` is the euclidean distance. - `COSINE` is the cosine distance. - `HAMMING` is the hamming distance. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. """ def __init__( self, connection_string: str, embedding_function: Embeddings, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, collection_metadata: Optional[dict] = None, pre_delete_collection: bool = False, logger: Optional[logging.Logger] = None, relevance_score_fn: Optional[Callable[[float], float]] = None, ) -> None: self.connection_string = connection_string self.embedding_function = embedding_function self.collection_name = collection_name self.collection_metadata = collection_metadata self._distance_strategy = distance_strategy self.pre_delete_collection = pre_delete_collection self.logger = logger or logging.getLogger(__name__) self.override_relevance_score_fn = relevance_score_fn self.EmbeddingStore = get_embedding_store( self.distance_strategy, collection_name ) self.__post_init__() def __post_init__( self, ) -> None: self._conn = self.connect() self.create_hnsw_extension() self.create_collection() @property def distance_strategy(self) -> DistanceStrategy: if isinstance(self._distance_strategy, DistanceStrategy): return self._distance_strategy if self._distance_strategy == DistanceStrategy.EUCLIDEAN.value: return DistanceStrategy.EUCLIDEAN elif self._distance_strategy == DistanceStrategy.COSINE.value: return DistanceStrategy.COSINE elif self._distance_strategy == DistanceStrategy.HAMMING.value: return DistanceStrategy.HAMMING else: raise ValueError( f"Got unexpected value for distance: {self._distance_strategy}. " f"Should be one of {', '.join([ds.value for ds in DistanceStrategy])}." ) @property def embeddings(self) -> Embeddings: return self.embedding_function @classmethod def connection_string_from_db_params( cls, driver: str, host: str, port: int, database: str, user: str, password: str, ) -> str: """Return connection string from database parameters.""" return f"postgresql+{driver}://{user}:{password}@{host}:{port}/{database}" def connect(self) -> sqlalchemy.engine.Connection: engine = sqlalchemy.create_engine(self.connection_string) conn = engine.connect() return conn @property def distance_function(self) -> Any: if self.distance_strategy == DistanceStrategy.EUCLIDEAN: return "l2sq_dist" elif self.distance_strategy == DistanceStrategy.COSINE: return "cos_dist" elif self.distance_strategy == DistanceStrategy.HAMMING: return "hamming_dist" def create_hnsw_extension(self) -> None: try: with Session(self._conn) as session: statement = sqlalchemy.text("CREATE EXTENSION IF NOT EXISTS lantern") session.execute(statement) session.commit() except Exception as e: self.logger.exception(e) def create_tables_if_not_exists(self) -> None: try: self.create_collection() except ProgrammingError: pass def drop_table(self) -> None: try: self.EmbeddingStore.__table__.drop(self._conn.engine) except ProgrammingError: pass def drop_tables(self) -> None: self.drop_table() def _hamming_relevance_score_fn(self, distance: float) -> float: return distance def _select_relevance_score_fn(self) -> Callable[[float], float]: """ The 'correct' relevance function may differ depending on a few things, including: - the distance / similarity metric used by the VectorStore - the scale of your embeddings (OpenAI's are unit normed. Many others are not!) - embedding dimensionality - etc. """ if self.override_relevance_score_fn is not None: return self.override_relevance_score_fn # Default strategy is to rely on distance strategy provided # in vectorstore constructor if self.distance_strategy == DistanceStrategy.COSINE: return self._cosine_relevance_score_fn elif self.distance_strategy == DistanceStrategy.EUCLIDEAN: return self._euclidean_relevance_score_fn elif self.distance_strategy == DistanceStrategy.HAMMING: return self._hamming_relevance_score_fn else: raise ValueError( "No supported normalization function" f" for distance_strategy of {self._distance_strategy}." "Consider providing relevance_score_fn to Lantern constructor." ) def _get_op_class(self) -> str: if self.distance_strategy == DistanceStrategy.COSINE: return "dist_cos_ops" elif self.distance_strategy == DistanceStrategy.EUCLIDEAN: return "dist_l2sq_ops" elif self.distance_strategy == DistanceStrategy.HAMMING: return "dist_hamming_ops" else: raise ValueError( "No supported operator class" f" for distance_strategy of {self._distance_strategy}." ) def _get_operator(self) -> str: if self.distance_strategy == DistanceStrategy.COSINE: return "<=>" elif self.distance_strategy == DistanceStrategy.EUCLIDEAN: return "<->" elif self.distance_strategy == DistanceStrategy.HAMMING: return "<+>" else: raise ValueError( "No supported operator" f" for distance_strategy of {self._distance_strategy}." ) def _typed_arg_for_distance( self, embedding: List[Union[float, int]] ) -> List[Union[float, int]]: if self.distance_strategy == DistanceStrategy.HAMMING: return list(map(lambda x: int(x), embedding)) return embedding @property def _index_name(self) -> str: return f"langchain_{self.collection_name}_idx" def create_hnsw_index( self, dims: int = ADA_TOKEN_COUNT, m: int = 16, ef_construction: int = 64, ef_search: int = 64, **_kwargs: Any, ) -> None: """Create HNSW index on collection. Optional Keyword Args for HNSW Index: engine: "nmslib", "faiss", "lucene"; default: "nmslib" ef: Size of the dynamic list used during k-NN searches. Higher values lead to more accurate but slower searches; default: 64 ef_construction: Size of the dynamic list used during k-NN graph creation. Higher values lead to more accurate graph but slower indexing speed; default: 64 m: Number of bidirectional links created for each new element. Large impact on memory consumption. Between 2 and 100; default: 16 dims: Dimensions of the vectors in collection. default: 1536 """ create_index_query = sqlalchemy.text( "CREATE INDEX IF NOT EXISTS {} " "ON {} USING hnsw (embedding {}) " "WITH (" "dim = :dim, " "m = :m, " "ef_construction = :ef_construction, " "ef = :ef" ");".format( quoted_name(self._index_name, True), quoted_name(self.collection_name, True), self._get_op_class(), ) ) with Session(self._conn) as session: # Create the HNSW index session.execute( create_index_query, { "dim": dims, "m": m, "ef_construction": ef_construction, "ef": ef_search, }, ) session.commit() self.logger.info("HNSW extension and index created successfully.") def drop_index(self) -> None: with Session(self._conn) as session: # Drop the HNSW index session.execute( sqlalchemy.text( "DROP INDEX IF EXISTS {}".format( quoted_name(self._index_name, True) ) ) ) session.commit() def create_collection(self) -> None: if self.pre_delete_collection: self.delete_collection() self.drop_table() with self._conn.begin(): try: self.EmbeddingStore.__table__.create(self._conn.engine) except ProgrammingError as e: # Duplicate table if e.code == "f405": pass else: raise e def delete_collection(self) -> None: self.logger.debug("Trying to delete collection") self.drop_table() @contextlib.contextmanager def _make_session(self) -> Generator[Session, None, None]: """Create a context manager for the session, bind to _conn string.""" yield Session(self._conn) def delete( self, ids: Optional[List[str]] = None, **kwargs: Any, ) -> None: """Delete vectors by ids or uuids. Args: ids: List of ids to delete. """ with Session(self._conn) as session: if ids is not None: self.logger.debug( "Trying to delete vectors by ids (represented by the model " "using the custom ids field)" ) stmt = delete(self.EmbeddingStore).where( self.EmbeddingStore.custom_id.in_(ids) ) session.execute(stmt) session.commit() @classmethod def _initialize_from_embeddings( cls, texts: List[str], embeddings: List[List[float]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, pre_delete_collection: bool = False, **kwargs: Any, ) -> Lantern: """ Order of elements for lists `ids`, `embeddings`, `texts`, `metadatas` should match, so each row will be associated with correct values. Postgres connection string is required "Either pass it as `connection_string` parameter or set the LANTERN_CONNECTION_STRING environment variable. - `texts` texts to insert into collection. - `embeddings` an Embeddings to insert into collection - `embedding` is :class:`Embeddings` that will be used for embedding the text sent. If none is sent, then the multilingual Tensorflow Universal Sentence Encoder will be used. - `metadatas` row metadata to insert into collection. - `ids` row ids to insert into collection. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is the name of the table in which embedding data will be stored The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `distance_strategy` is the distance strategy to use. (default: EUCLIDEAN) - `EUCLIDEAN` is the euclidean distance. - `COSINE` is the cosine distance. - `HAMMING` is the hamming distance. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. """ if ids is None: ids = [str(uuid.uuid1()) for _ in texts] if not metadatas: metadatas = [{} for _ in texts] connection_string = cls.__get_connection_string(kwargs) store = cls( connection_string=connection_string, collection_name=collection_name, embedding_function=embedding, pre_delete_collection=pre_delete_collection, distance_strategy=distance_strategy, ) store.add_embeddings( texts=texts, embeddings=embeddings, metadatas=metadatas, ids=ids, **kwargs ) store.create_hnsw_index(**kwargs) return store def add_embeddings( self, texts: List[str], embeddings: List[List[float]], metadatas: List[dict], ids: List[str], **kwargs: Any, ) -> None: with Session(self._conn) as session: for text, metadata, embedding, id in zip(texts, metadatas, embeddings, ids): embedding_store = self.EmbeddingStore( embedding=embedding, document=text, cmetadata=metadata, custom_id=id, ) session.add(embedding_store) session.commit() def add_texts( self, texts: Iterable[str], metadatas: Optional[List[dict]] = None, ids: Optional[List[str]] = None, **kwargs: Any, ) -> List[str]: if ids is None: ids = [str(uuid.uuid4()) for _ in texts] embeddings = self.embedding_function.embed_documents(list(texts)) if not metadatas: metadatas = [{} for _ in texts] with Session(self._conn) as session: for text, metadata, embedding, id in zip(texts, metadatas, embeddings, ids): embedding_store = self.EmbeddingStore( embedding=embedding, document=text, cmetadata=metadata, custom_id=id, ) session.add(embedding_store) session.commit() return ids def _results_to_docs_and_scores(self, results: Any) -> List[Tuple[Document, float]]: """Return docs and scores from results.""" docs = [ ( Document( page_content=result.EmbeddingStore.document, metadata=result.EmbeddingStore.cmetadata, ), result.distance if self.embedding_function is not None else None, ) for result in results ] return docs def similarity_search( self, query: str, k: int = 4, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: embedding = self.embedding_function.embed_query(text=query) return self.similarity_search_by_vector( embedding=embedding, k=k, filter=filter, ) def similarity_search_with_score( self, query: str, k: int = 4, filter: Optional[dict] = None, ) -> List[Tuple[Document, float]]: embedding = self.embedding_function.embed_query(query) docs = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, filter=filter ) return docs def similarity_search_with_score_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, ) -> List[Tuple[Document, float]]: results = self.__query_collection(embedding=embedding, k=k, filter=filter) return self._results_to_docs_and_scores(results) def __query_collection( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, ) -> List[Any]: with Session(self._conn) as session: set_enable_seqscan_stmt = sqlalchemy.text("SET enable_seqscan = off") set_init_k = sqlalchemy.text("SET hnsw.init_k = :k") session.execute(set_enable_seqscan_stmt) session.execute(set_init_k, {"k": k}) filter_by = None if filter is not None: filter_clauses = [] for key, value in filter.items(): IN = "in" if isinstance(value, dict) and IN in map(str.lower, value): value_case_insensitive = { k.lower(): v for k, v in value.items() } filter_by_metadata = self.EmbeddingStore.cmetadata[ key ].astext.in_(value_case_insensitive[IN]) filter_clauses.append(filter_by_metadata) else: filter_by_metadata = self.EmbeddingStore.cmetadata[ key ].astext == str(value) filter_clauses.append(filter_by_metadata) filter_by = sqlalchemy.and_(*filter_clauses) embedding = self._typed_arg_for_distance(embedding) query = session.query( self.EmbeddingStore, getattr(func, self.distance_function)( self.EmbeddingStore.embedding, embedding ).label("distance"), ) # Specify the columns you need here, e.g., EmbeddingStore.embedding if filter_by is not None: query = query.filter(filter_by) results: List[QueryResult] = ( query.order_by( self.EmbeddingStore.embedding.op(self._get_operator())(embedding) ) # Using PostgreSQL specific operator with the correct column name .limit(k) .all() ) return results def similarity_search_by_vector( self, embedding: List[float], k: int = 4, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Document]: docs_and_scores = self.similarity_search_with_score_by_vector( embedding=embedding, k=k, filter=filter ) return _results_to_docs(docs_and_scores) @classmethod def from_texts( cls: Type[Lantern], texts: List[str], embedding: Embeddings, metadatas: Optional[List[dict]] = None, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, ids: Optional[List[str]] = None, pre_delete_collection: bool = False, **kwargs: Any, ) -> Lantern: """ Initialize Lantern vectorstore from list of texts. The embeddings will be generated using `embedding` class provided. Order of elements for lists `ids`, `texts`, `metadatas` should match, so each row will be associated with correct values. Postgres connection string is required "Either pass it as `connection_string` parameter or set the LANTERN_CONNECTION_STRING environment variable. - `connection_string` is fully populated connection string for postgres database - `texts` texts to insert into collection. - `embedding` is :class:`Embeddings` that will be used for embedding the text sent. If none is sent, then the multilingual Tensorflow Universal Sentence Encoder will be used. - `metadatas` row metadata to insert into collection. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is the name of the table in which embedding data will be stored The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `distance_strategy` is the distance strategy to use. (default: EUCLIDEAN) - `EUCLIDEAN` is the euclidean distance. - `COSINE` is the cosine distance. - `HAMMING` is the hamming distance. - `ids` row ids to insert into collection. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. """ embeddings = embedding.embed_documents(list(texts)) return cls._initialize_from_embeddings( texts, embeddings, embedding, metadatas=metadatas, ids=ids, collection_name=collection_name, pre_delete_collection=pre_delete_collection, distance_strategy=distance_strategy, **kwargs, ) @classmethod def from_embeddings( cls, text_embeddings: List[Tuple[str, List[float]]], embedding: Embeddings, metadatas: Optional[List[dict]] = None, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, ids: Optional[List[str]] = None, pre_delete_collection: bool = False, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, **kwargs: Any, ) -> Lantern: """Construct Lantern wrapper from raw documents and pre- generated embeddings. Postgres connection string is required "Either pass it as `connection_string` parameter or set the LANTERN_CONNECTION_STRING environment variable. Order of elements for lists `ids`, `text_embeddings`, `metadatas` should match, so each row will be associated with correct values. - `connection_string` is fully populated connection string for postgres database - `text_embeddings` is array with tuples (text, embedding) to insert into collection. - `embedding` is :class:`Embeddings` that will be used for embedding the text sent. If none is sent, then the multilingual Tensorflow Universal Sentence Encoder will be used. - `metadatas` row metadata to insert into collection. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is the name of the table in which embedding data will be stored The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `ids` row ids to insert into collection. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. - `distance_strategy` is the distance strategy to use. (default: EUCLIDEAN) - `EUCLIDEAN` is the euclidean distance. - `COSINE` is the cosine distance. - `HAMMING` is the hamming distance. """ texts = [t[0] for t in text_embeddings] embeddings = [t[1] for t in text_embeddings] return cls._initialize_from_embeddings( texts, embeddings, embedding, metadatas=metadatas, ids=ids, collection_name=collection_name, pre_delete_collection=pre_delete_collection, distance_strategy=distance_strategy, **kwargs, ) @classmethod def from_existing_index( cls: Type[Lantern], embedding: Embeddings, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, pre_delete_collection: bool = False, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, **kwargs: Any, ) -> Lantern: """ Get instance of an existing Lantern store.This method will return the instance of the store without inserting any new embeddings Postgres connection string is required "Either pass it as `connection_string` parameter or set the LANTERN_CONNECTION_STRING environment variable. - `connection_string` is a postgres connection string. - `embedding` is :class:`Embeddings` that will be used for embedding the text sent. If none is sent, then the multilingual Tensorflow Universal Sentence Encoder will be used. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is the name of the table in which embedding data will be stored The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `ids` row ids to insert into collection. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. - `distance_strategy` is the distance strategy to use. (default: EUCLIDEAN) - `EUCLIDEAN` is the euclidean distance. - `COSINE` is the cosine distance. - `HAMMING` is the hamming distance. """ connection_string = cls.__get_connection_string(kwargs) store = cls( connection_string=connection_string, collection_name=collection_name, embedding_function=embedding, pre_delete_collection=pre_delete_collection, distance_strategy=distance_strategy, ) return store @classmethod def __get_connection_string(cls, kwargs: Dict[str, Any]) -> str: connection_string: str = get_from_dict_or_env( data=kwargs, key="connection_string", env_key="LANTERN_CONNECTION_STRING", ) if not connection_string: raise ValueError( "Postgres connection string is required" "Either pass it as `connection_string` parameter" "or set the LANTERN_CONNECTION_STRING variable." ) return connection_string @classmethod def from_documents( cls: Type[Lantern], documents: List[Document], embedding: Embeddings, collection_name: str = _LANGCHAIN_DEFAULT_COLLECTION_NAME, distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY, ids: Optional[List[str]] = None, pre_delete_collection: bool = False, **kwargs: Any, ) -> Lantern: """ Initialize a vector store with a set of documents. Postgres connection string is required "Either pass it as `connection_string` parameter or set the LANTERN_CONNECTION_STRING environment variable. - `connection_string` is a postgres connection string. - `documents` is list of :class:`Document` to initialize the vector store with - `embedding` is :class:`Embeddings` that will be used for embedding the text sent. If none is sent, then the multilingual Tensorflow Universal Sentence Encoder will be used. - `collection_name` is the name of the collection to use. (default: langchain) - NOTE: This is the name of the table in which embedding data will be stored The table will be created when initializing the store (if not exists) So, make sure the user has the right permissions to create tables. - `distance_strategy` is the distance strategy to use. (default: EUCLIDEAN) - `EUCLIDEAN` is the euclidean distance. - `COSINE` is the cosine distance. - `HAMMING` is the hamming distance. - `ids` row ids to insert into collection. - `pre_delete_collection` if True, will delete the collection if it exists. (default: False) - Useful for testing. """ texts = [d.page_content for d in documents] metadatas = [d.metadata for d in documents] connection_string = cls.__get_connection_string(kwargs) kwargs["connection_string"] = connection_string return cls.from_texts( texts=texts, pre_delete_collection=pre_delete_collection, embedding=embedding, metadatas=metadatas, ids=ids, collection_name=collection_name, distance_strategy=distance_strategy, **kwargs, ) def max_marginal_relevance_search_with_score_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs selected using the maximal marginal relevance with score to embedding vector. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. fetch_k (int): Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult (float): Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Tuple[Document, float]]: List of Documents selected by maximal marginal relevance to the query and score for each. """ results = self.__query_collection(embedding=embedding, k=fetch_k, filter=filter) embedding_list = [result.EmbeddingStore.embedding for result in results] mmr_selected = maximal_marginal_relevance( np.array(embedding, dtype=np.float32), embedding_list, k=k, lambda_mult=lambda_mult, ) candidates = self._results_to_docs_and_scores(results) return [r for i, r in enumerate(candidates) if i in mmr_selected] def max_marginal_relevance_search( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query (str): Text to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. fetch_k (int): Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult (float): Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Document]: List of Documents selected by maximal marginal relevance. """ embedding = self.embedding_function.embed_query(query) return self.max_marginal_relevance_search_by_vector( embedding, k=k, fetch_k=fetch_k, lambda_mult=lambda_mult, filter=filter, **kwargs, ) def max_marginal_relevance_search_with_score( self, query: str, k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[dict] = None, **kwargs: Any, ) -> List[Tuple[Document, float]]: """Return docs selected using the maximal marginal relevance with score. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query (str): Text to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. fetch_k (int): Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult (float): Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Tuple[Document, float]]: List of Documents selected by maximal marginal relevance to the query and score for each. """ embedding = self.embedding_function.embed_query(query) docs = self.max_marginal_relevance_search_with_score_by_vector( embedding=embedding, k=k, fetch_k=fetch_k, lambda_mult=lambda_mult, filter=filter, **kwargs, ) return docs def max_marginal_relevance_search_by_vector( self, embedding: List[float], k: int = 4, fetch_k: int = 20, lambda_mult: float = 0.5, filter: Optional[Dict[str, str]] = None, **kwargs: Any, ) -> List[Document]: """Return docs selected using the maximal marginal relevance to embedding vector. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding (str): Text to look up documents similar to. k (int): Number of Documents to return. Defaults to 4. fetch_k (int): Number of Documents to fetch to pass to MMR algorithm. Defaults to 20. lambda_mult (float): Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Document]: List of Documents selected by maximal marginal relevance. """ docs_and_scores = self.max_marginal_relevance_search_with_score_by_vector( embedding, k=k, fetch_k=fetch_k, lambda_mult=lambda_mult, filter=filter, **kwargs, ) return _results_to_docs(docs_and_scores)