Terms

A client is the basic object for connecting to the DashVector server-side, similar to a connection in a relational database. To establish a connection, you create a client object and provide an API key. You can then use the client object to perform collection operations, such as creating a collection or retrieving a list of collections.

A cluster is a billable resource management unit, similar to a database in a relational database system. DashVector offers different cluster product specifications to meet various business needs. You can manage clusters in the console. Management operations include creating, upgrading, and releasing a cluster. After a cluster is created, you can perform operations on its collections and partitions.

A collection is a group of documents of the same type, similar to a table in a relational database. Each collection must have a unique name, which you can use to retrieve a specific collection object. The collection object lets you perform document operations, such as inserting a doc or retrieving a doc, and partition operations, such as creating a partition.

A partition is a logical division of documents within a collection. If you specify a partition for a document operation, such as inserting a doc or retrieving a doc, the operation is scoped to that partition. Using partitions correctly can improve the performance of document operations.

A doc, or document, is the basic data unit in DashVector, similar to a row in a relational database. A doc has the following properties: an ID (primary key), a vector, and fields (key-value pairs). A doc is the basic input structure for the Insert a doc, Update a doc, and Insert or update a doc operations. It is also the output structure for the Retrieve a doc and Fetch a doc operations.

A field is a basic unit of a doc and is similar to a column in a relational database. Each doc can contain multiple fields.

A vector, also known as an embedding vector, is a set of multi-dimensional data created by an embedding model that extracts features from unstructured data. In DashVector, a vector is a basic data unit of a doc that describes the features of the unstructured data. For example, [0.1, 0.2, 0.3, 0.4] is a 4-dimension vector.

A sparse vector is a vector in which most elements are zero. In DashVector, you can use sparse vectors to represent information such as term frequency. For example, {1:0.4, 10000:0.6, 222222:0.8} is a sparse vector in which the elements at the 1st, 10,000th, and 222,222nd positions have non-zero values. These positions can represent three keywords, and their corresponding values represent the keyword weights. All other elements are zero.

Queries per second (QPS) is the maximum number of query requests that can be sent to the DashVector service API per second. A higher QPS value means that more requests can be handled in the same period. For example, a QPS of 5 means you can make 5 API calls in one second.

An API key is a secret key used to access Vector Retrieval Service (DashVector). DashVector uses API keys for call authentication and billing. API keys are managed through your Alibaba Cloud account. Each account can have a maximum of three active API keys at one time.

You must store and use your API keys securely. For more information, see Protect and correctly use API keys.