similarity_search_with_score_by_vector() — langchain Function Reference

Source Code

libs/partners/qdrant/langchain_qdrant/vectorstores.py lines 541–627

    def similarity_search_with_score_by_vector(
        self,
        embedding: list[float],
        k: int = 4,
        filter: MetadataFilter | None = None,  # noqa: A002
        search_params: models.SearchParams | None = None,
        offset: int = 0,
        score_threshold: float | None = None,
        consistency: models.ReadConsistency | None = None,
        **kwargs: Any,
    ) -> list[tuple[Document, float]]:
        """Return docs most similar to embedding vector.

        Args:
            embedding: Embedding vector to look up documents similar to.
            k: Number of Documents to return.
            filter: Filter by metadata.
            search_params: Additional search params
            offset:
                Offset of the first result to return.
                May be used to paginate results.
                Note: large offset values may cause performance issues.
            score_threshold:
                Define a minimal score threshold for the result.
                If defined, less similar results will not be returned.
                Score of the returned result might be higher or smaller than the
                threshold depending on the Distance function used.
                E.g. for cosine similarity only higher scores will be returned.
            consistency:
                Read consistency of the search. Defines how many replicas should be
                queried before returning the result.
                Values:
                - int - number of replicas to query, values should present in all
                        queried replicas
                - 'majority' - query all replicas, but return values present in the
                    majority of replicas
                - 'quorum' - query the majority of replicas, return values present in
                    all of them
                - 'all' - query all replicas, and return values present in all replicas
            **kwargs:
                Any other named arguments to pass through to QdrantClient.search()

        Returns:
            List of documents most similar to the query text and distance for each.

        """
        if filter is not None and isinstance(filter, dict):
            warnings.warn(
                "Using dict as a `filter` is deprecated. Please use qdrant-client "
                "filters directly: "
                "https://qdrant.tech/documentation/concepts/filtering/",
                DeprecationWarning,
                stacklevel=2,
            )
            qdrant_filter = self._qdrant_filter_from_dict(filter)
        else:
            qdrant_filter = filter

        query_vector = embedding
        if self.vector_name is not None:
            query_vector = (self.vector_name, embedding)  # type: ignore[assignment]

        results = self.client.search(
            collection_name=self.collection_name,
            query_vector=query_vector,
            query_filter=qdrant_filter,
            search_params=search_params,
            limit=k,
            offset=offset,
            with_payload=True,
            with_vectors=False,  # LangChain does not expect vectors to be returned
            score_threshold=score_threshold,
            consistency=consistency,
            **kwargs,
        )
        return [
            (
                self._document_from_scored_point(
                    result,
                    self.collection_name,
                    self.content_payload_key,