Dashvector

继承自: BaseReader

DashVector 阅读器。

参数

名称

类型	描述	默认值	api_key
str	DashVector API 密钥。	必需	endpoint
DashVector 集群端点。	DashVector API 密钥。	源代码位于 llama-index-integrations/readers/llama-index-readers-dashvector/llama_index/readers/dashvector/base.py	endpoint

load_data

class DashVectorReader(BaseReader):
    """
    DashVector reader.

    Args:
        api_key (str): DashVector API key.
        endpoint (str): DashVector cluster endpoint.

    """

    def __init__(self, api_key: str, endpoint: str):
        """Initialize with parameters."""
        try:
            import dashvector
        except ImportError:
            raise ImportError(
                "`dashvector` package not found, please run `pip install dashvector`"
            )

        self._client: dashvector.Client = dashvector.Client(
            api_key=api_key, endpoint=endpoint
        )

    def load_data(
        self,
        collection_name: str,
        vector: Optional[List[float]],
        topk: int,
        filter: Optional[str] = None,
        include_vector: bool = True,
        partition: Optional[str] = None,
        output_fields: Optional[List[str]] = None,
        sparse_vector: Optional[Dict[int, float]] = None,
    ) -> List[Document]:
        """
        Load data from DashVector.

        Args:
            collection_name (str): Name of the collection.
            vector (List[float]): Query vector.
            topk (int): Number of results to return.
            filter (Optional[str]): doc fields filter
                conditions that meet the SQL where clause specification.detail in https://help.aliyun.com/document_detail/2513006.html?spm=a2c4g.2510250.0.0.40d25637QMI4eV
            include_vector (bool): Whether to include the embedding in the response.Defaults to True.
            partition (Optional[str]): The partition name
                to query. Defaults to None.
            output_fields (Optional[List[str]]): The fields
                to return. Defaults to None, meaning all fields
            sparse_vector (Optional[Dict[int, float]]): The
                sparse vector to query.Defaults to None.

        Returns:
            List[Document]: A list of documents.

        """
        collection = self._client.get(collection_name)
        if not collection:
            raise ValueError(
                f"Failed to get collection: {collection_name}," f"Error: {collection}"
            )

        ret = collection.query(
            vector=vector,
            topk=topk,
            filter=filter,
            include_vector=include_vector,
            partition=partition,
            output_fields=output_fields,
            sparse_vector=sparse_vector,
        )
        if not ret:
            raise Exception(f"Failed to query document," f"Error: {ret}")

        doc_metas = ret.output
        documents = []

        for doc_meta in doc_metas:
            node_content = json.loads(doc_meta.fields["_node_content"])
            document = Document(
                id_=doc_meta.id,
                text=node_content["text"],
                metadata=node_content["metadata"],
                embedding=doc_meta.vector,
            )
            documents.append(document)

        return documents

从 DashVector 加载数据。

load_data(collection_name: str, vector: Optional[List[float]], topk: int, filter: Optional[str] = None, include_vector: bool = True, partition: Optional[str] = None, output_fields: Optional[List[str]] = None, sparse_vector: Optional[Dict[int, float]] = None) -> List[Document]

collection_name

名称

类型	描述	默认值	api_key
集合的名称。	DashVector API 密钥。	vector	endpoint
List[float]	查询向量。	topk	endpoint
int	返回结果的数量。	filter	endpoint
Optional[str]	满足 SQL where 子句规范的文档字段过滤条件。详情参见 https://help.aliyun.com/document_detail/2513006.html?spm=a2c4g.2510250.0.0.40d25637QMI4eV	include_vector	检查点工作流运行
bool	是否在响应中包含嵌入。默认为 True。	True	partition
要查询的分区名称。默认为 None。	满足 SQL where 子句规范的文档字段过滤条件。详情参见 https://help.aliyun.com/document_detail/2513006.html?spm=a2c4g.2510250.0.0.40d25637QMI4eV	output_fields	检查点工作流运行
Optional[List[str]]	要返回的字段。默认为 None，表示所有字段	sparse_vector	检查点工作流运行
Optional[Dict[int, float]]	要查询的稀疏向量。默认为 None。	返回值	检查点工作流运行

List[Document]

描述	默认值
List[Document]：文档列表。	返回顶部

load_data

def load_data(
    self,
    collection_name: str,
    vector: Optional[List[float]],
    topk: int,
    filter: Optional[str] = None,
    include_vector: bool = True,
    partition: Optional[str] = None,
    output_fields: Optional[List[str]] = None,
    sparse_vector: Optional[Dict[int, float]] = None,
) -> List[Document]:
    """
    Load data from DashVector.

    Args:
        collection_name (str): Name of the collection.
        vector (List[float]): Query vector.
        topk (int): Number of results to return.
        filter (Optional[str]): doc fields filter
            conditions that meet the SQL where clause specification.detail in https://help.aliyun.com/document_detail/2513006.html?spm=a2c4g.2510250.0.0.40d25637QMI4eV
        include_vector (bool): Whether to include the embedding in the response.Defaults to True.
        partition (Optional[str]): The partition name
            to query. Defaults to None.
        output_fields (Optional[List[str]]): The fields
            to return. Defaults to None, meaning all fields
        sparse_vector (Optional[Dict[int, float]]): The
            sparse vector to query.Defaults to None.

    Returns:
        List[Document]: A list of documents.

    """
    collection = self._client.get(collection_name)
    if not collection:
        raise ValueError(
            f"Failed to get collection: {collection_name}," f"Error: {collection}"
        )

    ret = collection.query(
        vector=vector,
        topk=topk,
        filter=filter,
        include_vector=include_vector,
        partition=partition,
        output_fields=output_fields,
        sparse_vector=sparse_vector,
    )
    if not ret:
        raise Exception(f"Failed to query document," f"Error: {ret}")

    doc_metas = ret.output
    documents = []

    for doc_meta in doc_metas:
        node_content = json.loads(doc_meta.fields["_node_content"])
        document = Document(
            id_=doc_meta.id,
            text=node_content["text"],
            metadata=node_content["metadata"],
            embedding=doc_meta.vector,
        )
        documents.append(document)

    return documents