Manage databases with Milvus Manager - Vector Retrieval Service for Milvus

Milvus Manager provides visual management for databases and Collections, including creating databases and Collections, defining fields and indexes, inserting sample data, and performing vector searches.

Prerequisite

A Milvus instance is required. If you do not have one, see Create a Milvus instance.

Prerequisites

Go to the Milvus Manager page.
1. Log on to the Alibaba Cloud Vector Retrieval Service Milvus Edition console.
2. In the left navigation pane, click Instance List, and then select the target instance to open its details page.
3. At the top of the instance details page, click the Milvus Manager tab to connect to the Milvus instance. For more information, see Connect to a Milvus instance with Milvus Manager.
In the left navigation pane, click Databases to open the database management page.
The navigation tree on the left lets you browse and manage all databases and their Collections in the instance.

Database management

The Milvus Manager overview page displays instance information such as the number of databases and Collections, and allows you to create and delete databases.

Create a database

On the overview page, click Create Database. In the dialog box that appears, enter a database name and click Confirm.

Delete a database

The default database (default) cannot be deleted.
Before you delete a database, make sure that all Collections within it have been deleted.
You must manually enter the database name to confirm the deletion.

View database details

On the overview page, click a database name to open its details page and view Collection information.

Collection management

A Collection is the core unit in Milvus for storing vector data. Milvus Manager supports full lifecycle management of Collections.

Collection list

The database details page lists all Collections in the current database. The list includes the following information:

Parameter	Description
Name	The name of the Collection. You can search or filter by name.
Status	Loaded or not loaded.
Approximate Entity count	The total number of Entities in the Collection. Note When a Collection is not loaded, this count may include deleted Entities and is for reference only.
Description	A description of the Collection.
Alias	The configured alias.
Created At	The time when the Collection was created.
Actions	Release, Copy, Delete.

The top of the page provides the following action buttons:

+ Collection: Create a new Collection.
+ External Collection: Create an external collection (Alibaba Cloud DLF Paimon table).
Import File: Batch import data files.
Search by Name: Filter by Collection name.

Collection operations

Operation	Prerequisite	Description
Load	Indexes have been created for all vector fields.	Load the Collection into memory to enable queries.
Release	The Collection is in the loaded state.	Release the Collection from memory.
Delete	None	Permanently delete the Collection and all its data.
Rename	None	Rename the Collection.
Copy	None	Create a new Collection by copying the configuration of the current Collection.
Details	None	Go to the Collection details page to view schema, data, and other information.

Create a Collection

On the Collection list page, click + Collection. In the creation panel that appears, configure the following settings:

Basic information:

Name: The name of the Collection.
Description: A description of the Collection.

ID, metadata field, and vector field configuration:

Primary Key Type: Supports Int64 and VarChar.
Auto Generate ID: When enabled, the system automatically generates primary key values.
Field Name: A custom field name.
Field Type: Such as FloatVector and BinaryVector.
Dimensions: The number of vector dimensions must be specified for vector fields.

Advanced settings:

Consistency: Supports Bounded (bounded), Strong (strong), Session (session), and Eventually (eventual).
Enable Dynamic Schema: Allows inserting additional fields beyond the schema, which are stored as dynamic metadata in $meta.
Auto Load Collection After Creation: When enabled, the system automatically creates an AUTOINDEX for fields and loads the Collection.

Collection details page

Click a Collection name to open its details page. The page is organized into the following tabs:

Tab	Description
Schema	View and manage the field definitions and indexes of the Collection.
Vector Search	Perform vector similarity searches for debugging and verifying search results.
Data	View and manage entity data in the Collection. Supports importing files in CSV and JSON formats, with a data size limit of 2 MB.
Partitions	Manage the partitions of the Collection. Supports creating partitions and searching by name. The default partition cannot be deleted.
Segments	View the underlying segment information. Supports Compact and Flush operations.
Properties	View the configurable properties of the Collection. Supports page-based configuration and reset.

Schema tab

Displays the schema definition of the Collection and supports index management. The schema includes the following information:

Basic information: Name, description, Entity count, status, number of replicas, features, autoid, consistency, and created time.

Schema table:

Parameter	Description
Field Name	The name of the field, such as id and data.
Field Type	Such as Int64 and FloatVector(128).
Index	The index name, such as AUTOINDEX.
Index Parameters	The index configuration parameters, such as metric_type:L2.
Default Value	The default value of the field.
Description	A description of the field.
Actions	Operations such as creating or deleting indexes.

Vector search tab

Run vector similarity searches to debug and verify search results.

Data tab

View, edit, and manage entity data in the Collection.

Note

File imports in CSV and JSON formats are supported, with a maximum data size of 2 MB.

Partitions tab

Manage the partitions of a Collection, including creating partitions and searching by name.

Note

The default partition cannot be deleted.

Segments tab

Inspect and operate on the underlying segments of the Collection for advanced observability and performance tuning. The following operations are supported:

Compact: Perform segment compaction on the Collection to optimize storage and query performance.
Flush: Persist in-memory data to disk.

Properties tab

Displays the configurable properties of the Collection. If a property value is not shown, it has not been explicitly configured. You can configure and reset properties on this page.

External Collection management

External Collections allow you to link DLF lake table data through external references, enabling efficient vector queries on DLF data without duplicating raw data. For detailed instructions, see Manage external Collections.