Docs

Torch Object Storage

User guide

Creating vector indexes

After creating a vector bucket, you can create vector indexes within it. These indexes allow you to perform similarity queries on the vector data. This topic explains how to create a vector index.

Notes

Each vector bucket can hold up to 100 vector indexes.

Steps

In the left navigation pane, click Vector bucket list - Beta. In the buckets list, click the vector bucket you want to manage.

On the Vector Index page, click Create Vector Indexes.

Set the vector index parameters and click Submit.

Parameter	Description
Vector Index Name	Enter a name for your vector index. After you create the index, you cannot change its name. The index name must: Be unique within each vector bucket. Be between 3 and 63 characters long. Consist only of lowercase letters (a-z), numbers (0-9), and hyphens (-). Start and end with a lowercase letter or number.
Dimension	The dimension value determines how many numerical values each vector will contain. All vectors added to the index must have the same number of values. Enter an integer between 1 and 4,096.
Similarity	The distance between the query vector and the stored vectors. euclidean: The Euclidean distance measures the straight-line distance between two points in space. It is useful for measuring numerical differences in space. cosine: The cosine distance measures the directional difference between two vectors. It is well-suited for calculating semantic similarity in high-dimensional spaces involving text and images.
Unfiltered metadata	Descriptive information for vector data that cannot be used as query filters. The following constraints apply: Metadata keys must be between 3 and 63 characters long. You can add up to 10 non-filterable metadata keys.

Related operations

After creating a vector index, you can view its basic information in the console. This includes the index name, Resource Unique Identification (TRN), and creation time.

Last updated: 2026.03.10 16:25:38